2010-11-17T00:00:03 which group are you using for it? 2010-11-17T00:00:10 janzert: global moderators 2010-11-17T00:00:28 janzert: like.. if they have a say in here, they should have mod powers too 2010-11-17T00:00:29 Accoun: yes, with the radius at 17 2010-11-17T00:00:50 heh my bot is too conservative 2010-11-17T00:00:51 amstan: makes sense I suppose 2010-11-17T00:01:06 and this version even more so 2010-11-17T00:01:08 and loses a lot more 2010-11-17T00:01:09 :-\ 2010-11-17T00:01:24 McLeopold> pleaze use 12 like on official 2010-11-17T00:01:42 janzert: Are you willing to nail down the map specs yet? 2010-11-17T00:02:00 amstan: aikon 2010-11-17T00:02:02 ? 2010-11-17T00:02:12 @seen a1k0n 2010-11-17T00:02:12 amstan: a1k0n was last seen in #aichallenge 2 days, 22 hours, 12 minutes, and 6 seconds ago: right, they announced immediately the final tournament would be on bigger maps for tron 2010-11-17T00:02:20 k, decent 2010-11-17T00:02:31 McLeopold: hope to do some work on it tonight and get it up tomorrow 2010-11-17T00:02:48 So you're going to modify the code? 2010-11-17T00:03:07 McLeopold: the only functional change I know of is going to be the radius of the maps 2010-11-17T00:03:15 the other is bug fixing 2010-11-17T00:03:18 I see a bug in the current 2010-11-17T00:03:26 It's generating planets over 100 ships. 2010-11-17T00:03:30 right 2010-11-17T00:03:44 McLeopold: cool, thanks 2010-11-17T00:04:04 that's what I meant by bug fixing, none of the maps actually generated ended up with that problem did they? 2010-11-17T00:04:13 yes 2010-11-17T00:04:19 ahh, didn't realize that 2010-11-17T00:04:28 http://zeroviz.us:8080/canvas?game_id=1657 2010-11-17T00:04:34 It's one I generated 2010-11-17T00:04:45 151 +5 neutrals 2010-11-17T00:04:46 ahh, yeah I meant on the contest server 2010-11-17T00:04:59 but yes I do plan on fixing that 2010-11-17T00:05:46 amstan: how about bocsimacko/mega? 2010-11-17T00:06:03 janzert: we can add them as we go 2010-11-17T00:06:11 McLeopold: is this the bug: 5*distance(p1, p2)-1 ? 2010-11-17T00:06:14 map 30 has a 106 pair 2010-11-17T00:06:18 what about me? 2010-11-17T00:06:18 uh oh, user defined groups disappeared? 2010-11-17T00:06:39 mega1: we're moving away from the mailing list system 2010-11-17T00:06:40 make that "map_r528_30.txt" 2010-11-17T00:06:43 mega1: added for posting access and moderator on theforums 2010-11-17T00:06:54 Mjothitnir: ahh, ok 2010-11-17T00:07:00 mega1: we'll just have a forum, and we were wondering if we should give you posting access to that special forum 2010-11-17T00:07:14 honestly, I don't know why ship count should be anything other than a random number between 1 and 100 2010-11-17T00:08:09 McLeopold: hmm, yeah. antimatroid might be able to clarify that although I think may just be something left from the original generator 2010-11-17T00:08:48 :-o 2010-11-17T00:09:06 mega1: such people also become global moderators 2010-11-17T00:09:24 McLeopold: that's probably the rage bot thingy 2010-11-17T00:09:30 amstan: did all the user defined groups just disappear? They seem to have here 2010-11-17T00:09:33 janzert: I have my own map generator. I'll share it on bitbucket. 2010-11-17T00:09:38 janzert: yeah, removed them 2010-11-17T00:09:40 I think that can be pretty important. 2010-11-17T00:09:44 amstan: ok 2010-11-17T00:09:47 janzert: contest organizer vs admin was redundant 2010-11-17T00:09:51 and the others too 2010-11-17T00:10:03 mega1: yes, but the entire map should be checked after generation and kept or discarded. 2010-11-17T00:10:18 Otherwise you'll introduce weird bias 2010-11-17T00:10:22 McLeopold: I really only want to make the absolute minimum changes I can to fix the stuff that cropped up with this one if possible 2010-11-17T00:10:34 McLeopold: sure 2010-11-17T00:10:44 janzert: then revert it 2010-11-17T00:10:50 janzert: I got to say, alot of the generation stuff doesn't make sense to me. 2010-11-17T00:11:01 I think it's maps are not random enough 2010-11-17T00:11:07 the neutrals with more than 100 ships are from line 184 2010-11-17T00:11:10 well I'm the wrong one to ask about it :) 2010-11-17T00:11:16 the bug mega1 was talking aobut 2010-11-17T00:11:57 you want to min the number with 100 2010-11-17T00:12:08 no, that's not good enough 2010-11-17T00:12:20 why? 2010-11-17T00:12:26 It makes lots of planets at exatly 100. 2010-11-17T00:12:34 it only does it for 1 planet 2010-11-17T00:12:40 the non-snipable one 2010-11-17T00:12:40 pick a number between 1 and 10, then trim to no more than 2 2010-11-17T00:12:45 that is not a 50/50 split 2010-11-17T00:12:47 no sorry, can be snipable, 2010-11-17T00:12:52 claimable without losing home planet 2010-11-17T00:13:17 does one such planet in any way help to avoid draws? 2010-11-17T00:13:17 well that was a little hack put in for other people, feel free to work out a nicer solution :P 2010-11-17T00:13:36 mega1: only just, i think it'd work even better if it was non-snipable 2010-11-17T00:14:07 well, it reduces the set of obvious draw maps anyway, and doesn't really exclude any exciting ones 2010-11-17T00:14:19 although it does then assume all bots wont play stupidly 2010-11-17T00:14:19 you know guys, all this fuss about maps is a waste of time. the final contest will not be able to run but a handful of maps, and it will necessarily be biased in a random unknown way 2010-11-17T00:14:37 you need more than 5 maps 2010-11-17T00:14:41 i think 2010-11-17T00:14:53 McLeopold: it shouldn't introduce any skew just limiting it to 100 2010-11-17T00:14:53 krokkrok: I'm wasting my time precisely to control the way it's biased. 2010-11-17T00:15:13 chess tournaments are run on one map. 2010-11-17T00:15:23 that's different 2010-11-17T00:15:25 *** fmeyer has quit IRC (Ping timeout: 240 seconds) 2010-11-17T00:15:41 McLeopold: err, actually I guess it does 2010-11-17T00:15:45 yes it is different. for one thing, it is much easier to judge 2010-11-17T00:15:48 but not a lot :/ 2010-11-17T00:16:04 all future contest games should be picked just so it's easy to judge? 2010-11-17T00:16:08 doesn't sound overly exciting 2010-11-17T00:16:10 yes 2010-11-17T00:16:27 otherwise, the outcome is random 2010-11-17T00:16:33 *** TEAM_AI has joined #aichallenge 2010-11-17T00:16:38 hi 2010-11-17T00:16:39 it's not completely random 2010-11-17T00:16:43 there is a random factor yes 2010-11-17T00:16:57 can you calculate how large that random factor is? 2010-11-17T00:17:12 if you can't, you don't really know what it is you just judged 2010-11-17T00:17:25 can you do that for anything? 2010-11-17T00:17:31 of course 2010-11-17T00:17:37 actually it's easy to fix correctly http://pastebin.ca/1994156 2010-11-17T00:17:38 a foot race, nothing random at all 2010-11-17T00:17:38 let's trade a 100 little random factors that add up to near zero value to a single random value 100x the magnitude 2010-11-17T00:17:39 Top 10 players: bocsimacko(3641), bix0r4ever(3572), asavis(3514), protocolocon(3501), davidjliu(3470), dmj111(3435), shangas(3434), thedreamer(3433), _iouri_(3424), george(3417) 2010-11-17T00:17:46 there is randomness 2010-11-17T00:18:12 got a short request: does anybody know how long is the time-limit per round? (ms) 2010-11-17T00:18:21 *** delt0r has joined #aichallenge 2010-11-17T00:18:32 janzert: that doesn't fix it 2010-11-17T00:18:40 that's is bad 2010-11-17T00:18:40 *** delt0r___ has quit IRC (Ping timeout: 240 seconds) 2010-11-17T00:18:41 very bad 2010-11-17T00:18:50 you just changed the max ships for every neutral 2010-11-17T00:18:58 team_ai i think its 1 sec 2010-11-17T00:19:10 unless i'm mistaken? 2010-11-17T00:19:22 remove the dependency of population on the distance 2010-11-17T00:19:49 thx krokkrok, that's very little :D 2010-11-17T00:19:49 it's a poorly chosen variable name and completely untested yes 2010-11-17T00:20:07 mega1: that dependency is only done for 1 pair of planetes 2010-11-17T00:20:09 it can still be over 100 for starters 2010-11-17T00:20:15 we really need to get a closure on this map generator situation... 2010-11-17T00:20:27 it's fairly closed 2010-11-17T00:20:34 it's mostly aesthetics being worked out atm 2010-11-17T00:20:50 will each bot play on the exact same maps as all other bots in the final tournament? 2010-11-17T00:20:59 antimatroid: there is no other max_ships variable 2010-11-17T00:21:02 janzert: how's that? 2010-11-17T00:21:09 the constant is maxShips 2010-11-17T00:21:15 ah yeah, that'd be it, sorry :P 2010-11-17T00:21:19 i saw that and was like ahhh 2010-11-17T00:21:27 janzert: can you help me test it? click on people and there's a test user permissions and test if they can/cannot post in there 2010-11-17T00:21:30 like I said bad variable name choice though 2010-11-17T00:22:00 yep, gotta find it first 2010-11-17T00:22:24 antimatroid: where does 5 * the home planets distance come from? what's the logic? 2010-11-17T00:22:29 will each bot play on the exact same maps as all other bots in the final tournament? 2010-11-17T00:22:41 McLeopold: so that you can take it without losing your home planet to rage bot 2010-11-17T00:22:43 McLeopold: five is the growth of the home planet 2010-11-17T00:23:28 You cannot lose your base if you take this planet. 2010-11-17T00:23:30 oh! I know how to fix it 2010-11-17T00:23:38 happy coding @all & cya l8ter 2010-11-17T00:23:44 will each bot play on the exact same maps as all other bots in the final tournament? 2010-11-17T00:23:46 *** TEAM_AI has quit IRC () 2010-11-17T00:23:53 krokkrok: i doubt it 2010-11-17T00:23:53 amstan: hmm, it let me post after setting my user permissions to olgeorge 2010-11-17T00:24:00 what if there's some free-planets inbetween the 2 bases? 2010-11-17T00:24:03 krokkrok: not decided 2010-11-17T00:24:05 the..its a lottery? 2010-11-17T00:24:11 it's too expensive to do that 2010-11-17T00:24:16 janzert: how many posts does he have? is he a newly registered? 2010-11-17T00:24:20 you wont listen to how infeasible it is 2010-11-17T00:24:38 amstan: http://ai-contest.com/forum/memberlist.php?mode=viewprofile&u=1588 2010-11-17T00:24:41 antimatroid: why is it infeasible? 2010-11-17T00:24:51 janzert: ya.. he's newly registered.. 2010-11-17T00:24:55 i have to do that too then 2010-11-17T00:24:57 sigh 2010-11-17T00:24:57 mega1: he's asking for a round robin 2010-11-17T00:25:00 num_ships = random.randint(minShips, min(maxShips,5*distance(p1, p2))-1) 2010-11-17T00:25:16 antimatroid: he didn't 2010-11-17T00:25:19 don't cap the final result, cap the range 2010-11-17T00:25:37 janzert: try now 2010-11-17T00:25:38 that's what janzert did 2010-11-17T00:25:50 mega1: what do you think every bot playing every other bot on a map is? :P 2010-11-17T00:25:51 yeah, ok, next issue 2010-11-17T00:26:05 antimatroid: read it again 2010-11-17T00:26:13 what i asked for just now is that every bot gets to play on the same map as every other bot, otherwise, the outcome is randomly biased 2010-11-17T00:26:31 amstan: looks like it still works 2010-11-17T00:26:35 janzert: darn.. wtf? 2010-11-17T00:26:35 oh, okay, it's still random 2010-11-17T00:26:50 sorry :) 2010-11-17T00:26:59 janzert: any ideas? 2010-11-17T00:27:10 krokkrok: I hope it is like what you suggest and it's going to be random still because the player you get to play on map X is not the same as for other bots. 2010-11-17T00:27:24 mcleopold is right about that change too 2010-11-17T00:27:28 mega1: but it could be 2010-11-17T00:28:00 krokkrok: for an N player round robin that can be N*N*M matches 2010-11-17T00:28:15 i think idealy you want to play on the same map many times 2010-11-17T00:28:15 where M is the number of maps 2010-11-17T00:28:26 in chess, M is 1 2010-11-17T00:28:32 antimatroid: that's the best way to increase variance 2010-11-17T00:28:33 amstan: not really, I haven't done a lot with phpbb 2010-11-17T00:28:34 works for them 2010-11-17T00:28:58 gah, there's like 3 different conversations in here now :| 2010-11-17T00:29:11 also it doesnt have to be N*N, simply, 100*100 2010-11-17T00:29:17 *** Rubicon-|-Cross has quit IRC (Quit: Leaving) 2010-11-17T00:29:23 krokkrok: as I said above: "let's trade a 100 little random factors that add up to near zero value to a single random value 100x the magnitude" 2010-11-17T00:29:34 sarcasm ^^^ 2010-11-17T00:30:16 in other words you introduce a huge random value when choosing that simple map 2010-11-17T00:30:18 *single 2010-11-17T00:30:20 you'd still have the problem of choosing the 100 2010-11-17T00:30:37 if all players new what map it was, like in chess, nothing would be random at all 2010-11-17T00:30:49 krokkrok: different boring game 2010-11-17T00:30:52 that's true 2010-11-17T00:31:04 janzert: who did you assume the permissions of? 2010-11-17T00:31:24 amstan: olgeorge last guy to post in waterloo forum 2010-11-17T00:31:25 krokkrok: we want to prevent and opening book race 2010-11-17T00:31:31 why 2010-11-17T00:31:47 janzert: it doesnt' seem to work for me 2010-11-17T00:31:51 i don't see any post button 2010-11-17T00:32:17 too much effort and sweat, keeps the barrier to entry too high 2010-11-17T00:32:24 kills the fun, boring. 2010-11-17T00:32:49 and it's not how the contest was advertised. 2010-11-17T00:32:53 amstan: hmm, maybe I'm doing it the wrong way 2010-11-17T00:32:58 varying maps are way more interesting than solving a given map 2010-11-17T00:33:28 for sure 2010-11-17T00:33:30 yes 2010-11-17T00:33:33 a game that cannot be judged fairly is not a game anymore, its just a beauty pageant 2010-11-17T00:33:42 playing a "solved" game is not AI, its a table lookup 2010-11-17T00:33:43 janzert: you have to click test users permissions in their profile, then on top of the page next to "User Control Panel (0 new messages) • View your posts ]" there should be a "revert permissions" link too 2010-11-17T00:33:53 so all sports are beauty pagents? 2010-11-17T00:33:54 a single unknown map, isn't fair, for sure 2010-11-17T00:34:00 anitimatroid: indeed 2010-11-17T00:34:07 do you even follow sports??? 2010-11-17T00:34:12 krokkrok: I sympathize with that statement but you can still do measurements with some uncertainty 2010-11-17T00:34:17 depends on the sport 2010-11-17T00:34:25 i skateboard and play what you would probabyl call field hockey 2010-11-17T00:34:27 fairest would be if every pair of bots, played on a list of 100ish different maps 2010-11-17T00:34:47 burny: that's the N*N*M solution that infeasible 2010-11-17T00:34:52 that's 2010-11-17T00:34:57 can't type at all 2010-11-17T00:35:27 amstan: hmm, yeah I do that (click on the test users permissions) and it sends me to the information page with successfully transferred permissions... but before I didn't get the link for restore permission for some reason 2010-11-17T00:35:32 this time it worked 2010-11-17T00:35:36 the system that stands a chance is N*N on a random map 2010-11-17T00:35:37 what is feasible is use random match ELO to find an appromiation of the top 100 bots, then 100*100*m. 2010-11-17T00:35:43 janzert: now there should be no post button 2010-11-17T00:35:53 not to be picky but isn't that n(n-1)/2 2010-11-17T00:35:56 yep, it went away 2010-11-17T00:36:02 janzert: decent 2010-11-17T00:36:10 what are the chances of us getting a top x round robin right at the end amstan and janzert? 2010-11-17T00:36:10 janzert: now we have to transfer a bunch of discussions in there 2010-11-17T00:36:19 antimatroid: we'll probably do it 2010-11-17T00:36:23 antimatroid: we did it before 2010-11-17T00:36:28 Mjothitnir: yes, but that's a constant in my big-O 2010-11-17T00:36:31 amstan: yep looks like it's working now 2010-11-17T00:36:31 almost 2010-11-17T00:36:40 however, my point is really this, as I said on the forums: next time, pick a game that can be perfectly and fairly judged, make that a criteria of the game choice 2010-11-17T00:36:41 ah okay, i totally don't remember it 2010-11-17T00:36:51 antimatroid: I don't think it will be necessary but we'll see 2010-11-17T00:37:17 krokkrok: there is value in the random maps even if they just totally dropped the ball on this one 2010-11-17T00:37:17 we are not dealing with billion+ n's and m's so big-o is a pile of crap 2010-11-17T00:37:36 or for bot's different in ELO rank by X, have them play min(3, 100-x/2) maps, or so 2010-11-17T00:37:41 *** iFire has joined #aichallenge 2010-11-17T00:37:48 Mjothitnir: fair enough, 50*99*M 2010-11-17T00:37:52 *** McLeopold_ has joined #aichallenge 2010-11-17T00:37:56 *** McLeopold has quit IRC (Read error: Connection reset by peer) 2010-11-17T00:38:06 mega1: thats a couple hours 2010-11-17T00:38:09 I hope to run the main tournament until the likelihood of superiority is above 95% for at least each adjoining pair of bots 2010-11-17T00:38:18 but we'll see how that goes 2010-11-17T00:38:19 krokkrok: what's your M? 2010-11-17T00:38:30 1-4 2010-11-17T00:38:35 hello, no 2010-11-17T00:38:42 hello my friend! 2010-11-17T00:38:47 hell, no 2010-11-17T00:38:50 it would be good if a few different ranking methods were tried following the contest to see how each seems to do 2010-11-17T00:38:51 better? ;-) 2010-11-17T00:39:12 the fewer the maps the greater the variance 2010-11-17T00:39:29 only if they are kept secret 2010-11-17T00:39:39 and i mean more about map variety than anything 2010-11-17T00:39:47 at the current servers rate/24hrs (~120000) 100 players eat through ~12 maps per day 2010-11-17T00:39:53 but again, i think this contest is already hosed, so I am more concerned about the next 2010-11-17T00:40:11 krokkrok: your idea of a perfect competition sounds very dull to me :) 2010-11-17T00:40:54 why is correct judging dull? you want bias? isnt that what you are trying to remove from the maps?? why bother?? 2010-11-17T00:41:20 Mjothitnir: I'm not sure that's better than generating a new random map for each match 2010-11-17T00:41:22 i want general 2010-11-17T00:41:29 i will try to minise bias around that 2010-11-17T00:41:44 *** iFire` has quit IRC (Ping timeout: 255 seconds) 2010-11-17T00:41:47 minimise* 2010-11-17T00:41:49 it *cannot* be general animatroid. this is the crux of my point. this contest is entirely subjective 2010-11-17T00:42:04 krokkrok: the dullness comes from optimizing for a special case 2010-11-17T00:42:13 krokkrok: it wasn't until they started tuning the generator 2010-11-17T00:42:49 the problem is built into the nature of the contest, it cannot be avoided. the ordering is objectively unstable 2010-11-17T00:42:52 true, the final results are a bit random but it wasn't subjective 2010-11-17T00:43:08 the ranking is stable in the limit 2010-11-17T00:43:19 the randomess of maps, is part of the contest 2010-11-17T00:43:29 *** yasith has joined #aichallenge 2010-11-17T00:43:34 not everyone was working under restrictions of the old generator, so the rankings would have again been "subjective" 2010-11-17T00:43:37 it is subjective, it depends on the exact bots that enter the contest; there are always bots not entered that can radically change the outcome 2010-11-17T00:43:45 the only thing I don't like about this, is that I think it would actually be semi easy to write a 'perfect' AI 2010-11-17T00:43:55 you're wrong :P 2010-11-17T00:43:57 krokkrok: yes, as with chess 2010-11-17T00:44:12 yes, as with chess! 2010-11-17T00:44:34 any contest is always going to be subjective yes? 2010-11-17T00:44:36 krokkrok: http://dictionary.reference.com/browse/subjective that's really what you mean? 2010-11-17T00:44:37 what we are doing, actually, is acting like elks in their mating ritual hehe 2010-11-17T00:44:41 I thought that kind of shoots down your argument. 2010-11-17T00:44:44 krok, the winning bot is that one that should probably win the most games, on any unknown random map 2010-11-17T00:44:56 no, any contest is not subjective, a footrace is not subjective 2010-11-17T00:45:06 it doesn't help that everyone has a different meaning of what "ai" means 2010-11-17T00:45:30 meaning/understanding whatever 2010-11-17T00:45:34 krokkrok: so basically ruling out any contest where your opponent's actions affect your outcome? 2010-11-17T00:45:37 for my definition: the rules make ai not possible to do 2010-11-17T00:45:39 krokkrok: please come up with a fair way to create rankings in absence of a total order between players 2010-11-17T00:46:14 dont play games that dont have total order 2010-11-17T00:46:14 I want to program in the rules, some huristics, and make the ai learn how it's opponents play, to balance the huristics 2010-11-17T00:46:26 sigh: those are not games 2010-11-17T00:46:33 they're decision problems and performance rankings 2010-11-17T00:46:34 antimatroid: I know 2010-11-17T00:46:47 *I agree 2010-11-17T00:46:54 krokkrok: mario ai challenge would probably do you well 2010-11-17T00:46:55 the list of possible future games does list a game that fits this critera, so, yeah, these are games too 2010-11-17T00:47:04 but that is what krokkrok is advocating 2010-11-17T00:47:06 that's a decision problem for you :) 2010-11-17T00:47:14 krok, if there are 5 players, and player1 beats 3 people, but the person who beats player1 losesr to everyone else.. who should win? 2010-11-17T00:47:45 no r/s/p..like a footrace, absolute numbers 2010-11-17T00:47:56 a footrace is not a game 2010-11-17T00:48:04 sure it is 2010-11-17T00:48:07 no it's not 2010-11-17T00:48:20 semantics 2010-11-17T00:48:21 how could this contest possibly be changed, to be footrace like? 2010-11-17T00:48:24 well, okay, but it's not a game where players actions affect the others 2010-11-17T00:48:34 this contest cannot be changed. the next one can 2010-11-17T00:48:44 to a decision problem? 2010-11-17T00:48:47 assuming it OCULD be changed.. how could it? 2010-11-17T00:48:47 that's so boring 2010-11-17T00:48:50 burny: single player, conquer all planets as fast as possible with a known planet 2010-11-17T00:48:54 but yes, not as intersting 2010-11-17T00:48:59 *known map 2010-11-17T00:49:00 correct, if players actions can affect others, you get cheating and collusion and occlusion 2010-11-17T00:49:11 conquer all planets as fast as possible can be solved with brute force 2010-11-17T00:49:13 you don't get collusion for 2 player zreo sum games 2010-11-17T00:49:15 very easily 2010-11-17T00:49:15 there is no point 2010-11-17T00:49:21 *** amriedle has quit IRC (Quit: Lost terminal) 2010-11-17T00:49:34 antimatroid, this contest has dozens of ways you could collude 2010-11-17T00:49:38 burny: was just giving something that seemed to match krokkrok's specs :P 2010-11-17T00:50:06 krokkrok: such as? 2010-11-17T00:50:21 who cares i fpeople collude? 2010-11-17T00:50:32 enter two bots, one purposefully loses to the other 2010-11-17T00:50:32 hang on, i want to know how one even would? 2010-11-17T00:50:41 not to mention.. it's explicitly against the rules to collude? 2010-11-17T00:50:42 that's against the rules 2010-11-17T00:50:46 possibly, but against the rules 2010-11-17T00:50:55 possible* 2010-11-17T00:50:59 *** Mathnerd314 has quit IRC (Quit: ChatZilla 0.9.86-rdmsoft [XULRunner 1.9.2.12/20101026210630]) 2010-11-17T00:51:02 yes, against what few rules we have 2010-11-17T00:51:03 cheaters arent cheaters if dont catch them, they are champions 2010-11-17T00:51:04 plus: even if you did do that.. 2010-11-17T00:51:07 it wouldn't 2010-11-17T00:51:10 really matter 2010-11-17T00:51:17 burny, yes, yes it does 2010-11-17T00:51:21 how so? 2010-11-17T00:51:28 bumps your win percentage 2010-11-17T00:51:32 i can feed ELO to my friends 2010-11-17T00:51:39 not in a meaningful way 2010-11-17T00:51:49 this is exactly how the vegas top poker players win championships 2010-11-17T00:51:52 actually, with bayeselo, it's harder to do that than you think 2010-11-17T00:51:57 krokkrok: you can't have an interesting game where that isn't really possible 2010-11-17T00:52:00 easy as pie man 2010-11-17T00:52:14 antimatroid; yes you cna 2010-11-17T00:52:16 *can 2010-11-17T00:52:16 when the game is "pretty much decided" you message each other, then give the desired result 2010-11-17T00:52:17 pretty easy 2010-11-17T00:52:18 if you have a 10th place bot.. and you have a friend submit the exact same bot(with some minor frig-ups to throw hte game for you).. 2010-11-17T00:52:22 it can be done. just think about it. 2010-11-17T00:52:25 that bot won't be top 100 most likely 2010-11-17T00:52:53 and you gettignto beat that 100th place bot for free.. won't help your ranking, cause everyone else(that matters) will also beat it 2010-11-17T00:53:16 burny: you "message" each other once the match is already winnable by someone, then whoever has the higher ranking forfeits the game 2010-11-17T00:53:20 ok, I cannot teach you right now how to properly cheat, I can only inform you that indeed, collusion can be used and it can easily change the outcome 2010-11-17T00:53:21 it would be pretty easy to implement 2010-11-17T00:53:31 yeah, just encode signal in first move order 2010-11-17T00:53:37 contest organizers can inspect the source code ;) 2010-11-17T00:53:37 no, first move is too important 2010-11-17T00:53:42 do it once the game is "decided" 2010-11-17T00:53:45 how do they know what their ranking is? 2010-11-17T00:53:52 antimatroid: yeah, true 2010-11-17T00:53:59 *** Utkarsh has quit IRC (Ping timeout: 264 seconds) 2010-11-17T00:54:06 you send out single fleeets in a specific way to message your ranking 2010-11-17T00:54:06 I haven't given it much thought :) 2010-11-17T00:54:11 they dont need to know what their ranking is 2010-11-17T00:54:11 the one with the highest gets to win 2010-11-17T00:54:17 people have been over this in the forums 2010-11-17T00:54:25 no... how do you know YOUR ranking? 2010-11-17T00:54:26 oh.. 2010-11-17T00:54:31 you dont 2010-11-17T00:54:32 you mean like.. to submit 10 of the same bot? 2010-11-17T00:54:36 its not necessary 2010-11-17T00:54:40 the rankings are prespecified 2010-11-17T00:54:40 or a thousand! 2010-11-17T00:54:47 and instead of getting 10th place.. 2010-11-17T00:54:48 if the two bots areexactly the same shouldn't they be making the same moves (and so are decernable with out special "bad moves") 2010-11-17T00:54:54 you'll get 10th, 11th, 12th, 13th, etc.? 2010-11-17T00:55:09 the question is this: can someone cheat without being detected? 2010-11-17T00:55:22 Mjorthitnir: you could have large groups of people colluding, with a defined preference ordering of who gets to win 2010-11-17T00:55:25 it's very unlikely 2010-11-17T00:55:37 and the source can be inspected for the winning bots 2010-11-17T00:55:37 it' 2010-11-17T00:55:37 antimatroid: how large? that's the question ;) 2010-11-17T00:55:38 my proposition.. is that someone can't cheat in a meaningful way, without being detected 2010-11-17T00:55:50 krokkrok: even then, people put up with that risk for a more interesting competition 2010-11-17T00:55:51 the only way I can see of, to cheat without being detected.. 2010-11-17T00:55:53 i'm just pointing out that it can be one, i don't think it'd work very well 2010-11-17T00:56:04 think evil. lets say you wanted to win an ai contest. be a judge. then, run the bots in secret [like this contest] and then only show the ones where your bot wins! 2010-11-17T00:56:19 is to have a very good bot, that throws the match to all weak players, and plays well against top100 players.. but also throws it against you 2010-11-17T00:56:22 antimatroid: i know it can work well, this is how WOW players get gladiator titles 2010-11-17T00:56:27 krokkrok: if you don't trust the organisers of any contest, there's no point being there 2010-11-17T00:56:38 i don't really play video games 2010-11-17T00:56:44 excluding flash games and minesweeper 2010-11-17T00:56:57 yes i can tell, you are not a gamer and thus don't know all the ways to cheat 2010-11-17T00:57:20 krokkrok: the simple fact is that there is a very large class of interesting games that people want to play, which you wish to exclude 2010-11-17T00:57:34 ah, well, the definition of interesting comes into play 2010-11-17T00:57:39 krok, if the top 10 people beat you.. then they'll also beat your colluders 2010-11-17T00:57:46 so it doesn't REALLY matter 2010-11-17T00:58:03 it does matter 2010-11-17T00:58:12 because its not a>b>c, its hightest ELO score 2010-11-17T00:58:15 that is subjective, but the point is that we want a competition that will cater to these type of games 2010-11-17T00:58:23 unless you cheat in a HUGE way.. it won't make much difference 2010-11-17T00:58:36 when cheating, go big, thats what I say 2010-11-17T00:58:36 ELO score is NOT being used for final score krok 2010-11-17T00:58:56 what is then 2010-11-17T00:58:59 it doesnt even have to be ELO, simply win ratio 2010-11-17T00:59:11 *** Utkarsh has joined #aichallenge 2010-11-17T00:59:31 krok: can your win ratio be higher than top 10 bot? 2010-11-17T00:59:31 i have to go now, but I ask you this; can you create a game that meets our monkey-brain desire for 'interesting' and yet is still mathematically sound? 2010-11-17T00:59:34 it would be very hard to cheat in a way that actually placed you right at the top 2010-11-17T00:59:40 and be undetected 2010-11-17T00:59:47 it is not hard to cheat undetected. trust me 2010-11-17T00:59:49 gotta go 2010-11-17T00:59:54 *** krokkrok has quit IRC (Quit: Page closed) 2010-11-17T01:00:05 * antimatroid rolls his eyes 2010-11-17T01:00:10 :) 2010-11-17T01:00:26 he's right, that it's easy to cheat... but wrong, that it's not easy to cheat in a meaningful way 2010-11-17T01:00:32 so how many bots does he have in the contest :) 2010-11-17T01:00:38 haha 2010-11-17T01:00:41 0? 2010-11-17T01:00:59 i don't technically have one running on the main server atm :P 2010-11-17T01:01:09 i have previously 2010-11-17T01:01:13 and who cares.. without hard-core cheating.. you can't bump up your rank by more than about 10% 2010-11-17T01:02:07 unless you like.. enter 10million bots that play mostly randomly.. and hope one of them gets real lucky 2010-11-17T01:02:30 *** davidjliu has quit IRC (Quit: Page closed) 2010-11-17T01:03:08 it's fairly easy to detect cheating though, with huristics 2010-11-17T01:03:43 stuff like.. is there any 2 bots that have more than 90% similarity in the compiled byte code? 2010-11-17T01:03:50 I'm tempted to try this theory out on tcp :) 2010-11-17T01:04:03 hehe 2010-11-17T01:04:09 what theory? my theory about it not mattering? 2010-11-17T01:04:09 i'd be interested to see the games 2010-11-17T01:04:13 signaling 2010-11-17T01:04:18 tcp it is easier, you are told who are playing before the game 2010-11-17T01:04:46 anyway, i'm going out to socialise for a few hours, it's been a while 2010-11-17T01:04:59 actually 2010-11-17T01:05:05 unless the level of randomness in your AI is fairly high.. 2010-11-17T01:05:07 what is "socialise" 2010-11-17T01:05:21 you're going to beat the same 80 people, and lose to the same 20 people, on both(or very nearly the same).. 2010-11-17T01:05:43 and so having one of them throw the game to your other one.. doesn't increase your rank in any substantial way 2010-11-17T01:05:49 $10 beer and burger 2010-11-17T01:05:57 :) 2010-11-17T01:07:32 I hope for final though, if it is using a ranking like ELO.. that it will be a multi-pass ranking 2010-11-17T01:07:33 *** Utkarsh has quit IRC (Ping timeout: 245 seconds) 2010-11-17T01:07:47 since bots won't be 'getting better and worse' as per real life people, 2010-11-17T01:07:57 i'd be fairly happy with the way rankings are done now tbh 2010-11-17T01:08:38 if someone loses their second game, to a weak player(who got lucky in their first game), it will lower their rank 2010-11-17T01:08:55 similarly the other way around 2010-11-17T01:09:10 all you have to do, is have multi-pass ranking.. 2010-11-17T01:09:36 after preliminary rankings are made: re-apply the same game's from the preliminary ranking.. 2010-11-17T01:09:57 and tell every opponent that their rank is prelim(but adjust their new 'real' rank) 2010-11-17T01:10:07 repeat until the ranks don't change 2010-11-17T01:10:15 burny: all the games are taken into account each time the rankings are run 2010-11-17T01:10:30 yes, but the order the games happen in, matters 2010-11-17T01:10:32 bayeselo isn't a simple iterative elo calculation 2010-11-17T01:10:35 which is fine in RL 2010-11-17T01:10:49 only in that it effects who you play next 2010-11-17T01:11:12 if a crummier bot beat a good bot, when they were all ranked 1000.. 2010-11-17T01:11:26 then the crummy bot's ranking will not have gone up significantly 2010-11-17T01:11:27 you can feed it the same set of games in any order you want and it will turn out the same rankings 2010-11-17T01:12:03 ok, 4 people 4 games... 2010-11-17T01:12:47 *** Utkarsh has joined #aichallenge 2010-11-17T01:12:53 if there are too few games to make up for a bad draw there isn't anything you can do about 2010-11-17T01:13:08 1: A>B C>D 2: a>c d>b 3: b>c a>D 2010-11-17T01:13:40 c won 1 game and lost 2.. but their ranking will be lowest 2010-11-17T01:14:10 *** Tobu has joined #aichallenge 2010-11-17T01:14:14 *** Tobu has joined #aichallenge 2010-11-17T01:14:19 even though it should be a>b>c>d 2010-11-17T01:15:25 er, I messed that up a bit.. but I'm sure you know what I mean 2010-11-17T01:16:10 *** mega1 has quit IRC (Remote host closed the connection) 2010-11-17T01:16:22 the first game: B losing to A, hurt it's ranking.. because A wasn't considered to be "better than" B at the time 2010-11-17T01:16:32 *** mega1 has joined #aichallenge 2010-11-17T01:17:02 burny: I think it uses the current rankings, not the rankings at the time of the game 2010-11-17T01:17:11 burny: there is no "at the time" in BayesElo 2010-11-17T01:17:25 game order does not matter 2010-11-17T01:17:36 yes it does 2010-11-17T01:17:38 Top 10 players: bocsimacko(3644), bix0r4ever(3546), asavis(3541), protocolocon(3481), thedreamer(3463), dmj111(3441), shangas(3440), davidjliu(3417), george(3416), _iouri_(3411) 2010-11-17T01:18:01 In regular Elo it does, In glicko, it does, but not in BayesElo 2010-11-17T01:18:10 the order doesn't matter in conjunction with the opponents ranking AT THE TIME.. 2010-11-17T01:18:27 but the opponents ranking 'at the time' in this contest.. is wrong, quite often 2010-11-17T01:19:12 it appears that way because they aren't showing the confidence ranges. 2010-11-17T01:19:38 if my 3500rank bot takes 5 games to climb from 1000 to 3500(winning 4/5).. the person who beat me, will not be credited with a 3500point win 2010-11-17T01:19:42 @topic 2010-11-17T01:19:42 amstan: Official Google AI Challenge: http://ai-contest.com/ || Channel Logs: http://contestbot.hypertriangle.com/ || No stupid questions. Don't ask to ask. || New MapGen is being used: http://tinyurl.com/aimapgen || Turn count WILL stay at 200! 2010-11-17T01:20:09 burny: why not? 2010-11-17T01:20:10 Well, yes, but that isn't how BayesElo works at all. 2010-11-17T01:20:23 because my rank wasn't 3500 whe it beat me 2010-11-17T01:20:46 @topic add Mailing List moving to the http://tinyurl.com/behindthescenesforum 2010-11-17T01:20:46 *** contestbot changes topic to "Official Google AI Challenge: http://ai-contest.com/ || Channel Logs: http://contestbot.hypertriangle.com/ || No stupid questions. Don't ask to ask. || New MapGen is being used: http://tinyurl.com/aimapgen || Turn count WILL stay at 200! || Mailing List moving to the http://tinyurl.com/behindthescenesforum" 2010-11-17T01:20:54 the reason ELO works for something like.. chess.. and car racing... is because people "get better".. or have "off days" etc. 2010-11-17T01:21:00 I repeat, it does not matter what your rank is when you play someone. The entire result set is calculated every time. 2010-11-17T01:21:01 burny: can you give me a corrected above example that you think causes the problem? 2010-11-17T01:21:07 x 2010-11-17T01:21:27 Elo for chess is different, rank does matter at the time of match. 2010-11-17T01:21:35 I don't think the differences will be significant.. but they could be 2010-11-17T01:21:35 This is BayesElo, not the same. 2010-11-17T01:23:01 1: A>B C>D 2: A>C D>B 3: B>C A>D which games are wrong here? 2010-11-17T01:23:07 or results 2010-11-17T01:23:12 k got the basic framework to maximize my value working, except its slow as hell 2010-11-17T01:23:15 for the example you want 2010-11-17T01:23:23 k, sec.. 2010-11-17T01:25:28 A>B>C>D>B, what is this? the klein bottle of r/s/p?? :) 2010-11-17T01:25:48 they're meant to be game results 2010-11-17T01:25:54 A beats B etc 2010-11-17T01:25:55 I know 2010-11-17T01:26:14 rock>scizzors>paper>rock 2010-11-17T01:26:50 ah, specifically to the line above :} 2010-11-17T01:27:03 err, thought you were refering to... 2010-11-17T01:27:05 *** smellyhippy has quit IRC (Ping timeout: 240 seconds) 2010-11-17T01:27:24 1: A>C B>D 2: B>A D>C 3: C>B A>D should yield A>B>C>D 2010-11-17T01:28:19 *** virdo has quit IRC (Ping timeout: 265 seconds) 2010-11-17T01:28:26 Bayesian Elo Rating for fixed sceel players 2010-11-17T01:28:46 if someone upload ne version 2010-11-17T01:29:08 @later tell jmcarthur see http://ai-contest.com/forum/viewtopic.php?f=20&t=1114 2010-11-17T01:29:08 amstan: Yep. 2010-11-17T01:29:19 if someone upload ne version all games with old version mast be deleted from reiting calculation 2010-11-17T01:29:30 if someone upload new version all games with old version mast be deleted from reiting calculation 2010-11-17T01:29:50 googling, I can't find any math on the bayeselo 2010-11-17T01:29:52 Accoun: dude.. shut up 2010-11-17T01:30:31 Accoun: they aren't used 2010-11-17T01:30:48 in current realisation old games weth old version keep in sesetm - it bug 2010-11-17T01:31:16 Accoun: they aren't used for ranking though 2010-11-17T01:31:25 Accoun: it's fair to say a new submission should be treated like a brand new contestant 2010-11-17T01:31:27 you can go read the code if you want to check 2010-11-17T01:31:58 It will play it's own matches and achieve it's own ranking separate from the previous entry 2010-11-17T01:32:18 *** smellyhippy has joined #aichallenge 2010-11-17T01:32:18 *** smellyhippy has joined #aichallenge 2010-11-17T01:32:41 The bots that played to old version still get credit for beating it, since it still has it's own ranking 2010-11-17T01:33:30 burny: I'm having a hard time figuring out why you think the above should be ordered that way 2010-11-17T01:33:56 *** perestrelka has quit IRC (Ping timeout: 255 seconds) 2010-11-17T01:34:11 McLeopold_: the bots that played the old version don't get credit for those games either 2010-11-17T01:34:39 so then, the old entry is deleted 2010-11-17T01:34:43 so to speak 2010-11-17T01:34:59 So if all my previous opponents resubmit what happens to my rating/elo 2010-11-17T01:35:20 the only factor the old submission has on the new is that the initial pairings the new one receives are based on the old ones rankings until it gets it's own ranking 2010-11-17T01:35:44 *** sigh has quit IRC (Ping timeout: 255 seconds) 2010-11-17T01:35:46 McLeopold_: effectively yes although the games are still archived for people to view of course 2010-11-17T01:35:47 McLeopold_> games with old versions mast be deleted from history of all bots 2010-11-17T01:36:00 Accoun: they are for the purposes of ranking 2010-11-17T01:36:17 they are just left in the profile list so people can see the games their bot has played 2010-11-17T01:36:26 *** perestrelka has joined #aichallenge 2010-11-17T01:36:35 So if all my previous opponents resubmit what happens to my rating/elo 2010-11-17T01:36:44 you start over 2010-11-17T01:36:52 since no games will be ranked 2010-11-17T01:37:03 you won't have one until you play a current submission 2010-11-17T01:37:16 But, that is unlikely to happen all at once 2010-11-17T01:37:27 need mark unused games in history 2010-11-17T01:37:35 if it not used 2010-11-17T01:38:00 that is quite hard to do with the way everything fits together 2010-11-17T01:38:29 *** hellman has joined #aichallenge 2010-11-17T01:38:41 true, but it adds a layer to the yanking around of the elo (in my head at least, I was soley blaming BayesElo for that, know I know that it is more) 2010-11-17T01:39:17 btw, burny here is how bayeselo comes out with your example above http://pastebin.ca/1994212 2010-11-17T01:39:58 it seems ok even though there are really too few games to really tell much there of course :) 2010-11-17T01:40:24 but it should help demonstrate that bayeselo doesn't care about the order of games 2010-11-17T01:41:07 in TCP servers Bayesian Elo mast not be used 2010-11-17T01:41:29 in TCP servers need chees version 2010-11-17T01:41:52 why is D above C: if C beat a better player than D did? 2010-11-17T01:41:58 Accoun: I'm not too worried about the rankings on the tcp server. 2010-11-17T01:42:01 without + - 2010-11-17T01:42:09 because it beat c 2010-11-17T01:42:21 Accoun: if you lose a game, you have something to work on 2010-11-17T01:42:41 Accoun: and I can't tell when there is a new submission based on name. 2010-11-17T01:43:35 McLeopold_> yes, if all users will be change name for new versions 2010-11-17T01:43:53 burny: also note that this is the competition version of bayeselo that is modified not to give a first player advantage by default 2010-11-17T01:43:55 Accoun: but I can't enforce that 2010-11-17T01:44:11 if you download a standard bayeselo you need to set advantage to 0 2010-11-17T01:44:17 Accoun: I could rank by elo - the lower confidence 2010-11-17T01:44:29 Then only established bots will be ranked high. 2010-11-17T01:44:48 Whole History Rating would be better for TCP servers 2010-11-17T01:44:59 but someone would have to implement it of course 2010-11-17T01:45:13 janzert: do you know the bayeselo commands for that? 2010-11-17T01:45:31 IIRC, it's just "advantage 0" in the elo mode 2010-11-17T01:45:46 before running mm to generate the ratings 2010-11-17T01:45:51 I'm using that 2010-11-17T01:46:24 but it comes after the mm 2010-11-17T01:46:40 hmm, maybe I'm remembering wrong 2010-11-17T01:47:04 I'd try it without, after mm and before mm to see what difference it makes if any 2010-11-17T01:47:43 advantage with no number after it should show you what the current setting is 2010-11-17T01:51:30 McLeopold_: did you get a chance to test the radius distribution? 2010-11-17T01:51:46 yeah, I'm still looking at the graph 2010-11-17T01:51:49 ok 2010-11-17T01:51:53 I'm pretty sure a log is more correct 2010-11-17T01:51:59 I just have to find the best formula 2010-11-17T01:52:28 need to generate new map for eath game 2010-11-17T01:52:34 we basically need a linear distribution right? 2010-11-17T01:52:48 or genetate over 999 map as alternative 2010-11-17T01:53:17 Accoun: I'm pretty sure I went over that with you the other day, so I'm not going to rehash it again now 2010-11-17T01:53:31 janzert: i'd suggest just doing the same random selection as the original generator, then convert to r, theta 2010-11-17T01:54:31 Zannick: I'm debating whether to even make a distribution change :/ 2010-11-17T01:54:52 maxRadius and fixing maximum planets are the only sure changes 2010-11-17T01:55:05 *** medrimonia has quit IRC (Ping timeout: 240 seconds) 2010-11-17T01:55:09 *** medrimonia has joined #aichallenge 2010-11-17T01:55:12 i'm mostly worried at how there seem to be more maps with absurdly close configurations 2010-11-17T01:55:54 you mean between the starter planets, overall or clusters? 2010-11-17T01:56:17 *** rebelxt_ has quit IRC (Quit: Page closed) 2010-11-17T01:56:18 janz, I think it should be that way because A lost to a better player, and C beat a better player 2010-11-17T01:56:42 overall, but there are a good handful of starting clusters 2010-11-17T01:57:20 but from the looks of that page you linked, adn the lack of math on the internet.. I think the bayesan is working basically the way I was sugesting 2010-11-17T01:57:31 the overall is probably because the radius is too small I would think right? 2010-11-17T01:57:43 i suspect that plays a factor 2010-11-17T01:57:48 I check the sqrt random radius thing and it's correct 2010-11-17T01:57:51 *checked 2010-11-17T01:57:59 the maps are basically only 1/4 the size 2010-11-17T01:58:05 with up to twice the planets 2010-11-17T01:58:07 really? 2010-11-17T01:58:12 Zannick: yes 2010-11-17T01:58:21 i thought it was only slightly more than 3/4 the size 2010-11-17T01:58:29 pi r^2 instead of 4 r^2 2010-11-17T01:58:36 *** tapwater has quit IRC (Quit: tapwater) 2010-11-17T01:59:08 does anyone know, if anyone has done the math/whatever, to make a 'perfect' bot? 2010-11-17T01:59:47 janzert: I didn't catch it when you explained your position on what Accoun suggests 2010-11-17T01:59:49 where the olny way to beat it, is to add some randomness, or to make your bot, assume it is playing against that particular bot? 2010-11-17T01:59:52 when was it? 2010-11-17T02:00:02 Zannick: you're right, they're 1/4 smaller :P 2010-11-17T02:00:36 mega1: The number of maps thing? 2010-11-17T02:00:40 yes 2010-11-17T02:01:03 mega1: probably somewhere in this past Friday to Saturday 2010-11-17T02:01:32 although I'm pretty sure I've seen antimatroid going over it since then as well 2010-11-17T02:01:53 so it might have been earlier and I'm remembering that one as the recent one 2010-11-17T02:02:01 yes we had a discussion too. What Accoun says makes sense to me. 2010-11-17T02:02:54 depends on how many games we can expect to play in the finals 2010-11-17T02:03:40 basically having a small set of maps keeps a large number of different bots playing on it so you get a good rating across those games basically 2010-11-17T02:04:26 janzert: power( rand() , 0.630929 ) 2010-11-17T02:04:41 so closer to sqrt than log 2010-11-17T02:04:52 small set of maps is the best way to introduce a lot of randomness into the ratings 2010-11-17T02:04:55 McLeopold_: is there any math behind that number or just testing? 2010-11-17T02:05:01 just testing 2010-11-17T02:05:08 but kinda math 2010-11-17T02:05:22 the number is trial and error, but I don't understand it 2010-11-17T02:06:09 It means that the set of radii from 0 - 0.5 are 1/3 as likely as the set from 0.5 - 1 2010-11-17T02:06:34 that property holds with with any 0-x and x - 3x 2010-11-17T02:06:40 There are maps on which I lose very often. The fewer maps there are the greater the chance of getting a very favorable or very bad set of maps. 2010-11-17T02:06:56 mega1: yes, it shouldn't be purely random 2010-11-17T02:07:21 There should be a good analysis of different map types so there is a good set of different types of maps 2010-11-17T02:07:24 yes, there is a tension with that and the rating between a large number and a small number 2010-11-17T02:07:44 I'm sorry? 2010-11-17T02:07:56 large and small number? 2010-11-17T02:08:07 of maps 2010-11-17T02:08:20 basically, we want uniform randomness 2010-11-17T02:08:22 *** narnach has quit IRC (Quit: Leaving.) 2010-11-17T02:08:23 small number of maps = not accurate ratings, large number = takes long time 2010-11-17T02:08:32 large number of maps to ensure a good distribution over all possible maps and small number of maps to help the ratings otu 2010-11-17T02:08:40 yep 2010-11-17T02:08:56 But, how do you make sure you don't have different map types? 2010-11-17T02:09:03 or, do have different map types 2010-11-17T02:09:07 I believe the maps are the bigger source of randomness. 2010-11-17T02:09:45 It's hard to make a representative sample with 30 maps. 2010-11-17T02:10:09 if we could take 2 maps, and calc their similarity 2010-11-17T02:10:20 we could ensure no 2 maps are too similar in style 2010-11-17T02:10:36 kinda like making sure planets are not too close 2010-11-17T02:10:37 you can cluster them by results 2010-11-17T02:10:39 mega1: I admit the 30 maps was a bit of a test, it's what I consider the extreme low end 2010-11-17T02:10:52 if 'most' maps end up having the starting locations be far from each other, I think that should have been part of the original problem description.. although I also think stating planets near each other is nearly always a draw 2010-11-17T02:11:19 as Mjothitnir calculated at the current rate 12 maps for a top 100 round robin takes a day 2010-11-17T02:11:43 shouldn't need a full 100 2010-11-17T02:11:50 we need to calculate a few different dimensions of maps, then ensure no 2 maps have the same values for all dimenseions 2010-11-17T02:12:01 starting distance is a big factor, so that should be one 2010-11-17T02:12:23 doing a full round robin with the nearest 10 should be sufficient.. and half of the maps for nearest 20, and 30% for nearest 30, etc. 2010-11-17T02:12:39 you must take a representative sample from the generator 2010-11-17T02:12:43 the cluster factor of a map could be another 2010-11-17T02:12:59 if you start applying heuristics you're very likely to distort it significantly 2010-11-17T02:13:22 mega1: but if we can think of the most important factors, it will help 2010-11-17T02:13:43 the best way to get a representative sample from a infintely large random population.. 2010-11-17T02:13:52 is to take a large random sample 2010-11-17T02:13:54 umm: I'm trying to take a representative sample from the generator 2010-11-17T02:14:07 that's almost infinite 2010-11-17T02:14:46 with no huristics 2010-11-17T02:15:04 or: use huristics just to determine how large that sample should be 2010-11-17T02:15:54 McLeopold_: going back to radius distribution, the chance of a number should be weighted by the number itself gives the correct distribution. 2010-11-17T02:16:17 ? 2010-11-17T02:16:18 if player A is vastly better than player B.. but B always beats A on a certain type of map.. and that map type only shows up say, 3% of the time. if it exists 3times in a set of 30.. is that bad? 2010-11-17T02:16:32 or you create a list of the possible "types" and have the generator create maps of each type. 2010-11-17T02:16:34 the current code sqrt(random.random()) code is correct 2010-11-17T02:17:05 we already have types: line a point symmetric 2010-11-17T02:17:12 and 2010-11-17T02:17:17 mega1: I think it is power( rand() , 0.63.. ) 2010-11-17T02:17:25 is it against the rules to abuse known facts about the RNG being used, and the map generation code? 2010-11-17T02:17:29 mega1: what's your reasoning? 2010-11-17T02:17:39 Top 10 players: bocsimacko(3630), bix0r4ever(3594), asavis(3503), protocolocon(3475), luksian(3465), dmj111(3446), davidjliu(3445), shangas(3439), george(3424), _iouri_(3414) 2010-11-17T02:17:47 McLeopold_: I plotted a sample of 1000 random points generated by it 2010-11-17T02:17:58 stupid rules if u can change winer without changing rules 2010-11-17T02:18:02 looked at a roughly uniform distribution in a circle 2010-11-17T02:18:15 eg. there's 4billion possible seeds, and I know that 7% of those 4billion maps have some trait(even though only 3% of the total population has that), then the AI should favor that trait slightly 2010-11-17T02:18:22 trivially follows from ratios of areas of circles 2010-11-17T02:18:54 it think it should be circumference, not area 2010-11-17T02:19:01 why? 2010-11-17T02:19:10 what kind of uniform distribution do we want? 2010-11-17T02:19:18 because we are dealing with points, not area 2010-11-17T02:19:41 this is the chance of placing it on any particular circumference 2010-11-17T02:19:58 circle has smaller area, smaller average distance, and smaller maximum distance, than a square of equal circumfrance 2010-11-17T02:20:10 what I think we want that the expected number of planets in any region is linearly proportional to its area 2010-11-17T02:20:15 "types" is more than just symmetry (point/line), but number of planets, radius, spacing, etc 2010-11-17T02:20:28 er, no it doesn't.. 2010-11-17T02:20:41 mega1: right so we need the chance of placing at radius to be half that of placing it at radius 2 2010-11-17T02:20:48 no 2010-11-17T02:20:52 ya, I agree.. circumfrance is likely better over all(for preserving traits of the old maps dimensions) 2010-11-17T02:21:18 take two circles 2010-11-17T02:21:33 if the map is quantized to integer radius I'm positive that is right 2010-11-17T02:21:43 one with r1=r and with r2=2*r 2010-11-17T02:21:52 I'm not quite sure it holds for a smooth radius 2010-11-17T02:22:01 the area of the larger is 4 times that of the smaller 2010-11-17T02:22:17 need add ELO + - field in "Current Rankings" 2010-11-17T02:22:32 hence you want 1/4 chance of producing a radius < r1 2010-11-17T02:22:39 that's it. 2010-11-17T02:23:57 hmm 2010-11-17T02:24:15 mega1: say they can be placed at r 1-4 2010-11-17T02:24:17 but as I said I had also plotted the generated points 2010-11-17T02:25:53 placing 10 planets on it on average by my method above places the same number of planets on each radii as the radius is 2010-11-17T02:26:04 your property also holds 2010-11-17T02:26:13 for r1=2 2010-11-17T02:26:38 so I think what we said is equivalent for the property you and I want 2010-11-17T02:27:15 possibly, I haven't fully internalized your version 2010-11-17T02:27:28 just saying that there is no need to touch _that_ part 2010-11-17T02:27:39 *** aerique has joined #aichallenge 2010-11-17T02:27:52 I'm still not convinced, I'm thinking... 2010-11-17T02:31:30 mega1: when you plotted generated points, did you compaire them to any other generation method? 2010-11-17T02:31:56 no, I only eyeballed it 2010-11-17T02:32:15 systematic distorsions are easy to see 2010-11-17T02:32:19 *** Utkarsh has quit IRC (Ping timeout: 265 seconds) 2010-11-17T02:33:09 so, if I had 1 planet at every x, y integer value, or a grid, that would be uniform? 2010-11-17T02:33:20 cropped by a circle, of course 2010-11-17T02:33:24 argh 2010-11-17T02:33:38 the current formula is screwed up 2010-11-17T02:33:52 mega1: ^^ 2010-11-17T02:33:59 McLeopold_: yes 2010-11-17T02:34:05 okay, we agree there 2010-11-17T02:34:58 it doesn't produce up to the maximum radius but instead the minRadius + sqrt(maxRadius-minRadius) 2010-11-17T02:35:35 since it is multiplying by the sqrt(random.random()) 2010-11-17T02:35:49 no, look at the ) 2010-11-17T02:35:59 *** sigh has joined #aichallenge 2010-11-17T02:36:13 math.sqrt(random.random()) 2010-11-17T02:36:36 math.sqrt(1) == 1 2010-11-17T02:36:56 the graph of math.sqrt(random.random()) slopes upward 2010-11-17T02:36:57 is there a way to find your own bot in the rankings easily? 2010-11-17T02:36:58 heh, umm oops 2010-11-17T02:37:13 still [0,1) 2010-11-17T02:37:16 yeah, ok. maybe it's time for me to get some sleep 2010-11-17T02:37:26 :) 2010-11-17T02:37:54 when's the deadline for this thing? in about 5days? 2010-11-17T02:38:13 do you mean what page in the rankings or? 2010-11-17T02:38:42 burny: your profile page will give you the ranking you are at 2010-11-17T02:38:58 k 2010-11-17T02:39:02 *** Utkarsh has joined #aichallenge 2010-11-17T02:39:09 look at the front page for the relevant dates 2010-11-17T02:39:11 int(rank/100)+1 = page, if you want to see your rank mates 2010-11-17T02:39:35 as long as min is always 0, it's okay 2010-11-17T02:41:03 curious about how bad the trivial bot changes I made will do.. think I for sure lose to rage bot, but should do ok against other bad bots 2010-11-17T02:41:18 janzert: that is wrong 2010-11-17T02:41:44 McLeopold_: what is? 2010-11-17T02:41:56 it should produce a random radius, then regenerate if it is within the min distance 2010-11-17T02:42:09 since it is not linear 2010-11-17T02:42:10 *** Mistmanov has joined #aichallenge 2010-11-17T02:45:36 McLeopold_: are you convinced about the sqrt part yet? 2010-11-17T02:45:40 I've fairly convinced myself that sqrt is correct 2010-11-17T02:45:48 no, still looking at it 2010-11-17T02:48:33 *** Utkarsh has quit IRC (Ping timeout: 265 seconds) 2010-11-17T02:51:29 mega1: do you agree about needing to regenerate if below min radius instead of the current method 2010-11-17T02:51:45 which line is that? 2010-11-17T02:51:54 I agree with McLeopold_ but at the moment I don't trust myself :) 2010-11-17T02:51:59 184? 2010-11-17T02:52:07 no 57? 2010-11-17T02:52:17 the body of rand_radius 2010-11-17T02:53:25 basically the current method is offsetting sqrt from zero to max_r - min_r 2010-11-17T02:53:45 if min <= math.sqrt(random.random()) * max: ...? 2010-11-17T02:53:46 instead of the correct sqrt from min_r to max_r 2010-11-17T02:54:17 the default bot ranks around an ELO of 500? 2010-11-17T02:54:25 yes, McLeopold_ is right 2010-11-17T02:54:33 http://pastebin.ca/1994268 2010-11-17T02:54:39 is my current version 2010-11-17T02:55:10 is there no do-while loop in python 2010-11-17T02:55:10 ? 2010-11-17T02:55:18 no 2010-11-17T02:55:27 other than that it's good 2010-11-17T02:55:46 it'll break if min_r is less than 0 of course, but I'm not worried about that right now 2010-11-17T02:55:55 mega1: for radius 0.5 to radius 1.0, you want about 1/4 of the numbers below 0.5 and 3/4 above? 2010-11-17T02:56:05 yes 2010-11-17T02:56:58 the easiest way to generate a equally-distributed random point in a circle.. 2010-11-17T02:57:13 is to generate 2 points, until x*x+y*y < r*r 2010-11-17T02:57:16 *** Utkarsh has joined #aichallenge 2010-11-17T02:57:33 so, a random number of 0.25 should map to 0.5 2010-11-17T02:58:16 yes, conveniently sqrt(0.25)=0.5 :-) 2010-11-17T02:59:35 okay then, sqrt it is 2010-11-17T02:59:48 Yay. One last thing to sort out 2010-11-17T02:59:59 Unfortunately this truly is a bit subjective 2010-11-17T03:00:12 uh-oh 2010-11-17T03:00:20 * burny is the master of removing subjection frm subjective topics 2010-11-17T03:00:21 mega1: yeah no kidding 2010-11-17T03:00:38 muh-hahahahaha 2010-11-17T03:00:43 (lavalamp's map gen: http://pastebin.com/zBafpvuM, just to diff against) 2010-11-17T03:01:02 Should the new maps have the same average density of planets (i.e. same area) or the same maximum distance or a compromise between the two? 2010-11-17T03:01:28 so, box or circle? 2010-11-17T03:01:29 it depends on the original problem description 2010-11-17T03:01:38 what did the original generator do? 2010-11-17T03:01:39 burny: it wasn't specified at all 2010-11-17T03:01:57 uniform in a square? 2010-11-17T03:02:02 the original generator fit all maps within a square with 24 distance on a side 2010-11-17T03:02:05 then: it should be what creates the most interesting maps.. 2010-11-17T03:02:14 yes, I believe uniform within that square 2010-11-17T03:02:24 interesting being: not an obivous stalemate 2010-11-17T03:02:38 the new generator does uniform within a circle. currently with max diameter of 24 2010-11-17T03:02:43 mega1: that pastebin didn't work for me? 2010-11-17T03:03:01 this means the new maps currently are smaller both in area and maximum distance 2010-11-17T03:03:12 mega1: because I kept the comma (whoops) 2010-11-17T03:03:12 strange, it works for me. Maybe remove the comma at the end 2010-11-17T03:03:13 and correspondingly have a higher density 2010-11-17T03:03:36 janz, I think the squared average distance, should be maintained, rather than the maximum 2010-11-17T03:03:46 and for this was the 12->17 change proposed, right? 2010-11-17T03:03:56 mega1: yes 2010-11-17T03:04:18 maxRadius 17 restores approximately the same maximum distance 2010-11-17T03:04:19 well, do we want square or circle? that has to be answered first 2010-11-17T03:04:22 *** grog has joined #aichallenge 2010-11-17T03:04:27 if I wanted to be "correct" I'd balance the non-overlapping areas of the square and the circle 2010-11-17T03:04:34 McLeopold_: I'm not changing it from the current system for that 2010-11-17T03:04:43 so circle then? 2010-11-17T03:04:47 15.3 would be my recommended radius 2010-11-17T03:04:55 McLeopold_: yes 2010-11-17T03:04:56 so compute the area of the square that isn't covered by the circle 2010-11-17T03:05:02 changing from circle bound in square (12), to square bound in circle (17) 2010-11-17T03:05:04 to get the same squared average distance 2010-11-17T03:05:10 and compute the area of the circle that isn't covered by square 2010-11-17T03:05:32 I think you want the circle with area 24*24 2010-11-17T03:05:33 (same square) 2010-11-17T03:05:36 mega1: doesn't that work out to making the same area? 2010-11-17T03:05:37 choose R to minimize the difference between the two areas (i.e. make them equal) 2010-11-17T03:06:22 r=17 circle has more area than a side=24 square 2010-11-17T03:06:37 burny: right 2010-11-17T03:06:43 13.5 would give same area 2010-11-17T03:06:47 janzert: hmm, that's a less scientific way to put it :-) 2010-11-17T03:06:49 that's what we are discussing 2010-11-17T03:07:14 mega1: just a different way of looking at it again :P 2010-11-17T03:07:23 sorry, for overcomplicating it 2010-11-17T03:07:39 13.5 would increase the overall relative-density of the planets 2010-11-17T03:07:59 I'd also make it a square 2010-11-17T03:08:00 for a 23 planet map, it should be the same 2010-11-17T03:08:10 15.3 would maintain relative density 2010-11-17T03:08:31 I guess I really meant to maintain the density for a 23 planet map sorry 2010-11-17T03:08:35 ha, yes square! Janzert is out voted! 2010-11-17T03:08:49 I even know of such a generator. 2010-11-17T03:08:59 you'll have to appeal to Jeff to get that change made 2010-11-17T03:09:20 if you choose 3 random points in a square and circle of equal surface area.. the triangle forced in the square will probably have a larger surface area than teh triangle formed in the circle 2010-11-17T03:09:41 but if it has to stay a circle then it is really unclear how to balance distances and area 2010-11-17T03:09:42 i.e. I'm not making any wholesale changes of the generator past or future that Jeff doesn't force on me :) 2010-11-17T03:10:29 you could do half sqaure, half circle 2010-11-17T03:10:34 to be fair to him though he didn't insist on this particular generator at the last change 2010-11-17T03:10:41 McLeopold_: heh 2010-11-17T03:10:59 or oval 2010-11-17T03:11:39 so I'd make it slightly larger in area than the square to allow distances to get closer 2010-11-17T03:11:50 to the distances in the square. 2010-11-17T03:11:55 I think that what should be maximized, is the standard deviation, of number of plants captured, on games by top100 bots 2010-11-17T03:12:12 mega1: that's the direction I'm leaning 2010-11-17T03:12:19 cool 2010-11-17T03:12:33 whatever radius results in maps having inconsistent number of captured planets, is the best radius for me 2010-11-17T03:12:59 in which case 15 may make sense 2010-11-17T03:13:47 it's about in the middle of old area and distance as well as overall average density 2010-11-17T03:13:55 14.5 2010-11-17T03:14:03 12 + 17 / 2 2010-11-17T03:14:07 and is a nice integer 2010-11-17T03:14:19 McLeopold_: farther from average density though 2010-11-17T03:14:21 geeze, your generating reals 2010-11-17T03:14:24 sqrt((12*12+17*17)/2) 2010-11-17T03:14:27 but since the number of planets varies asymmetrically around 23 2010-11-17T03:14:30 hmm, is that right? 2010-11-17T03:14:47 can you make radius a function of number of planets? 2010-11-17T03:14:58 or was the point to have different densities? 2010-11-17T03:15:07 that would cause less variation in density 2010-11-17T03:15:08 15 and 30, almost symmetrical 2010-11-17T03:15:13 I think different density was the point 2010-11-17T03:15:33 actually, it doesn't matter 2010-11-17T03:15:54 unless everything is in the unit circle 2010-11-17T03:16:09 for bot tactics that is 2010-11-17T03:16:27 it does I think 2010-11-17T03:16:36 it just makes the games develop slower with less density 2010-11-17T03:16:47 and you want fast games right? 2010-11-17T03:16:53 density is the same as changes to stating population and growth 2010-11-17T03:17:17 since bots may not look ahead until the end of the game this may be a factor 2010-11-17T03:17:21 if everythign is twice as far away, it's the same as if you had double growth 2010-11-17T03:17:40 Top 10 players: SUA.khb(3680), bocsimacko(3654), bix0r4ever(3552), asavis(3535), protocolocon(3472), dmj111(3457), shangas(3440), _iouri_(3427), george(3416), Xrillo(3410) 2010-11-17T03:17:41 but there are more important ones. 2010-11-17T03:18:02 *** perror has joined #aichallenge 2010-11-17T03:18:35 double distance = double growth and half game duration 2010-11-17T03:18:41 hmm, it does matter because of the minimum distances 2010-11-17T03:19:08 oh yes, double average distance also halves the minimum distance 2010-11-17T03:19:16 yep 2010-11-17T03:19:21 *** grog has quit IRC (*.net *.split) 2010-11-17T03:19:22 *** Mistmanov has quit IRC (*.net *.split) 2010-11-17T03:19:22 *** lavalamp_ has quit IRC (*.net *.split) 2010-11-17T03:19:22 *** eburnette has quit IRC (*.net *.split) 2010-11-17T03:19:22 *** Mjothitnir has quit IRC (*.net *.split) 2010-11-17T03:19:22 *** Ice_Harley has quit IRC (*.net *.split) 2010-11-17T03:19:55 blah, stupdi bot lost all3 games so far 2010-11-17T03:19:56 janzert: so I would go closer to 17 than 12 2010-11-17T03:20:12 need it to win atleast one, so I have a clue of what lvl it's at 2010-11-17T03:20:38 McLeopold_: so 15 right? :P 2010-11-17T03:20:44 I vote 15.3 2010-11-17T03:20:55 mega1: the generator you posted earlier is the one in svn? 2010-11-17T03:20:58 Yeah...I take my time. 15 is probably good 2010-11-17T03:21:01 if you want new maps to be as similar as possibel to old maps: but be round instead of square 2010-11-17T03:21:01 at least it seems to be 2010-11-17T03:21:27 janzert: what about the clustering of planets. Isn't there a bias in the generator? 2010-11-17T03:21:29 no, it's the svn version + two fixes suggested by antimatroid and implemented by lavalamp 2010-11-17T03:21:35 as far as I know 2010-11-17T03:21:46 I implemented the same changes on my server. 2010-11-17T03:21:47 (there is another one below it) 2010-11-17T03:22:31 fantistic, just lost to aibot_current_dev -93 elo 2010-11-17T03:23:36 Oh, I see it does have the max planet fix and changes the radius to 17 2010-11-17T03:23:39 so I lose 3/3, and another bot loses 5/15.. but I'm ranked above it(it beat me) 2010-11-17T03:24:05 the max planet fix was wrong though 2010-11-17T03:24:11 nm, my ratign just droped more(without playing more games) 2010-11-17T03:24:30 burny none of the servers update rating after every game 2010-11-17T03:24:33 for sure the ELO re-adjusts as games are played 2010-11-17T03:24:33 which we already discussed tonight 2010-11-17T03:24:55 burny: are you on zeroviz? 2010-11-17T03:24:56 but only every so often after a fixed interval 2010-11-17T03:25:02 zeroviz? 2010-11-17T03:25:12 the tcp server 2010-11-17T03:25:17 jan: ahhh 2010-11-17T03:25:38 it seemed to change after each game, but maybe it was laged behind(and the first one, was what it was before any game) 2010-11-17T03:26:10 although I guess I haven't actually looked at the tcp code 2010-11-17T03:26:25 it's just my understanding that it operates that way the same as the contest server 2010-11-17T03:26:34 mine does once a minute 2010-11-17T03:26:52 dhartmei wrong a script that I assume runs at a similar interval 2010-11-17T03:26:55 wrote 2010-11-17T03:27:19 roughly what rank does ragebot get? 2010-11-17T03:27:25 bad 2010-11-17T03:27:55 no clue how bad? 2010-11-17T03:28:09 -237 on my server 2010-11-17T03:28:32 I would expect it to beat the sample bot nearly 100% of the time 2010-11-17T03:28:39 McLeopold_: I don't see anything that would obviously cause clustering. Also remember true random tends to be more clustered than humans generally think it should be 2010-11-17T03:28:52 yes 2010-11-17T03:29:36 Is there something specific in the code that is causing skew or is it just from looking at generated maps? 2010-11-17T03:30:03 it's the line symmetry ones that looked skewed to me. I'm reviewing the code 2010-11-17T03:30:22 It seems like stuff is always too close to the line 2010-11-17T03:30:24 they are always on one side of the center 2010-11-17T03:30:33 all of them 2010-11-17T03:30:43 the center ones you mean? 2010-11-17T03:30:51 ya.. they're always on the same side of the center 2010-11-17T03:30:56 as per the code 2010-11-17T03:30:57 on one side of the center planet 2010-11-17T03:30:57 mega1: that just means center is really the edge right? 2010-11-17T03:31:02 no 2010-11-17T03:31:11 otherwise it's the center :-) 2010-11-17T03:31:28 but the equidistant neutral are thrown to one side 2010-11-17T03:31:34 it generates in teh range of 0 to r 2010-11-17T03:31:38 instead of -r to r 2010-11-17T03:32:14 mega1: ahh 2010-11-17T03:32:15 it seems deliberate 2010-11-17T03:32:25 but I can't imagine why 2010-11-17T03:33:09 probably it's the most fair solution 2010-11-17T03:33:11 unless maybe.. it was intended that the angle is re-generated for each planet 2010-11-17T03:33:28 and most general. 2010-11-17T03:33:41 mega1: what is? 2010-11-17T03:33:50 joking 2010-11-17T03:34:12 you need to learn to express sarasims on the internet better :P 2010-11-17T03:34:12 well stop it :) 2010-11-17T03:34:43 I'm just trying to get this done so I can go to bed and let the community bash it before I generate more maps 2010-11-17T03:34:52 mega1: yeah, that's it, the center neutrals are always one side of the actual center 2010-11-17T03:35:09 think my bot will almost always go to 200turns if it wins.. it has no clue how to attack 2010-11-17T03:35:32 but even if they are distributed evenly on the line section they will still tend to build clusters 2010-11-17T03:35:34 it should randomly choose a side 2010-11-17T03:35:40 it does 2010-11-17T03:36:02 McLeopold_: the center neutrals are mirrored either side of center aren't they? 2010-11-17T03:36:14 *** Mistmanov has joined #aichallenge 2010-11-17T03:36:14 *** grog has joined #aichallenge 2010-11-17T03:36:14 *** lavalamp_ has joined #aichallenge 2010-11-17T03:36:14 *** eburnette has joined #aichallenge 2010-11-17T03:36:14 *** Ice_Harley has joined #aichallenge 2010-11-17T03:36:17 it randomly choose a side.. and then generates all middle plants on that side 2010-11-17T03:36:22 look at thetaA and thetaB 2010-11-17T03:36:38 theta a/b are the NON middle ones 2010-11-17T03:36:47 oops, wrong symmetry 2010-11-17T03:36:56 *** narnach has joined #aichallenge 2010-11-17T03:36:56 *** narnach is now known as narnach_mb 2010-11-17T03:37:28 it seems to me that you should choose an angle and radius, then, if the two planets are too close, merge them into one 2010-11-17T03:37:29 also: do to rounding errors.. it's possible for symetry to result in one side having a distance of 2.00000 and the other 1.9999999 2010-11-17T03:37:37 that might be important to fix 2010-11-17T03:37:57 the epsilon is suppose to fix that 2010-11-17T03:37:59 *** Utkarsh has quit IRC (Ping timeout: 255 seconds) 2010-11-17T03:38:10 yeah, epsilon is to prevent that 2010-11-17T03:38:28 so if 2 planets have a distance that is real close to an integer, it throws one away? 2010-11-17T03:38:29 it's used in the function not_valid 2010-11-17T03:38:36 yes 2010-11-17T03:38:39 ok 2010-11-17T03:39:28 woot, 0for 4! 2010-11-17T03:39:34 ugh, so yeah the unbalanced center planets is definitely purposeful 2010-11-17T03:39:43 wish I knew why 2010-11-17T03:40:39 but I guess I don't see a pressing need to change it now 2010-11-17T03:40:39 it's the spice I tell you 2010-11-17T03:40:46 no, it's ok 2010-11-17T03:43:03 sleepy times for me 2010-11-17T03:43:05 laters 2010-11-17T03:43:12 well, picking 0-5 centers just because throws of the uniform densidty 2010-11-17T03:43:50 *** Ice_Harley has quit IRC (Ping timeout: 265 seconds) 2010-11-17T03:43:50 it would be better to merge if it was too close to the line 2010-11-17T03:44:46 yeah, that may be but too large of a change for too little benefit for me to consider now 2010-11-17T03:45:02 except for last week? 2010-11-17T03:45:07 heh 2010-11-17T03:45:21 janzert: no, I think that is a big deal 2010-11-17T03:45:21 let's try to finalize it 2010-11-17T03:45:39 the line symetry maps put too much on the center line 2010-11-17T03:45:57 remember I wasn't in favor of changing last week either 2010-11-17T03:46:10 my vote was to use the original generator 2010-11-17T03:46:23 do you guys really think is right for people with bots in the game to be desiging the maps??? 2010-11-17T03:46:34 no 2010-11-17T03:46:48 I'm tainted and withdraw. 2010-11-17T03:46:50 grog: the overall properties sure 2010-11-17T03:46:52 if it is an open process, then yes 2010-11-17T03:47:08 grog: seriously, you are right 2010-11-17T03:47:25 the organizers don't have experience on the maps 2010-11-17T03:47:47 we are not hand designing maps 2010-11-17T03:47:59 we are trying to produce a system that does not have bias 2010-11-17T03:47:59 I don't I ever had 900 elo on tcp 2010-11-17T03:48:03 +think 2010-11-17T03:48:08 it's getting close 2010-11-17T03:49:34 who is "a" and "aibot_current_dev"? 2010-11-17T03:49:46 a is flagcapper 2010-11-17T03:50:25 janzert: if 0-5 is a number that maintains uniform density, then great, otherwise, it is a bais 2010-11-17T03:50:29 changed his name generator 2010-11-17T03:51:27 this is not an open process, there are 4000 players and about 4 here 2010-11-17T03:51:29 McLeopold_: what is the big problem with non-uniform density? also I doubt it's lost in the noise density variation anyway 2010-11-17T03:52:26 well, yeah, it's nit picky, but I think overall, you'll have more planets on the line with all the line symmetrical maps 2010-11-17T03:53:01 grog: we are just cleaning up. Neither of us wanted to change the generator. 2010-11-17T03:53:04 so, I think the probabilty of a planet on the line is higher than anywhere else 2010-11-17T03:53:17 isn't that the point? 2010-11-17T03:53:21 *** Utkarsh has joined #aichallenge 2010-11-17T03:53:28 Well, I suppose it might be. 2010-11-17T03:53:32 other wise you could just rip that code out 2010-11-17T03:53:50 you'd still want to merge if you generated near the line 2010-11-17T03:53:59 http://zeroviz.us:8080/ down? 2010-11-17T03:54:00 otherwise you'd have to little on the line 2010-11-17T03:54:09 Accoun: no 2010-11-17T03:54:29 http://zeroviz.us:8080/ down for me 2010-11-17T03:54:48 Accoun: I'm getting connections from others 2010-11-17T03:54:53 Accoun: it's working for me 2010-11-17T03:55:17 IP the same? 2010-11-17T03:55:21 resolves to 98.247.248.39 2010-11-17T03:55:21 Except from people from china... 2010-11-17T03:55:23 for me that is 2010-11-17T03:55:29 98.247.248.39 995 ? 2010-11-17T03:56:02 Accoun: that's correct 2010-11-17T03:56:34 down for me now 2010-11-17T03:56:50 that sucks 2010-11-17T03:57:31 screenshot time 2010-11-17T03:57:36 903 elo 2010-11-17T03:57:50 janzert: i suppose the other thing is that the clusters never seem to be in the center 2010-11-17T03:57:59 just the edges 2010-11-17T03:58:10 connect: Connection timed out 2010-11-17T03:58:10 kill: No such process 2010-11-17T03:58:19 McLeopold_: there are more edges then inner areas 2010-11-17T03:58:23 and http down 2010-11-17T03:59:14 mega1: I mean the edges of the line 2010-11-17T03:59:22 so around 2 points 2010-11-17T03:59:48 hmm, I haven't noticed 2010-11-17T04:00:23 from proxi worck well 2010-11-17T04:00:25 I think I know why 2010-11-17T04:01:07 If you pick a theta near the line, you have to have a large radius so the two planets are the minimum distance 2010-11-17T04:01:22 so for those theta, only large radius will qualify 2010-11-17T04:01:52 to compensate, extra center neutrals should be generated, but they should be closer to the center 2010-11-17T04:02:15 so, the end points of the line will have higher density 2010-11-17T04:02:21 http://ai-contest.com/forum/viewtopic.php?f=3&t=1094&p=7397#p7397 2010-11-17T04:02:23 which is where the clusters usually are 2010-11-17T04:02:31 McLeopold_> im baned? 2010-11-17T04:02:43 McLeopold_> from proxi worck well 2010-11-17T04:02:56 so you all have about 12 hours to find the major bugs I've left in the generator that are sure to rune the final tournament ;) 2010-11-17T04:03:09 Accoun: no, I don't know your problem. did you try ping? 2010-11-17T04:03:11 s/rune/ruin 2010-11-17T04:03:24 janzert: I'm describing the last one 2010-11-17T04:03:25 ^^ 2010-11-17T04:03:54 McLeopold_> from proxy worck well 2010-11-17T04:03:54 accoun is way too paranoid for his own good 2010-11-17T04:04:04 Accoun: refresh your dns cache 2010-11-17T04:04:18 Accoun: or try by ip 2010-11-17T04:04:38 McLeopold_> unban please 2010-11-17T04:04:41 Accoun: I'm not going to solve you problem. But, I assure you I didn't ban you. 2010-11-17T04:04:47 Accoun: http://98.247.248.39:8080/ 2010-11-17T04:05:21 McLeopold_: at this point I think post in the forum thread and see if you can get enough consensus that it *needs* to be changed 2010-11-17T04:05:37 so, are you too tired to think? 2010-11-17T04:05:56 McLeopold_> is up 2010-11-17T04:06:06 strange 2010-11-17T04:06:11 just read above and give me a quick impression, that's all I ask 2010-11-17T04:06:36 to me that is at least a medium sized change and I really don't want to change it unless it is really going to cause major problems, well at least semi-major ones ;) 2010-11-17T04:06:56 Accoun: http://www.google.com/search?sourceid=chrome&client=ubuntu&channel=cs&ie=UTF-8&q=refresh+your+dns+cache 2010-11-17T04:07:14 amstan: he's connected now 2010-11-17T04:07:31 okay, it's medium size, but it make sense? 2010-11-17T04:07:38 McLeopold_: it's clearly that though.. lol, it's magically up after i gave him the link 2010-11-17T04:07:57 I think it was magically up after I unbanned him/ 2010-11-17T04:08:08 jk 2010-11-17T04:08:12 McLeopold_: you banned him? lol 2010-11-17T04:08:15 no 2010-11-17T04:08:30 McLeopold_: I'm not sure one way or the other 2010-11-17T04:08:34 He's a good bot. I need to know how to beat him. 2010-11-17T04:08:46 janzert: okay, thanks for looking. 2010-11-17T04:09:39 to expand a little bit, I think *some* clustering is actually good but I can see where too much could cause problems 2010-11-17T04:09:50 oh, clustering is fine 2010-11-17T04:09:57 I just don't want it to be predictable 2010-11-17T04:10:02 janzert: oh.. what do you think about my framework redesign post? 2010-11-17T04:10:07 but like I said before, I think the players have a much better feel for this than I do 2010-11-17T04:10:28 amstan: mostly that I don't want to think about it yet :/ 2010-11-17T04:10:33 lol 2010-11-17T04:10:34 :) 2010-11-17T04:10:42 but overall it looked good 2010-11-17T04:10:46 okay, but I've got the fix if you want it 2010-11-17T04:11:13 if I get antimatroid to agree? 2010-11-17T04:11:25 post about it in the forums and I'll see what the responses are by tomorrow 2010-11-17T04:11:30 k 2010-11-17T04:17:17 *** Accoun has quit IRC (Ping timeout: 255 seconds) 2010-11-17T04:17:29 *** Accoun has joined #aichallenge 2010-11-17T04:17:39 Top 10 players: bocsimacko(3665), bix0r4ever(3538), protocolocon(3490), asavis(3467), dmj111(3462), shangas(3440), _iouri_(3425), george(3421), SUA.khb(3414), Xrillo(3410) 2010-11-17T04:17:44 *** mega1_ has joined #aichallenge 2010-11-17T04:18:27 *** mega1 has quit IRC (Ping timeout: 250 seconds) 2010-11-17T04:21:20 nice, someone is using unicode names on my server 2010-11-17T04:23:39 *** McLeopold_ has quit IRC (Quit: Leaving) 2010-11-17T04:25:37 *** mega1_ has quit IRC (Ping timeout: 272 seconds) 2010-11-17T04:26:52 *** hellman has quit IRC (Ping timeout: 272 seconds) 2010-11-17T04:30:35 *** smellyhippy has quit IRC (Ping timeout: 250 seconds) 2010-11-17T04:30:49 *** Naktibalda has joined #aichallenge 2010-11-17T04:36:08 *** smellyhippy has joined #aichallenge 2010-11-17T04:37:41 janzert: are ratings still being updated? 2010-11-17T04:38:58 @rankings 2010-11-17T04:38:59 janzert: Top 10 players: bocsimacko(3663), bix0r4ever(3563), protocolocon(3485), asavis(3477), dmj111(3468), shangas(3447), george(3425), _iouri_(3424), Xrillo(3415), hutabhu(3402) 2010-11-17T04:39:00 *** hellman has joined #aichallenge 2010-11-17T04:39:07 yes 2010-11-17T04:40:04 mm right 2010-11-17T04:42:36 *** bhasker has quit IRC (Ping timeout: 240 seconds) 2010-11-17T04:43:09 *** Utkarsh has quit IRC (Ping timeout: 250 seconds) 2010-11-17T04:47:52 *** Utkarsh has joined #aichallenge 2010-11-17T04:49:56 *** bhasker has joined #aichallenge 2010-11-17T04:58:45 *** grog has quit IRC (Ping timeout: 265 seconds) 2010-11-17T05:05:16 *** Palmik has joined #aichallenge 2010-11-17T05:17:41 Top 10 players: bocsimacko(3660), bix0r4ever(3534), asavis(3487), protocolocon(3477), dmj111(3461), shangas(3447), _iouri_(3420), george(3420), Xrillo(3410), hutabhu(3390) 2010-11-17T05:23:13 *** phxtrack has quit IRC (Ping timeout: 265 seconds) 2010-11-17T05:26:27 McLeopold_> again working only from proxy 2010-11-17T05:26:56 *** Utkarsh has quit IRC (Ping timeout: 240 seconds) 2010-11-17T05:31:26 *** Utkarsh has joined #aichallenge 2010-11-17T05:36:31 *** Utkarsh has quit IRC (Ping timeout: 272 seconds) 2010-11-17T05:39:55 what am i agreeing to? 2010-11-17T05:40:09 i have never really had any issues with stuff mcleopold wants to do 2010-11-17T05:41:49 *** Utkarsh has joined #aichallenge 2010-11-17T05:59:53 *** aerique has quit IRC (Quit: ...) 2010-11-17T06:02:33 *** yasith has quit IRC (Read error: Operation timed out) 2010-11-17T06:16:05 *** yasith has joined #aichallenge 2010-11-17T06:17:41 Top 10 players: bocsimacko(3664), bix0r4ever(3567), protocolocon(3487), dmj111(3460), asavis(3457), shangas(3451), george(3426), luksian(3416), _iouri_(3414), Xrillo(3408) 2010-11-17T06:17:41 *** grog has joined #aichallenge 2010-11-17T06:17:57 *** Utkarsh has quit IRC (Read error: Connection reset by peer) 2010-11-17T06:20:50 *** grog has quit IRC (Client Quit) 2010-11-17T06:24:59 *** Utkarsh has joined #aichallenge 2010-11-17T06:27:59 *** bhasker_ has joined #aichallenge 2010-11-17T06:28:00 *** bhasker has quit IRC (Read error: Connection reset by peer) 2010-11-17T06:31:52 *** Rubicon-|-Cross has joined #aichallenge 2010-11-17T06:37:39 *** hellman has quit IRC (Remote host closed the connection) 2010-11-17T06:37:39 *** bhasker_ has quit IRC (Read error: Connection reset by peer) 2010-11-17T06:38:06 *** bhasker has joined #aichallenge 2010-11-17T06:44:08 *** Utkarsh has quit IRC (Ping timeout: 265 seconds) 2010-11-17T06:49:00 *** Utkarsh has joined #aichallenge 2010-11-17T06:54:00 *** perestrelka has quit IRC (Quit: leaving) 2010-11-17T07:17:43 Top 10 players: bocsimacko(3649), bix0r4ever(3608), Raschi(3509), luksian(3471), protocolocon(3470), asavis(3470), dmj111(3446), shangas(3432), george(3416), _iouri_(3404) 2010-11-17T07:35:34 *** mceier has joined #aichallenge 2010-11-17T07:37:18 *** Frontier has joined #aichallenge 2010-11-17T07:48:13 where is a secret TCP server? 2010-11-17T07:48:19 @TCP 2010-11-17T07:48:20 Accoun: TCP = (#1) http://www.benzedrine.cx/planetwars/, or (#2) http://72.44.46.68/, or (#3) http://zeroviz.us:8080, or (#4) 98.247.248.39:995 (zeroviz.us) 2010-11-17T07:54:14 *** Utkarsh has quit IRC (Ping timeout: 250 seconds) 2010-11-17T07:59:38 *** Utkarsh has joined #aichallenge 2010-11-17T08:13:50 *** Utkarsh has quit IRC (Ping timeout: 265 seconds) 2010-11-17T08:17:36 *** sigh has quit IRC (Remote host closed the connection) 2010-11-17T08:17:43 Top 10 players: bocsimacko(3646), bix0r4ever(3604), Raschi(3483), luksian(3463), protocolocon(3455), asavis(3443), dmj111(3439), shangas(3420), george(3414), _iouri_(3403) 2010-11-17T08:18:08 *** delt0r___ has joined #aichallenge 2010-11-17T08:19:05 *** Utkarsh has joined #aichallenge 2010-11-17T08:20:20 *** delt0r has quit IRC (Ping timeout: 276 seconds) 2010-11-17T08:27:47 *** Utkarsh has quit IRC (Ping timeout: 240 seconds) 2010-11-17T08:27:50 *** p4p4 has joined #aichallenge 2010-11-17T08:38:32 *** virdo has joined #aichallenge 2010-11-17T08:40:17 *** Utkarsh has joined #aichallenge 2010-11-17T08:43:03 *** virdo has quit IRC (Ping timeout: 265 seconds) 2010-11-17T08:52:54 @seen McLeopold 2010-11-17T08:52:54 eburnette: McLeopold was last seen in #aichallenge 8 hours, 27 minutes, and 2 seconds ago: yeah, ok, next issue 2010-11-17T08:53:00 @seen McLeopold_ 2010-11-17T08:53:00 eburnette: McLeopold_ was last seen in #aichallenge 4 hours, 31 minutes, and 40 seconds ago: nice, someone is using unicode names on my server 2010-11-17T08:55:40 @later tell McLeopold_ Feature request for your tcp server - let me specify bot name patterns that I do and don't want to be matched with. For example if I'm testing 3 versions of bot Xyz.n then I'd like them to play other bots and not each other 2010-11-17T08:55:40 eburnette: Ready to serve, my lord. 2010-11-17T08:56:42 that's a pretty good idea actually 2010-11-17T08:57:05 Patterns could be specified in the user name, like username="Mybot.n|incl=*|excl=Mybot.*" (choose your favorite syntax, doesn't matter) 2010-11-17T08:58:33 *** krokkrok has joined #aichallenge 2010-11-17T09:07:23 1 2010-11-17T09:08:08 2 2010-11-17T09:11:34 3 2010-11-17T09:11:50 * tests 2010-11-17T09:12:14 what's the command to emote 2010-11-17T09:12:20 ;) 2010-11-17T09:12:43 (is thinking) like that 2010-11-17T09:12:58 to get the * in front 2010-11-17T09:14:43   2010-11-17T09:15:41 *** krokkrok has quit IRC (Quit: Page closed) 2010-11-17T09:16:13 * Mistmanov is thinking 2010-11-17T09:16:23 slash me text 2010-11-17T09:16:32 * eburnette tests 2010-11-17T09:16:53 * eburnette is looking for a collaborator 2010-11-17T09:17:43 Top 10 players: bocsimacko(3645), bix0r4ever(3621), protocolocon(3449), Raschi(3443), asavis(3428), dmj111(3428), luksian(3423), george(3409), shangas(3406), hutabhu(3389) 2010-11-17T09:18:17 Anybody know who is bix0r4ever? just curious 2010-11-17T09:18:31 @seen bix0r4ever 2010-11-17T09:18:31 eburnette: I have not seen bix0r4ever. 2010-11-17T09:19:27 * eburnette is really depressed that bots he wrote a month ago are beating his new improved bots 2010-11-17T09:19:46 familiar problem 2010-11-17T09:20:06 what do you mean bots he made a month ago? 2010-11-17T09:20:12 added loads of "improvements", and my old bot is beating my new bot 66-34 2010-11-17T09:21:08 I mean for example Monferno.5 is pretty old but it's kicking butt compared to Houndoom.exp which suxors. 2010-11-17T09:21:39 *** mceier has quit IRC (Quit: leaving) 2010-11-17T09:21:41 yeah, i think everyone has had issues with "making it smarter" actually making it stupider 2010-11-17T09:22:27 oh, i first read that as you were depressed that bix0r's bot is getting beaten by the ones he wrote months ago, hence my confusion :P 2010-11-17T09:22:29 Most programs, let's say for example a sorting algorithm, if I write a new version I can tell if it's better or not. This is not like that. 2010-11-17T09:22:41 no, it's way too complicated :| 2010-11-17T09:24:22 I haven't figured a way to do iterative improvements. Maybe if I had a supercomputer and could run 1000 games a second after every change... 2010-11-17T09:24:36 *** Utkarsh has quit IRC (Ping timeout: 265 seconds) 2010-11-17T09:26:22 That's odd, Monferno.5 went from +110 to -55 in a handful of games, is the scoring on zeroviz working right? 2010-11-17T09:29:18 no idea sorry 2010-11-17T09:39:38 *** Utkarsh has joined #aichallenge 2010-11-17T09:41:35 im fuily demaralised 2010-11-17T09:41:51 no power 2010-11-17T09:41:56 any 2010-11-17T09:42:51 evil cnadian evil like last eye 2010-11-17T09:44:22 *** Utkarsh has quit IRC (Ping timeout: 260 seconds) 2010-11-17T09:48:39 *** Utkarsh has joined #aichallenge 2010-11-17T10:06:56 *** Utkarsh has quit IRC (Ping timeout: 276 seconds) 2010-11-17T10:12:58 *** medrimonia1 has joined #aichallenge 2010-11-17T10:15:05 *** Utkarsh has joined #aichallenge 2010-11-17T10:17:43 Top 10 players: bocsimacko(3624), bix0r4ever(3598), Raschi(3459), protocolocon(3449), luksian(3437), dmj111(3426), george(3406), shangas(3405), asavis(3397), davidjliu(3393) 2010-11-17T10:18:47 *** JamesMG has joined #aichallenge 2010-11-17T10:21:54 *** Rubicon-|-Cross has quit IRC (Quit: Leaving) 2010-11-17T10:23:44 *** Utkarsh has quit IRC (Ping timeout: 250 seconds) 2010-11-17T10:24:17 *** p4p4 has quit IRC (Ping timeout: 240 seconds) 2010-11-17T10:29:03 *** Utkarsh has joined #aichallenge 2010-11-17T10:34:47 *** medrimonia1 has quit IRC (Quit: Leaving.) 2010-11-17T10:35:22 *** medrimonia1 has joined #aichallenge 2010-11-17T10:49:45 *** hellman has joined #aichallenge 2010-11-17T11:01:20 *** p4p4 has joined #aichallenge 2010-11-17T11:17:45 Top 10 players: bocsimacko(3635), bix0r4ever(3592), luksian(3477), protocolocon(3460), dmj111(3436), asavis(3412), shangas(3412), george(3406), hutabhu(3396), _iouri_(3385) 2010-11-17T11:25:42 in local news, apparently australian politicians are still bitching about people making "anonymous comments on the internet" 2010-11-17T11:25:47 *** bhasker has quit IRC (Quit: bhasker) 2010-11-17T11:25:48 fuck people can be retarded here 2010-11-17T11:26:42 *** bhasker has joined #aichallenge 2010-11-17T11:26:49 *** mega1 has joined #aichallenge 2010-11-17T11:33:45 *** superflit has quit IRC (Quit: superflit) 2010-11-17T11:33:58 *** superflit has joined #aichallenge 2010-11-17T11:34:34 antimatroid: same thing in korea. 2010-11-17T11:34:55 antimatroid: personally, i am bit worried about australian developments regarding the net. 2010-11-17T11:35:02 brx: i expected more of australia 2010-11-17T11:35:22 a lot of people seem to have discovered the internet, now they want every accessible page safe for a 5 year old to look at 2010-11-17T11:35:26 it's ridiculous 2010-11-17T11:35:32 * brx nods 2010-11-17T11:35:42 and sad. 2010-11-17T11:35:50 meh, bring on the black net 2010-11-17T11:36:00 *** narnach_mb has quit IRC (Quit: Leaving.) 2010-11-17T11:36:11 okay, time to start working on the bot :) 2010-11-17T11:36:22 it'd 3 30am, almost time for bed me thinks :P 2010-11-17T11:36:36 it's 5:30pm here 2010-11-17T11:36:49 i'm sitting in a bar on campus, waiting for some friends to arrive 2010-11-17T11:37:38 i finished my last exam for undergrad last tuesday :) 2010-11-17T11:37:53 congrats :) 2010-11-17T11:38:42 *** superflit has quit IRC (Ping timeout: 260 seconds) 2010-11-17T11:44:41 *** Utkarsh has quit IRC (Read error: Connection reset by peer) 2010-11-17T11:46:59 *** mega1 has quit IRC (Ping timeout: 265 seconds) 2010-11-17T11:52:19 *** McLeopold has joined #aichallenge 2010-11-17T11:52:24 *** Utkarsh has joined #aichallenge 2010-11-17T11:52:39 antimatroid: you there? 2010-11-17T11:55:21 *** bhasker has quit IRC (Quit: bhasker) 2010-11-17T11:55:38 McLeopold: yeah 2010-11-17T11:55:42 i'm not that good at going to bed 2010-11-17T11:55:55 I think I see a small issue with the map generator 2010-11-17T11:55:57 oh, you just entered, nevermind, shoot 2010-11-17T11:56:08 ? 2010-11-17T11:56:20 is that what you posted in the forums about? 2010-11-17T11:56:22 For the neutral that should be takeable against RageBot 2010-11-17T11:56:29 No this is slightly different 2010-11-17T11:56:34 Even from my second post. 2010-11-17T11:57:00 *** tapwater has joined #aichallenge 2010-11-17T11:57:07 Line 184: min( 100, 5 * distance(p1, p2) - 1) 2010-11-17T11:57:17 p1 and p2 are suppose to be the home planets? 2010-11-17T11:58:14 line 184 gets a workout, it's the only line number i know off by heart :P 2010-11-17T11:58:23 i'd be happy with such changes 2010-11-17T11:58:29 mega1: what about you? 2010-11-17T11:58:37 i'm indifferent at this stage tbh, i've got what i really wanted 2010-11-17T11:58:57 no, it's not a map thing, but an actual bug 2010-11-17T11:59:25 even the planet on your side isn't guaranteed not to be unsnipable 2010-11-17T11:59:31 p1 and p2 might not be the home planets if center neutrals where created 2010-11-17T11:59:38 the variable might be overwritten 2010-11-17T11:59:48 planet_max = min(100, 5 * distance(planets[1], planets[2]) - 1) 2010-11-17T11:59:52 p1 and p2 are always players home planets 2010-11-17T12:00:12 it adds p0 as very centre neutral, then home plants, then the rest of centre neutrals, then the rest of the neutrals 2010-11-17T12:01:20 so wait, centre neutrals can be more than 5 can't they? 2010-11-17T12:01:39 could they have been before with the old one? 2010-11-17T12:01:48 or was that thrown out 2010-11-17T12:02:17 uhh... 2010-11-17T12:02:37 p1 and p2 get overwritten 2010-11-17T12:03:32 oh, i see, sorry i thought you were just talking planet numbers 2010-11-17T12:03:37 (it's 4am) 2010-11-17T12:03:51 hawaii? 2010-11-17T12:03:59 tasmania 2010-11-17T12:04:15 (australia) 2010-11-17T12:04:33 cool, I might visit there next year 2010-11-17T12:04:55 it's an awesome place :) 2010-11-17T12:05:06 *** medrimonia1 has left #aichallenge 2010-11-17T12:05:15 yeah, so, sometimes you're not calculating off the true home planets 2010-11-17T12:05:22 see my line up code above 2010-11-17T12:05:26 i may be riding across the state in january with a couple of friends, if we can get a date we're all free 2010-11-17T12:05:43 yeah, okay 2010-11-17T12:05:44 hmmm 2010-11-17T12:06:12 I'm putting a change on pastebin in a minute 2010-11-17T12:06:17 save distp1p2 in a variable or something? 2010-11-17T12:06:48 what are your thoughts on making the "special" neutrals non-snipable? 2010-11-17T12:06:53 i think that's a good idea 2010-11-17T12:07:01 but it does let stupid bots off easy at the end 2010-11-17T12:07:04 yes, I'm writing the code, one minute 2010-11-17T12:07:05 particularly for the lower ranked bots 2010-11-17T12:07:10 *** superflit has joined #aichallenge 2010-11-17T12:07:17 no, they will still not win. :) 2010-11-17T12:07:40 yes, but you might have two shit bots that are otherwise equal, but one doesn't handle those cases well 2010-11-17T12:07:48 excluding those maps will make them appear identical 2010-11-17T12:08:27 i generally care about the top 20 or so, but i can understand why people that have spent time but aren't up there would possibly want that 2010-11-17T12:08:30 if that makes sense at all? 2010-11-17T12:08:53 what we don't want is a top bot to draw against a horrible bot 2010-11-17T12:09:21 bocsimackos current upload lost to a 600+ ranked bot the other day 2010-11-17T12:09:27 because we can't guarantee a full round robin, that would be devistating if you drew the wrong map early on 2010-11-17T12:09:40 well, a loss is different 2010-11-17T12:09:50 wouldn't that be worse? 2010-11-17T12:10:02 for the ratings, yes, but you deserve a loss 2010-11-17T12:10:13 a map where there are no options is different 2010-11-17T12:10:32 i do agree ranking the top bots is most important 2010-11-17T12:10:36 add the change :) 2010-11-17T12:10:45 okay, hold on... 2010-11-17T12:10:47 i'm just trying to think of people other than myself 2010-11-17T12:10:54 i don't do it often :P 2010-11-17T12:11:27 crap, dictionaries for planets, wtf 2010-11-17T12:11:32 :) 2010-11-17T12:15:17 *** Bobng has joined #aichallenge 2010-11-17T12:15:22 what? 2010-11-17T12:16:03 i don't understand how there are "statutes of limitations on crimes" 2010-11-17T12:17:45 Top 10 players: bocsimacko(3639), bix0r4ever(3594), luksian(3476), protocolocon(3464), davidjliu(3454), dmj111(3437), shangas(3417), asavis(3414), george(3414), Raschi(3384) 2010-11-17T12:21:06 McLeopold: I'm going to head to bed, anything else before I do? 2010-11-17T12:21:18 almost there, 2 minutes 2010-11-17T12:22:18 http://pastebin.com/vGdn1pyC 2010-11-17T12:23:23 *** Frontier has left #aichallenge 2010-11-17T12:23:25 *** mega1 has joined #aichallenge 2010-11-17T12:24:17 http://pastebin.com/vGdn1pyC 2010-11-17T12:24:29 mega1: found actual bug in generator 2010-11-17T12:24:39 mega1: not just philisophical issue 2010-11-17T12:24:42 looks good to me 2010-11-17T12:24:51 it was a bad bug too 2010-11-17T12:24:54 McLeopold: what is the bug? 2010-11-17T12:25:30 the RageBot check 2010-11-17T12:25:48 It wasn't always calculating from the home planets, because the variables sometimes got overwritten 2010-11-17T12:26:13 plus, you should read the logs from 10 minutes ago 2010-11-17T12:26:31 and my two posts 2010-11-17T12:26:58 I don't think I'll convince everyone, but I won't feel better if it was you that rejected me. :) 2010-11-17T12:27:12 I mean, I would feel better. 2010-11-17T12:27:39 yep, well for any of those things i'm indifferent 2010-11-17T12:27:48 although your solutions do seem better 2010-11-17T12:29:38 just not my bot :( 2010-11-17T12:30:36 ok, I've read it 2010-11-17T12:30:43 thx 2010-11-17T12:30:57 agree or disagree? 2010-11-17T12:32:20 *** fawek has joined #aichallenge 2010-11-17T12:32:31 can you just sum it up so that I'm not answering to the wrong thing 2010-11-17T12:33:25 making the special neutrals non-snipable? 2010-11-17T12:36:07 *** bhasker has joined #aichallenge 2010-11-17T12:38:16 Anybody know if Phil Hassey has an entry? 2010-11-17T12:39:39 who is phil hassey? 2010-11-17T12:40:35 mega1: #1, make the special neutrals unsnipable. 2010-11-17T12:40:58 mega1: #2, fix the bias toward putting planets and the end points of the line in line symmetric maps 2010-11-17T12:41:34 mega1: the math on both should be checked 2010-11-17T12:41:48 #1, I don't know whether the single planet is effective at reducing the number of draws, so no idea 2010-11-17T12:42:00 since I suck at circle distribution. :) 2010-11-17T12:42:25 mega1: but at least the obvious move isn't to sit there 2010-11-17T12:42:39 I think that is enough to make a difference between a good bot and a bad one 2010-11-17T12:42:40 d0ugal: it's the one called galcon 2010-11-17T12:42:58 McLeopold: if it's obvious then even a bad bot will find it 2010-11-17T12:43:00 d0ugal: it's just the official galcon ai, essentially. not really a competitor 2010-11-17T12:43:13 mega1: true 2010-11-17T12:44:09 I can't tell without actually looking at games and measuring draw rates. 2010-11-17T12:44:20 #2 sounds good 2010-11-17T12:44:37 although I forgot what the bug was 2010-11-17T12:44:47 or what caused it 2010-11-17T12:44:56 the minimum distance check, right? 2010-11-17T12:45:03 check draw rates here: http://zeroviz.us:8080/by_Map.html 2010-11-17T12:45:11 uh yeah 2010-11-17T12:45:14 no 2010-11-17T12:45:25 the minimim distance check was already implemented 2010-11-17T12:45:43 but minimum distance plays a role in this other bias 2010-11-17T12:46:57 it might also be a bad first move 2010-11-17T12:47:10 i think it's worth having it be non-snipable 2010-11-17T12:47:16 but i'm clambering in to bed now 2010-11-17T12:47:17 night 2010-11-17T12:47:55 cya 2010-11-17T12:48:43 hmm, the radius wasn't fixed right... 2010-11-17T12:48:43 McLeopold: the greater the minimum distance the greater the emptiness in the middle 2010-11-17T12:48:50 yeah 2010-11-17T12:49:05 but more so on line symmetry 2010-11-17T12:49:28 why? this is that I can't recall 2010-11-17T12:49:28 and the greater density at the end points of the line 2010-11-17T12:49:58 we didn't discuss it last night, which is why you can't recall? 2010-11-17T12:50:28 for a theta close to the center line, small radius will be thrown out, large ones will be kept 2010-11-17T12:51:01 That's why I think merging the two planets into 1 for small radii is best 2010-11-17T12:51:25 so the radius still has an even chance of being selected near the center line 2010-11-17T12:52:05 This doesn't change the fairness of the maps. It's a pretty small nitpick. 2010-11-17T12:53:01 mcleopold those fancy tables on your servers slow the page load down a lot:-\ 2010-11-17T12:53:06 McLeopold: the code in question is 181-203, right? 2010-11-17T12:53:27 um.. 2010-11-17T12:54:23 yeah, the not valid check produces the bias 2010-11-17T12:54:38 line 197 2010-11-17T12:55:07 and if it fails it takes another random theta and r 2010-11-17T12:55:32 I think it should merge, not fail 2010-11-17T12:55:46 what would merging do? 2010-11-17T12:56:13 it would mean a planet still has an even chance of being generated anywhere on the center line 2010-11-17T12:56:15 not just the ends 2010-11-17T12:58:46 wait, the center line is not any more problematic than the others? 2010-11-17T12:59:09 what others? other lines? 2010-11-17T12:59:13 yes 2010-11-17T12:59:22 which lines? 2010-11-17T12:59:43 the ones that go through the center 2010-11-17T12:59:57 for radial? 2010-11-17T12:59:59 the center of the circle 2010-11-17T13:00:16 for either? 2010-11-17T13:00:26 for radial, it's okay 2010-11-17T13:00:43 for line, it's more of a problem the closer you get to the center line 2010-11-17T13:00:55 because of the minimum distance 2010-11-17T13:01:16 radial, the twin planet is always on the other side of the map, so doesn't get rejected 2010-11-17T13:01:57 the same thing happens with radial: the closer you are to the central point density goes down 2010-11-17T13:02:43 *** Migi32 has joined #aichallenge 2010-11-17T13:02:45 but the rejection area is only a circle of the minimum distance in the center 2010-11-17T13:03:06 it a small circle around each already generated planet 2010-11-17T13:03:07 for line, it's a box the whole length of the map 2010-11-17T13:03:19 a box?! 2010-11-17T13:03:30 what makes you think that? 2010-11-17T13:03:37 yes, or half the minimum distance to the center line 2010-11-17T13:03:56 sorry, I can't see that in the code 2010-11-17T13:04:12 please point me to the right place 2010-11-17T13:04:24 okay... 2010-11-17T13:04:40 well, line 197 2010-11-17T13:04:50 not_valid checks the minimum distance 2010-11-17T13:05:12 it checks against all planets, but... 2010-11-17T13:05:37 it is also going to check against the 2 twin planets just created 2010-11-17T13:05:52 in line, they will always be on other sides of the line... 2010-11-17T13:06:18 so if they are within half the minimum distance, they will be rejected 2010-11-17T13:06:25 yes, so far 2010-11-17T13:06:33 no 2010-11-17T13:07:32 so they are rejected, but I still can't see the bias 2010-11-17T13:08:02 Okay, I'm thinking now there isn't a bias toward the ends... 2010-11-17T13:08:32 the 0-5 center line neutrals is suppose to compensate for the void in the center 2010-11-17T13:08:47 just like the center planet compensates for the void on a radial map 2010-11-17T13:09:17 yeah, that's a good point 2010-11-17T13:09:23 0-5 seems arbitrary, that's why I prefer a merge if a point was picked near the center line 2010-11-17T13:10:10 you still haven't said how you merge and what 2010-11-17T13:10:15 two planets? 2010-11-17T13:10:19 add their growths? 2010-11-17T13:10:36 no, just pick one planet and move it onto the line 2010-11-17T13:10:41 same growth and ships 2010-11-17T13:10:50 remove the second planet 2010-11-17T13:10:57 center line neutral created! 2010-11-17T13:11:25 why is it desirable? 2010-11-17T13:11:43 so the center area has the same probability of being picked as the rest 2010-11-17T13:11:57 in radial, it 100% :) 2010-11-17T13:12:31 if, on average, there is really 0-5 center planets created with my method, then the code is fine 2010-11-17T13:14:12 the "center area" is the min distance wide box around the center line? 2010-11-17T13:14:50 yes 2010-11-17T13:15:43 *** mceier has joined #aichallenge 2010-11-17T13:15:48 I see. 2010-11-17T13:16:15 What you propose does fix the proportion of planets in that box. 2010-11-17T13:16:47 It does not fix it in the half boxes on either side of the line. 2010-11-17T13:17:28 it kinda pulls planets toward the center, due to the minimum distance 2010-11-17T13:17:45 Top 10 players: bocsimacko(3644), bix0r4ever(3602), luksian(3491), protocolocon(3461), dmj111(3452), asavis(3435), george(3422), shangas(3421), Raschi(3401), Xrillo(3377) 2010-11-17T13:18:48 if you consider the planet to be in both boxes? 2010-11-17T13:22:48 *** fawek has quit IRC (Quit: No Ping reply in 180 seconds.) 2010-11-17T13:24:12 *** narnach has joined #aichallenge 2010-11-17T13:34:18 *** fawek has joined #aichallenge 2010-11-17T13:35:25 sorry for the dealy 2010-11-17T13:35:41 my point is that there are still region where the distribution is uneven 2010-11-17T13:36:32 the practical aspect is how many planets are equal distant from the players 2010-11-17T13:36:46 how many are 1 step closer to one of them 2010-11-17T13:37:28 there is more sniping opportunity as we approach the center line 2010-11-17T13:37:28 1 step won't happend in line symmetry? 2010-11-17T13:38:08 I think it can happen. 2010-11-17T13:38:38 not if a planet can't be within 1 of the center line 2010-11-17T13:38:56 because the line connecting the player to the planet comes from an angle unrelated to angle of the line connecting the pair of planets 2010-11-17T13:39:44 so I must say again, this time for #2, that I have no idea how it would work out 2010-11-17T13:40:28 yeah 2010-11-17T13:44:28 yep, 1 step does happen 2010-11-17T13:44:41 farthur out 2010-11-17T13:47:28 mega1: #2 isn't big enough to warrent a change 2010-11-17T13:48:12 bhasker: the tables shouldn't be anymore data than the regular tcp server 2010-11-17T13:48:16 without evidence to the contrary I'd lean towards leaving it alone 2010-11-17T13:48:26 mega1: yeah, I agree 2010-11-17T13:48:28 hmm its not data i think its all the fancy Javascript 2010-11-17T13:48:40 I though it was pushing to the end points, but I don't think that anymore 2010-11-17T13:48:51 bhasker: javascript should be cached once 2010-11-17T13:49:02 what maps are the three servers running? 2010-11-17T13:49:12 just the render of that takes longer at least thats what i feel 2010-11-17T13:49:16 benzendrine is running the old ones still, I guess 2010-11-17T13:49:25 official is radius12, mine is radius17, tcp is original generator 2010-11-17T13:49:48 After janzert finalizes, today, I will regenerate. 2010-11-17T13:50:08 *** Naktibalda has quit IRC (Quit: ChatZilla 0.9.86 [Firefox 3.6.8/20100722155716]) 2010-11-17T13:50:18 bhasker: oh, the planet id and growth rate display? 2010-11-17T13:50:50 no just this page 2010-11-17T13:50:50 http://zeroviz.us:8080/malazanbot.html 2010-11-17T13:50:55 i think all the sorting it tries to do 2010-11-17T13:50:57 slows it down 2010-11-17T13:50:59 at load 2010-11-17T13:51:09 oh, I don't limit the result set 2010-11-17T13:51:18 i think you send it sorted by time upfront 2010-11-17T13:51:21 sorting is client side 2010-11-17T13:51:24 yea 2010-11-17T13:51:30 but i think it does a sort anyway 2010-11-17T13:51:31 at load 2010-11-17T13:51:39 yes, double sort the same way. 2010-11-17T13:51:44 @rainbow 2010-11-17T13:51:44 McLeopold: (rainbow ) -- Returns colorized like a rainbow. 2010-11-17T13:51:46 yea that slows the load a lot 2010-11-17T13:52:08 no, it doesn't execute until the jquery document load fires 2010-11-17T13:52:32 so, you may not see it as fast, but the browser loaded it just the same 2010-11-17T13:52:46 I can turn off the initial sort, it's not a big deal 2010-11-17T13:53:56 yea probably the initial sort turned off is better 2010-11-17T13:55:20 bhasker: looks like I build the table upside down. :) 2010-11-17T13:55:46 hmm something i weird with its building, isee some old games first at page load 2010-11-17T13:55:50 i guess the upside down explains it 2010-11-17T13:55:56 then it resorts 2010-11-17T13:56:17 err now its in reverse 2010-11-17T13:56:24 you need to order by time desc maybe? 2010-11-17T13:56:27 in your sql 2010-11-17T13:56:36 now all the latest games are at the bottom 2010-11-17T13:59:16 question: this visualizer has the option of creating a mapfile out of any turn in a game. really handy for testing http://spz.nofate.com/planetwars/visualizer/?game_id=7707441&server=Official 2010-11-17T13:59:31 any chance of incorporating that into the tcp server? 2010-11-17T13:59:35 bhasker: fixed 2010-11-17T13:59:59 Mistmanov: let me check 2010-11-17T14:01:21 thanks 2010-11-17T14:07:46 much better now mcleopold doesn't take forever to load 2010-11-17T14:07:58 cool 2010-11-17T14:08:08 I'm halfway to the game state extractor 2010-11-17T14:08:35 *** Bobng has quit IRC (Ping timeout: 255 seconds) 2010-11-17T14:09:00 cool =) 2010-11-17T14:10:12 i wonder if it would be possible to build an visualizer which can overlay custom data 2010-11-17T14:10:13 per turn 2010-11-17T14:10:16 on the graph 2010-11-17T14:10:19 or on the canvas 2010-11-17T14:10:36 i should brush up my javascript skillz 2010-11-17T14:11:03 http://ai-contest.com/forum/viewtopic.php?f=18&t=751&start=40 2010-11-17T14:11:07 was working on something like that 2010-11-17T14:11:39 hmm 2010-11-17T14:12:32 but having a visualizer that automatically displays all the logging for a turn would also be nice 2010-11-17T14:12:40 but should be done locally ofcourse 2010-11-17T14:15:07 Mistmanov: game state is now there 2010-11-17T14:15:16 thanks, gonna try it out =) 2010-11-17T14:16:10 in the visualizer what does the first number indicate 2010-11-17T14:16:13 i see like 2+5 2010-11-17T14:16:16 16+5 2010-11-17T14:16:17 etc 2010-11-17T14:16:21 planet id, growth rate 2010-11-17T14:16:26 ah doh 2010-11-17T14:16:31 it could look better 2010-11-17T14:17:07 mcleopold: works like a charm =) many thanks 2010-11-17T14:17:45 Top 10 players: bocsimacko(3647), bix0r4ever(3597), luksian(3500), protocolocon(3458), dmj111(3451), asavis(3443), shangas(3421), george(3419), Raschi(3403), Xrillo(3377) 2010-11-17T14:21:13 *** perror has quit IRC (Quit: Bye all !) 2010-11-17T14:21:44 wow, nice work McLeopold 2010-11-17T14:21:51 thx :) 2010-11-17T14:22:29 *** hellman has quit IRC (Remote host closed the connection) 2010-11-17T14:26:22 http://zeroviz.us:8080/canvas?game_id=2799 2010-11-17T14:26:45 other than my obvious fail in the center, look at how I capture planet 13 in the upper right 2010-11-17T14:27:29 oh nice, you got that working 2010-11-17T14:27:34 cool 2010-11-17T14:27:37 btw 2010-11-17T14:27:46 I think the growthrate isn't necessary 2010-11-17T14:27:49 I tried to get mine to do that correctly, but haven't seen it do that yet 2010-11-17T14:28:06 as the radius already visualizes that and it bloats it a bit 2010-11-17T14:28:33 I have a hard time telling the difference 2010-11-17T14:28:48 oh 2010-11-17T14:42:40 *** lavalamp__ has joined #aichallenge 2010-11-17T14:46:54 *** Xrillo has joined #aichallenge 2010-11-17T14:54:43 *** amstan has quit IRC (Remote host closed the connection) 2010-11-17T14:59:16 *** Accoun has quit IRC () 2010-11-17T15:02:30 *** amstan has joined #aichallenge 2010-11-17T15:02:30 *** ChanServ sets mode: +o amstan 2010-11-17T15:05:02 *** amstan has quit IRC (Remote host closed the connection) 2010-11-17T15:08:19 McLeopold: could you please show everyone within the last 24 hours in the system? 2010-11-17T15:08:43 Click Rantings 2010-11-17T15:08:49 uh, Ratings 2010-11-17T15:10:29 this is the place for rantings... 2010-11-17T15:10:45 Error323: I added a note to help you. :) 2010-11-17T15:11:10 McLeopold: :D 2010-11-17T15:11:54 E323 http://zeroviz.us:8080/canvas?game_id=5199 2010-11-17T15:12:02 my new version still sucks 2010-11-17T15:12:03 lol 2010-11-17T15:12:13 it makes too many stupid mistakes 2010-11-17T15:12:15 *** Accoun has joined #aichallenge 2010-11-17T15:13:06 btw E323 your value function is it monotonically increasing as the game goes for games that you win? 2010-11-17T15:14:49 *** Mistmanov has quit IRC (Quit: Page closed) 2010-11-17T15:15:12 McLeopold: Without code to add center line planets and instead merging planets too close, maps would average just under 1 planet on the line 2010-11-17T15:15:21 bhasker: this is my heuristic function, not alphabeta or anything 2010-11-17T15:15:27 I gave up on that given the time left 2010-11-17T15:15:45 yea mine is heuristic too, i am not doing any minmaxing 2010-11-17T15:15:52 i am just scoring my moves 2010-11-17T15:15:56 and executing the best ones 2010-11-17T15:15:59 janzert: mega1 and I hashed it out earlier, there is a slight bais, but it's not worth the change 2010-11-17T15:16:06 yeah, ok 2010-11-17T15:16:07 i take my available ships and run through the plaents and see what moves i can do 2010-11-17T15:16:16 and score each one based on the end state after the move is executed 2010-11-17T15:16:18 janzert: and my fix just introduces a different bias 2010-11-17T15:16:31 but check out the unsnipable fix 2010-11-17T15:16:50 even if we don't use the, there is still the incorrect home planet bug 2010-11-17T15:16:58 ah antbot started up 2010-11-17T15:17:45 Top 10 players: bocsimacko(3634), bix0r4ever(3616), luksian(3522), Raschi(3456), asavis(3452), protocolocon(3452), dmj111(3444), shangas(3420), george(3416), Xrillo(3382) 2010-11-17T15:18:10 incorrect home planet? you mean the off by one in calculating safety from ragebot? 2010-11-17T15:18:57 no, I mean bad code with overloaded variables 2010-11-17T15:19:20 the variables p1 and p2 are not always the home planets 2010-11-17T15:19:28 it needs to be planet[1] planet[2] 2010-11-17T15:19:40 where? 2010-11-17T15:20:12 http://pastebin.com/P2ktiwD6 2010-11-17T15:20:25 line 193 in that version 2010-11-17T15:21:02 line 187 in the repository 2010-11-17T15:21:06 ahh, I see in the ragebot safe planet calc 2010-11-17T15:21:34 antimatroid was for the antisnipe code 2010-11-17T15:21:55 I think mega1 was too... 2010-11-17T15:22:15 hmm, no that doesn't have anything to do with sniping that I can see 2010-11-17T15:22:18 No, mega1 was indifferent 2010-11-17T15:22:24 oh, nm 2010-11-17T15:22:27 misunderstood 2010-11-17T15:22:40 I don't think I'll add antisniping 2010-11-17T15:22:56 I'd rather not add additional constraints 2010-11-17T15:23:01 to the generator 2010-11-17T15:23:17 I already wrote it, and it stays within spec 2010-11-17T15:23:21 instead watch actual map statistics and remove ones that do bad 2010-11-17T15:23:28 That would work too 2010-11-17T15:23:46 But what about when you do the final? Will bad maps be removed then too? 2010-11-17T15:23:56 yes, that is what I'm hoping 2010-11-17T15:24:07 Oh, then who cares? 2010-11-17T15:24:13 right :) 2010-11-17T15:24:42 Okay. If you make the change to the p1 and p2 variables. We can call it good and I can regenerate. 2010-11-17T15:24:44 but of course I don't have anything written to actually monitor draw percentages and such yet so... 2010-11-17T15:25:05 Ha! I do. 2010-11-17T15:25:52 yeah but not integrated with the contest server :) 2010-11-17T15:25:54 You'll need some type of criteria. 2010-11-17T15:26:10 X percent draws with only the top 100 or something 2010-11-17T15:26:24 and is there anything other than draw percentage that should be watched 2010-11-17T15:26:35 yeah, that's my basic idea at the moment 2010-11-17T15:26:41 *** greghaynes has joined #aichallenge 2010-11-17T15:26:41 I don't think so. 2010-11-17T15:27:01 You're really just trying to remove maps that skew results. 2010-11-17T15:27:14 Because not everyone will play every map 2010-11-17T15:27:19 against everyone 2010-11-17T15:27:42 There is the part about getting a fair sampling 2010-11-17T15:28:00 But unless we can agree on map metrics, that's a hugh argument to be had. 2010-11-17T15:28:25 yes, and that is somewhat separate from the 'online' monitoring anyway 2010-11-17T15:28:50 Okay, waiting for r538... 2010-11-17T15:30:39 *** iamluck has quit IRC (Quit: KVIrc 4.0.2 Insomnia http://www.kvirc.net/) 2010-11-17T15:31:57 oh, shoot 2010-11-17T15:32:18 @roulette 2010-11-17T15:32:18 McLeopold: *click* 2010-11-17T15:32:26 I forgot I had planned on fixing the map translation code as well 2010-11-17T15:32:52 ? 2010-11-17T15:33:01 just translate by maxRadius instead of maxRadius + 2 would be the simplest 2010-11-17T15:33:08 is there a problem with that? 2010-11-17T15:33:12 function translate_planets 2010-11-17T15:33:30 no real problem I suppose more cosmetic 2010-11-17T15:33:37 is that to center it so everything is a positive coord? 2010-11-17T15:33:49 the smallest x or y we can get is 2 right now instead of 0 2010-11-17T15:33:50 yes 2010-11-17T15:34:08 well, that's a quick fix 2010-11-17T15:34:25 just wish I had remembered to do it last night 2010-11-17T15:35:32 hmm, whether to do it anyway or just leave it now 2010-11-17T15:35:51 please make both changes now 2010-11-17T15:35:58 pretty please 2010-11-17T15:36:08 ok, I'll go ahead and commit it. 2010-11-17T15:36:21 yeah 2010-11-17T15:36:28 not going to regenerate for probably 8 more hours as I don't have time right now 2010-11-17T15:36:40 I'm going to on my server 2010-11-17T15:36:48 sounds good 2010-11-17T15:36:56 another chance to find problems :) 2010-11-17T15:36:57 I'll post to the forum as well 2010-11-17T15:37:03 *** nullkuhl has quit IRC (Ping timeout: 250 seconds) 2010-11-17T15:37:17 *** sigh has joined #aichallenge 2010-11-17T15:38:29 commited 2010-11-17T15:39:11 sweet 2010-11-17T15:39:17 I really hope that is the last changes 2010-11-17T15:39:24 me too 2010-11-17T15:39:31 I will stop all arguments about maps now. 2010-11-17T15:39:39 heh, good luck 2010-11-17T15:39:59 Well, I started most of them, so, it shouldn't be hard. :) 2010-11-17T15:40:11 *** iamluck has joined #aichallenge 2010-11-17T15:45:30 McLeopold: so maxradius is 15 now? 2010-11-17T15:45:39 yes 2010-11-17T15:45:45 janzert committed 2010-11-17T15:45:50 I'm regenerating for my server 2010-11-17T15:46:03 oh down to 15? 2010-11-17T15:46:15 He said he'd do official later. 2010-11-17T15:46:21 cool, will you put a tar/zip up of those maps? 2010-11-17T15:46:23 lavalamp_: so max dist is 30 2010-11-17T15:46:27 yes 2010-11-17T15:46:33 k... still in a circle? 2010-11-17T15:46:36 thanks 2010-11-17T15:46:52 lavalamp__: yes 2010-11-17T15:47:02 ok 2010-11-17T15:47:12 men I am kinda getting tired of the mapchanges now comeon guys :S 2010-11-17T15:47:12 not that it would matter for my dumb bot :( 2010-11-17T15:47:15 http://zeroviz.us:8080/maps.zip 2010-11-17T15:47:32 2 more weeks roughly 2010-11-17T15:47:34 Error323: this should be final 2010-11-17T15:47:58 I heard that before... 2010-11-17T15:48:07 That's why we were pushing last night and this morning to vette all issues. 2010-11-17T15:48:23 Personnally, I find this maps a lot more interesting! 2010-11-17T15:48:43 Yeah, the line symmetry makes it tough. :) 2010-11-17T15:49:55 well its really nice yes, just the timeconstraint thats becomming more and more of an issue 2010-11-17T15:50:06 Error323: these changes are just to fix 'bugs' introduced by the replacement map generator, so you can consider the feature freeze as being last Saturday 2010-11-17T15:50:22 If the changes were that the size was four times bigger, it would have been very annoying, but here, there's nothing that changes too much from before 2010-11-17T15:50:43 how do you guys figure out how many ships to hold back to defend against hypothetical threats (e.g. ragebot)? 2010-11-17T15:50:46 Create a vote for extending the contest till January? :) 2010-11-17T15:51:09 janzert: I understand, but still I have to regenerate with the mapgen and see how the bot is doing now :] 2010-11-17T15:51:10 That's just killing my bot right now :( 2010-11-17T15:51:35 iamluck: I really don't think any extension is going to happen 2010-11-17T15:51:37 janzert: so is the latest in svn yet? 2010-11-17T15:51:40 No don't extend the contest until January, I've got to work and sleep!!!! 2010-11-17T15:51:44 Error323: yes 2010-11-17T15:52:06 lavalamp_: 100 - distance * growth? :) 2010-11-17T15:52:10 also see above McLeopold posted a zip with a fresh run 2010-11-17T15:52:41 * lavalamp__ rolls eyes 2010-11-17T15:52:54 lavalamp_: I'm just verifying for the planet how much ships the enemy can send to my planets, and how much I can send 2010-11-17T15:53:02 McLeopold: be great if you could post about it in the forums as well 2010-11-17T15:53:03 *** amstan has joined #aichallenge 2010-11-17T15:53:03 *** ChanServ sets mode: +o amstan 2010-11-17T15:53:14 k 2010-11-17T15:53:17 yo 2010-11-17T15:53:18 gotta run now, see ya all 2010-11-17T15:53:24 *** sigh has quit IRC (Remote host closed the connection) 2010-11-17T15:53:27 medrimonia, that seems to make my bot super conservative... 2010-11-17T15:53:35 what's new? 2010-11-17T15:53:50 map generator finalized 2010-11-17T15:54:42 The hard part is to find a way to keep free_ships, without taking too much risks 2010-11-17T15:55:03 what's free_ships to you? 2010-11-17T15:55:28 Ships that you can send without taking the risk of loosing a planet 2010-11-17T15:55:55 antimatroid: you're mod 2010-11-17T15:56:34 *** p4p4 has quit IRC (Quit: ChatZilla 0.9.84 [SeaMonkey 2.0a3/20090223135443]) 2010-11-17T15:56:48 hmm... how do you calculate that? every one of their planets vs every one of yours? 2010-11-17T15:57:29 I've a matrix with the state of every planet during the next SIZEMAP turns 2010-11-17T15:57:52 same here 2010-11-17T15:58:31 And I look at the number of ship of all the planets I will own at distance(source_id,dest_id) 2010-11-17T15:58:33 I can defend against ships in flight no problem... the problem is defending against ships that haven't been launched yet 2010-11-17T15:58:48 And the same for the enemy 2010-11-17T15:59:23 hmmm that's exactly what I was trying to do, maybe I have a bug :/ 2010-11-17T15:59:51 regarding maps: aren't the ones with fewer planets more likely to be drawish? 2010-11-17T16:00:18 I don't think so. 2010-11-17T16:00:30 There is less opportunity for error, I suppose. 2010-11-17T16:00:42 a two planet map is very likely to be a draw 2010-11-17T16:01:02 true, but min is 15 2010-11-17T16:01:25 Janzert mentioned they would monitor maps with lots of draws and remove them. 2010-11-17T16:01:33 3-planet map ought to always be a draw, too. After that I think it depends on the map. 2010-11-17T16:01:50 I've read that. It can be a subjective process. 2010-11-17T16:02:21 lavalamp__: check my pm 2010-11-17T16:02:53 mega1: how about you? do you care about http://www.ai-contest.com/forums/viewforum.php?f=20 ? 2010-11-17T16:04:26 amstan: yes 2010-11-17T16:04:47 *** yasith has quit IRC (Ping timeout: 272 seconds) 2010-11-17T16:04:53 geeze, with all my posting, I look like an organizer... 2010-11-17T16:05:18 McLeopold: it's pretty clear though, you're green, admins are red 2010-11-17T16:05:30 ? 2010-11-17T16:05:39 oh, my color changed today, I didn't know what that meant 2010-11-17T16:05:47 lol 2010-11-17T16:05:52 so.. i deleted the old groups 2010-11-17T16:06:09 and organizers are now admins, trusted ppl that run around are green 2010-11-17T16:07:08 amstan: Never gonna run around... and desert you 2010-11-17T16:07:46 so... as a moderator, you gain some powers to organize the forums and you also have a say in the behind a scenes forum(read: mailing list) 2010-11-17T16:08:12 k 2010-11-17T16:08:45 Weee, I win on the new maps! 2010-11-17T16:09:28 *** McLeopold has quit IRC (Quit: Leaving) 2010-11-17T16:09:51 mega1: k.. you get added, antimatroid's left 2010-11-17T16:10:05 mega1: this is you? http://www.ai-contest.com/forums/memberlist.php?mode=viewprofile&u=305 2010-11-17T16:11:08 amstan: yes, that's me 2010-11-17T16:11:25 mega1: k, cool 2010-11-17T16:12:21 amstan: what you're doing is a good idea, but you may want to do it by placing peolpe in a group other than mods, if possible. Some (not me) could percieve that as some sort of favoritism, esp. if a lot of peolpe with high-ranking bots end up there. 2010-11-17T16:12:46 someone made me not a moderator? 2010-11-17T16:13:07 Zannick: hmm? 2010-11-17T16:13:18 eh 2010-11-17T16:13:27 amstan: why are you restricting this to moderators 2010-11-17T16:13:29 i want to have a different name on the forums anyway 2010-11-17T16:13:33 lavalamp__: the thing is that we need to assume the same ammount of trust for both people 2010-11-17T16:13:43 create a new permission group for that forum and put people in thar rather than giving them administrator powers over the forums 2010-11-17T16:13:53 You can set existing mods/admins to inherit those permissions 2010-11-17T16:14:17 That's fine, but there's no need for them to be mods, is there? 2010-11-17T16:14:23 pgpaskar: i know.. but if they're allowed to post in there it means we care enough about them to consider their suggestions 2010-11-17T16:14:42 pgpaskar: there's also a need for mods, i think those 2 groups intersect 2010-11-17T16:14:45 what does that have to do with moderators 2010-11-17T16:15:02 mods are also people that you trust 2010-11-17T16:15:18 no, mods are people who clean up and take care of spam/bad users in the forums 2010-11-17T16:15:22 *** Mjothvitnir has joined #aichallenge 2010-11-17T16:15:32 trust has to do with favouritism, and its very much separate 2010-11-17T16:15:41 yeah, don't worry about it 2010-11-17T16:15:43 so add people to both groups if you want... but I think they are separate concepts and should be visibly separate 2010-11-17T16:15:50 i'm not offended, i haven't been on in a while 2010-11-17T16:15:56 so like.. i need a name for those mods then 2010-11-17T16:15:58 lol 2010-11-17T16:16:07 s/mods/people that can post/g 2010-11-17T16:16:09 "suggestinators" 2010-11-17T16:16:21 lavalamp__: ew 2010-11-17T16:16:23 lol 2010-11-17T16:16:28 hehe 2010-11-17T16:16:35 No, it's one thing to have mods, but another thing to have special access to a particular forum. The groups may have some people who intersect but they are not the same, or at least, they shouldn;t be :| 2010-11-17T16:16:48 "peanut gallery" 2010-11-17T16:17:09 *** yasith has joined #aichallenge 2010-11-17T16:17:26 "the obsessed" 2010-11-17T16:17:32 k, it'll have to wait till later 2010-11-17T16:17:35 now: mazes 2010-11-17T16:17:42 Holders of the rainstick 2010-11-17T16:17:44 "no comments from the peanut gallery" would imply that that is the rest of us. 2010-11-17T16:17:46 Top 10 players: bocsimacko(3624), bix0r4ever(3600), luksian(3495), asavis(3466), Raschi(3454), dmj111(3438), _Astek_(3438), protocolocon(3430), shangas(3413), george(3412) 2010-11-17T16:17:46 I'm doing it right now 2010-11-17T16:17:53 mazes? 2010-11-17T16:18:12 the forums thing 2010-11-17T16:18:14 lol maze maps 2010-11-17T16:18:49 *** delt0r has joined #aichallenge 2010-11-17T16:19:33 * lavalamp__ scratches head 2010-11-17T16:20:02 *** delt0r___ has quit IRC (Ping timeout: 276 seconds) 2010-11-17T16:21:16 okay, i made an account that matches my name here: http://www.ai-contest.com/forums/memberlist.php?mode=viewprofile&u=1739 2010-11-17T16:21:36 * Zannick back to work 2010-11-17T16:22:12 *** amstan has quit IRC (Ping timeout: 260 seconds) 2010-11-17T16:25:27 *** amstan has joined #aichallenge 2010-11-17T16:25:27 *** ChanServ sets mode: +o amstan 2010-11-17T16:26:14 "discussion panel" or "focus panel" would be a good name 2010-11-17T16:26:52 or "panel" for short 2010-11-17T16:27:43 "board of directors" 2010-11-17T16:27:57 stakeholders? lol 2010-11-17T16:28:19 pgpaskar: where are you? 2010-11-17T16:28:30 Hi 2010-11-17T16:28:32 at home 2010-11-17T16:28:49 i should be studying for stats 2010-11-17T16:28:51 pgpaskar: you didn't come to mazes, you should come to mazes 2010-11-17T16:32:21 yeah... but im going to choose not to fail stat today 2010-11-17T16:32:38 but! mazes! 2010-11-17T16:32:57 mazes! http://csclub.uwaterloo.ca/events/MC4061-2010-11-17-04:30_PM 2010-11-17T16:37:32 sounds pretty cool 2010-11-17T16:38:22 *** amstan has quit IRC (Ping timeout: 272 seconds) 2010-11-17T16:40:25 *** Frontier has joined #aichallenge 2010-11-17T16:41:17 *** narnach has quit IRC (Quit: Leaving.) 2010-11-17T16:41:53 relevant: http://ai-contest.com/forums/viewtopic.php?f=20&t=1124 2010-11-17T16:42:03 to previous discussion, anyway 2010-11-17T16:47:44 *** Palmik has quit IRC (Remote host closed the connection) 2010-11-17T16:48:34 *** Rubicon-|-Cross has joined #aichallenge 2010-11-17T16:56:54 *** Mathnerd314 has joined #aichallenge 2010-11-17T16:57:32 does anyone know if there is any plan to have the contest server use the new map generator (prior to the "real" contest)?? 2010-11-17T17:00:22 Mjothvitnir: I think it will happen today. janzert is the one to ask. 2010-11-17T17:02:20 looks like janzert is absent, thanks. 2010-11-17T17:03:43 *** bhasker_ has joined #aichallenge 2010-11-17T17:06:06 *** sigh has joined #aichallenge 2010-11-17T17:06:11 *** yasith has quit IRC (Ping timeout: 255 seconds) 2010-11-17T17:06:50 *** bhasker has quit IRC (Ping timeout: 276 seconds) 2010-11-17T17:07:53 *** McLeopold has joined #aichallenge 2010-11-17T17:17:46 Top 10 players: Raschi(3704), bocsimacko(3658), bix0r4ever(3587), _Astek_(3501), asavis(3488), luksian(3468), protocolocon(3459), dmj111(3454), shangas(3439), george(3430) 2010-11-17T17:21:02 *** smellyhippy has quit IRC (Ping timeout: 255 seconds) 2010-11-17T17:21:42 *** fawek has quit IRC (Read error: Connection reset by peer) 2010-11-17T17:24:25 so I was analyzing some of the games my bot played, to see what it's doing right/wrong... and it seems like it's behaving exactly as the default AI, so I think I uploaded the wrong file.. 2010-11-17T17:24:39 but analyzing the play, there seems to be a counting error 2010-11-17T17:25:11 should I attempt to make my bot utilize this error: or give details to someone, so it can be fixed? 2010-11-17T17:26:50 the right thing to do would be to have it fixed 2010-11-17T17:27:18 *** smellyhippy has joined #aichallenge 2010-11-17T17:27:18 *** smellyhippy has joined #aichallenge 2010-11-17T17:28:47 counting error? 2010-11-17T17:29:09 can you show us? 2010-11-17T17:29:29 not until I get official word from one of the people running it :P 2010-11-17T17:30:19 what kind of error are we talking about here? battle outcome or something else? 2010-11-17T17:30:29 if you resubmit your bot: do the old games still remain in history?(just not effect rank?) 2010-11-17T17:30:56 curious. 2010-11-17T17:31:00 they will show up in the history of the people you played but not yours, I think 2010-11-17T17:31:12 are you talking about opposing fleets arriving at the same time at a planet? 2010-11-17T17:31:17 a neutral planet 2010-11-17T17:31:23 zannick, ya 2010-11-17T17:31:39 lava: do the re-plays still exist though, is what I meant? 2010-11-17T17:31:40 that logic is tricky to get correct 2010-11-17T17:31:45 yes they do. 2010-11-17T17:31:45 yeah, the result is highest player's total minus second highest player's total 2010-11-17T17:32:17 it seems weird, but that's the way that case was settled on a month and a half ago 2010-11-17T17:32:20 ok, cool. going to resubmit my proper bot now.. just didn't want to lose the history links 2010-11-17T17:32:25 there is no way the organizers could change the engine now, even if there were a bug 2010-11-17T17:32:47 time to head home 2010-11-17T17:32:48 brb 2010-11-17T17:33:11 lavalamp__: i disagree. how handling of cases shouldn't change, but if there's a bug that crashes the engine, that'd be fixed 2010-11-17T17:33:29 ok THAT kind of bug they'd fix ^^ 2010-11-17T17:33:42 but I don't think burny was talking about that 2010-11-17T17:34:04 there were very strange bugs in that case back at the beginning 2010-11-17T17:34:52 *** jaspervdj has quit IRC (Quit: NEVER GONNA GIVE YOU UP NEVER GONNA LET YOU DOWN) 2010-11-17T17:35:51 Would your bot take the 85 or the 30 in the middle? My bot got it right: http://ai-contest.com/visualizer.php?game_id=7728771 2010-11-17T17:36:08 Initially I thought it was wrong but nope... 2010-11-17T17:36:18 on first move? 2010-11-17T17:36:33 pastebin the gamestate lavalamp__? 2010-11-17T17:36:35 yeah first move 2010-11-17T17:36:50 attacking an 85 in teh middle on first turn is a foresure loss 2010-11-17T17:37:04 30 is in the middle, 85 nearby 2010-11-17T17:37:17 attacking a 30 in mid on first move isn't a good move either 2010-11-17T17:37:36 JamesMG: no tools with me atm 2010-11-17T17:37:50 the 85 would have to be atleast 8units from the middle to be a good move 2010-11-17T17:38:06 and starting positions atleast 20units apart 2010-11-17T17:38:11 mine goes for the 85 2010-11-17T17:38:21 yeah going for the 85 is correct on that map 2010-11-17T17:38:37 didn't look right to me at first 2010-11-17T17:38:39 http://pastebin.com/579Ys0KG 2010-11-17T17:38:45 in case anyone else wants it 2010-11-17T17:38:54 ^^ 2010-11-17T17:39:12 JamesMG do you knapsack it? 2010-11-17T17:39:27 no, ordered list of planets 2010-11-17T17:39:31 k 2010-11-17T17:39:44 I can run it through my knapsack minimax bot... 2010-11-17T17:39:54 eh, it's ok 2010-11-17T17:40:07 it's pretty clearly better, I think 2010-11-17T17:40:35 knapsack bot goes 85 and 10 2010-11-17T17:40:44 the 9cost and 10cost 3growth are very likely to be better choices, depending on the placement of stuff 2010-11-17T17:41:08 I don't see any 30population planets though 2010-11-17T17:41:17 middle planet is 2010-11-17T17:41:48 oh.. that one is shorter lol 2010-11-17T17:41:52 yeah 2010-11-17T17:41:57 planetID 0 2010-11-17T17:42:14 best bet on that map, is to wait 1frame, and then take the 30 2010-11-17T17:42:31 haha 2010-11-17T17:42:34 unless opponent takign hte 9 or the 10, prevents you from safely taking hte 85 2010-11-17T17:42:39 http://zeroviz.us:8080/map_r528_28.txt.html 2010-11-17T17:43:53 nice, dmj made exactly my move ^^ 2010-11-17T17:44:05 ouch, shows my bot's incompentance after turn 1 for sure 2010-11-17T17:44:34 oh, nver mind.. the 85 isn't near the middle 2010-11-17T17:44:40 the 85 is best move for sure 2010-11-17T17:44:58 it's like 10steps from middle 2010-11-17T17:45:23 and 3ish from starting point 2010-11-17T17:45:38 lol, Error323 lets a +3 planet with 0 ships on it sit there in one of those games 2010-11-17T17:46:28 is the math on that correct? wow 2010-11-17T17:46:33 I wish there was a pairing of dmj and E323 on that map, I'd like to see their moves go head to head 2010-11-17T17:46:47 I guess it is..., but weird 2010-11-17T17:47:07 I go for the neutral 10 on the side 2010-11-17T17:47:41 actually ya.. the 10 and the 85 both on first turn is best 2010-11-17T17:47:50 not the 85 2010-11-17T17:48:17 burny: yeah and that's what I do :) 2010-11-17T17:48:21 and the 9 on 3rd frame 2010-11-17T17:48:38 wish I could request that map on TCP now 2010-11-17T17:48:39 McLeopold, do you do knapsack or priority list or something else? 2010-11-17T17:48:45 made a change that should make my bot slightly less dumb on it 2010-11-17T17:48:49 actually.. second frame for the 9 2010-11-17T17:48:56 A weird combination of both 2010-11-17T17:49:01 haha 2010-11-17T17:51:27 kind of interesting that on zeroviz most games by far go to 200 turns 2010-11-17T17:51:42 yeah, bots are better 2010-11-17T17:51:48 or really dumb :) 2010-11-17T17:52:54 bah, my bot gives a compile error when I submit it.. 2010-11-17T17:53:02 it will Email me the compiler output? 2010-11-17T17:54:12 takes like an hour though? 2010-11-17T17:54:34 burny: I don't get those emails :/ 2010-11-17T17:54:56 how do you fix it then? 2010-11-17T17:55:01 @pastewars 2010-11-17T17:55:02 McLeopold: pastewars = http://visualizer.naktibalda.lt/ 2010-11-17T17:55:24 geussing? :/ 2010-11-17T17:55:38 if janzert show up you can ask him... check your spam folder though 2010-11-17T17:55:54 like.. guess that it can't use C++ style comments.. 2010-11-17T17:56:03 or that you can't declare a non namespaced class.. 2010-11-17T17:56:05 or some such? 2010-11-17T17:56:15 http://visualizer.naktibalda.lt/573 2010-11-17T17:56:23 ^^ my openings 2010-11-17T17:56:33 11 to the neutral only, wait to snipe 2010-11-17T17:56:52 do you take the 30 on frame2 then? 2010-11-17T17:57:03 only because I'm reacting 2010-11-17T17:57:28 My older bot didn't have antisniping code, so it took it with some extra 2010-11-17T17:57:53 anti sniping? 2010-11-17T17:58:03 yeah, looks like antbot likes to snipe scotty 2010-11-17T17:58:10 :) 2010-11-17T17:58:38 burny: anti sniping is the same as "don't go for the 30 in the center, that's a bad move" 2010-11-17T17:58:38 if you go to middle on first frame, and opponent doesn't take the 85(or do something stupid).. you lose 2010-11-17T17:59:04 ahhh 2010-11-17T17:59:07 yeah, that's why the 85 is bad, you give your opponent the opportunity to take the center 2010-11-17T17:59:22 yeah I need to put in some that-planet-can-be-sniped code... 2010-11-17T17:59:31 same 2010-11-17T17:59:39 after I fix so many other problems 2010-11-17T17:59:46 JamesMG: like this one: http://zeroviz.us:8080/canvas?game_id=1457 2010-11-17T17:59:47 the 85 should beat the 30 though, due to distances 2010-11-17T17:59:49 I wrote that code for the map generator 2010-11-17T18:00:04 oh god lavalamp__ :( 2010-11-17T18:00:22 ^^ ragebot fail 2010-11-17T18:00:27 how long ago was that? 2010-11-17T18:00:35 I bugged my code this morning and fixed it about an hour ago 2010-11-17T18:00:49 yesterday night? 2010-11-17T18:00:50 http://pastebin.com/vGdn1pyC 2010-11-17T18:00:54 oh 2010-11-17T18:00:56 look for the snipable function 2010-11-17T18:01:25 *** bhasker_ has quit IRC (Quit: bhasker_) 2010-11-17T18:01:43 p1 and p2 are your planet/closest enemy planet? 2010-11-17T18:01:55 p1 and p2 are the home planets 2010-11-17T18:01:59 p3 is the neutral 2010-11-17T18:02:17 it returns true if it is a bad target 2010-11-17T18:02:27 ah 2010-11-17T18:03:00 assuming the enemy hits 1 turn later with the same amount of ships 2010-11-17T18:03:10 so it is only useful for the first turn 2010-11-17T18:03:21 You would need a more generic function throughout the game 2010-11-17T18:03:41 err...not 1 turn later 2010-11-17T18:03:48 but 1 turn or later 2010-11-17T18:03:56 hm... I wonder if my bot would do the same thing if the payoff table engine were on... 2010-11-17T18:03:57 depending on how much closer you are 2010-11-17T18:06:26 JamesMG: remember my center fail? 2010-11-17T18:06:31 yeah McLeopold 2010-11-17T18:06:43 That's were some of my formulas break down 2010-11-17T18:06:52 ah 2010-11-17T18:06:53 somehow my bot thought those planets were snipable 2010-11-17T18:07:14 *** phreeza_ has quit IRC (Quit: Computer has gone to sleep) 2010-11-17T18:07:28 I'd have to see it again to understand I think 2010-11-17T18:07:55 http://zeroviz.us:8080/canvas?game_id=2799 2010-11-17T18:08:15 that map is tough 2010-11-17T18:08:15 oh, yeah I see 2010-11-17T18:08:27 Perhaps 100 to the good side of the map? 2010-11-17T18:08:32 if you don't make the leap out to those good planets... 2010-11-17T18:08:40 which looks really like a bad move at first glance 2010-11-17T18:08:50 wow that is a massive center fail :/ 2010-11-17T18:08:54 yeah, it's almost worth losing your home planet 2010-11-17T18:09:22 Turn 1: it looks snipable, Turn 2: oops too late, go for other stuff 2010-11-17T18:09:26 which map? http://imgur.com/Zqv8F.png 2010-11-17T18:09:51 *** narnach has joined #aichallenge 2010-11-17T18:10:21 that map is poopers 2010-11-17T18:10:27 *** amstan has joined #aichallenge 2010-11-17T18:10:27 *** ChanServ sets mode: +o amstan 2010-11-17T18:10:33 McLeopold: I'm surprised at the order for 35 ships to 26 instead of 25 on turn 2 2010-11-17T18:10:50 what map is that? I want to see if my bot played on it at all 2010-11-17T18:11:08 newmap68.txt 2010-11-17T18:11:26 yes, your in the list 2010-11-17T18:11:31 http://zeroviz.us:8080/newmap68.txt.html 2010-11-17T18:11:43 just sort by name 2010-11-17T18:11:58 a lot of 200 turns for lavalamp :) 2010-11-17T18:12:01 E323 tried to snipe me like I would've expected you to do McLeopold 2010-11-17T18:12:06 http://zeroviz.us:8080/canvas?game_id=5255 2010-11-17T18:12:06 MyBot.cc:267: error: 'sqrt' was not declared in this scope 2010-11-17T18:12:09 how do I fix that? 2010-11-17T18:12:10 yeah crap I fail on that map 2010-11-17T18:12:23 #include ? 2010-11-17T18:12:55 lavalamp_: only when player 1 2010-11-17T18:13:07 whoa what, my old bot makes different moves depending on which starter planet it gets 2010-11-17T18:13:10 which means you have a planet id bias? 2010-11-17T18:13:34 http://zeroviz.us:8080/canvas?game_id=4262 2010-11-17T18:13:53 I suppose I must, strange 2010-11-17T18:14:05 I still have it in some places 2010-11-17T18:14:07 wow I do have a lot of 200 turn games 2010-11-17T18:14:25 planet 25 and 26 are the same distance 2010-11-17T18:14:28 planetID bias explains why I hit 25 instead of 26 2010-11-17T18:14:32 so, which is better? 2010-11-17T18:14:37 but I don't hit either from the top starting position 2010-11-17T18:14:40 *** bhasker has joined #aichallenge 2010-11-17T18:14:48 on turn 1 anyways 2010-11-17T18:14:54 it's because I have epic battles that take 150+moves to result in a clear winner 2010-11-17T18:16:10 hm well good to know that everyone has problems on that map 2010-11-17T18:16:35 Solve that map and win all line symmetric maps. :) 2010-11-17T18:17:16 when the server changes things for player two, does it change all planet IDs or just the home worlds? 2010-11-17T18:17:47 Top 10 players: Raschi(3658), bocsimacko(3636), bix0r4ever(3572), asavis(3468), luksian(3439), _Astek_(3437), protocolocon(3432), dmj111(3426), shangas(3412), george(3398) 2010-11-17T18:17:49 *** bhasker_ has joined #aichallenge 2010-11-17T18:19:58 just which planet you own 2010-11-17T18:20:14 *** Zoost has quit IRC (Ping timeout: 240 seconds) 2010-11-17T18:20:17 so, only the owner numbers are reversed 2010-11-17T18:20:18 *** Zoost has joined #aichallenge 2010-11-17T18:20:36 oh it doesn't reorder planets at all? 2010-11-17T18:20:58 *** bhasker has quit IRC (Ping timeout: 272 seconds) 2010-11-17T18:21:03 I guess that wouldn't make sense 2010-11-17T18:21:22 so you can figure out if you're "really" player 2 based on which planet you start out with? 2010-11-17T18:21:43 so roughly a thousand people have entered the default bot with zero changes? 2010-11-17T18:21:59 burny, would'nt surprise me 2010-11-17T18:22:17 I did, but only by accident 2010-11-17T18:22:44 then when I saw the first game, wa slike "wtf?.. why is it sending 50 instead of 37?" 2010-11-17T18:23:21 howdydo 2010-11-17T18:23:49 janzert: the contest will end at 27th right? 2010-11-17T18:23:59 at 11:59 2010-11-17T18:24:05 IIRC 2010-11-17T18:24:17 that's what I recall also 2010-11-17T18:24:18 yeah but i've been reading stuff on the forums 2010-11-17T18:24:27 about ppl wanting to postpone 2010-11-17T18:24:45 http://zeroviz.us:8080/canvas?game_id=2644 2010-11-17T18:24:49 but imo that's a bad idea 2010-11-17T18:24:50 postpone why? 2010-11-17T18:24:55 cuz of mapchange 2010-11-17T18:25:13 pgpaskar: so.. what's going on? 2010-11-17T18:25:18 Hi 2010-11-17T18:25:22 is there even any official word that the new changes will be the official maps? 2010-11-17T18:25:24 McLeopold: that's what I expect from you :) 2010-11-17T18:25:25 Studying... stat lol 2010-11-17T18:25:26 I want to postpone but it's because I'm behind schedule XP 2010-11-17T18:25:27 also reddit 2010-11-17T18:25:37 pgpaskar: i see you made a group and stuff 2010-11-17T18:25:39 usually with this kind of thing, you don't get 'real' maps 2010-11-17T18:25:41 JamesMG: that was the "dumb" bot just going for it. 2010-11-17T18:25:53 yes 2010-11-17T18:25:55 The planet's can't be sniped 2010-11-17T18:26:04 pgpaskar: what's up next term? 2010-11-17T18:26:12 The growth is too high compared with the neutral ships 2010-11-17T18:26:15 cs cs cs math math, a deliciously difficult term 2010-11-17T18:26:16 and certainly no clear indication of what the density, possible costs, etc. would be 2010-11-17T18:26:22 pgpaskar: no coop? 2010-11-17T18:26:35 burny: yes, the new maps will be final 2010-11-17T18:26:38 nope 2010-11-17T18:26:42 McLeopold: proof I need to work on my defense/reinforcement code 2010-11-17T18:26:42 aww 2010-11-17T18:26:43 not next term anyway 2010-11-17T18:26:49 summer i have coop 2010-11-17T18:27:06 i am seq B 2010-11-17T18:27:06 final? er.. do you mean, these exact maps are going to be used for final judging? 2010-11-17T18:27:23 2x school work 2x school work school works school work ... 2010-11-17T18:27:23 pgpaskar: k 2010-11-17T18:27:34 *** JamesMG has quit IRC (Quit: Leaving.) 2010-11-17T18:27:34 burny: no but the generator that made them is 2010-11-17T18:27:43 burny: no, the generator is final. The maps will look like the ones on my server 2010-11-17T18:28:08 McLeopold: you are absolutely positive? 2010-11-17T18:28:26 did they fix the "line symetry, has all the 'middle' planets being on same side?" bug? 2010-11-17T18:28:31 I'm 99% sure. Review the logs here with me a Janzert 2010-11-17T18:28:50 burny: I started that theory, and I think I was wrong. 2010-11-17T18:28:56 burny: so no fix needed 2010-11-17T18:29:01 no, you aren't wrong 2010-11-17T18:29:48 unless it was fixed since the code I was shown 2010-11-17T18:30:06 burny: we've reviewed the code extensivly and had many a math argument. I positive it's a good, fair generator. 2010-11-17T18:30:22 ya, it's fair regardless 2010-11-17T18:30:25 burny: any bias left is too small to worry about 2010-11-17T18:30:39 the bias is in the middle.. so it's equal for both teams 2010-11-17T18:30:42 burny: the code was updated about 4 hours ago 2010-11-17T18:31:02 ahh.. I was looking at it about 45hours ago 2010-11-17T18:32:02 *** Prillicy has joined #aichallenge 2010-11-17T18:32:13 pgpaskar: hmmm, hold on, i don't like admins being in global mods too, i'll just give them mod powers 2010-11-17T18:33:06 *** Naktibalda has joined #aichallenge 2010-11-17T18:33:21 *** IADaveMark has joined #aichallenge 2010-11-17T18:33:45 Yo... I'm assuming that big list over there means there are people in-channel? 2010-11-17T18:34:01 (or not) 2010-11-17T18:34:19 lurkers, all 2010-11-17T18:34:27 Figures. 2010-11-17T18:34:49 What's the easiest way to either log or debug a C++-based bot? 2010-11-17T18:35:09 Just for the helluvit I tried cout but that caused a timeout 2010-11-17T18:35:20 IADaveMark: well i wrote my own assert functions with a logger 2010-11-17T18:35:29 log4cpp ? 2010-11-17T18:35:32 dave, what level of debuging? 2010-11-17T18:35:49 Best case, I would love to step through my code. *hmpf* 2010-11-17T18:35:53 *** Utkarsh has quit IRC (Ping timeout: 245 seconds) 2010-11-17T18:36:06 But barring that, I just want to be able to write values to the console or to a text file. 2010-11-17T18:36:08 the 'best' way, is to write your own game emulator, and run AI vs AI locally, and have a graphical viewer, which includes visual metrics of the descision process 2010-11-17T18:36:13 *** Utkarsh has joined #aichallenge 2010-11-17T18:36:36 burny: yea, though too lazy for it :P 2010-11-17T18:36:43 I don't want to bother writing something new like an emulator 2010-11-17T18:36:43 same lol 2010-11-17T18:37:11 It's taken me this long to get around to writing a bot because I'm too busy writing AI for a living. 2010-11-17T18:37:18 I just changed/added like 20lines to the default AI, just to see how good/bad that will do 2010-11-17T18:37:38 IADaveMark: you can write a test driver to pass in one game state to your bot, and then catch it in the debugger. 2010-11-17T18:37:39 I just write to stderr and use planetwarrior (which shows stderr) or as sujjested you use a logger and write to a file 2010-11-17T18:37:58 PlanetWarrior is that thing I see on the forums that someone wrote? 2010-11-17T18:38:09 it is one of them yes 2010-11-17T18:38:21 well, I would wager that there exists a bot which consists of a variety of functions that just give huristics on planets/fleets. and plays a near perfect game in under 50lines 2010-11-17T18:38:32 So no way that people have found to simply step through the code, eh? 2010-11-17T18:38:42 Ack... brb... pizza 2010-11-17T18:38:59 "ack".. "pizza"?.. those words don't go together.. 2010-11-17T18:39:09 Oh wow. The problem I'd been looking for the past half hour or so was I had int i = 0 when I meant to have i = 0 in a loop since I checked i after the loop. I kept getting the value i previously was once the i I wanted went out of scope. 2010-11-17T18:39:32 ouch 2010-11-17T18:40:16 you need to make that error more often os it will rise higher on your list of things to check. ;-) 2010-11-17T18:40:52 or.. neer ever, make loop counter be outside of scope 2010-11-17T18:41:01 and explicitly set teh value inside the loop 2010-11-17T18:41:12 *** Utkarsh has quit IRC (Ping timeout: 265 seconds) 2010-11-17T18:41:31 anyone arond yet, who's running this contest? 2010-11-17T18:41:34 well, yeah, never making the error works too, I guess. 2010-11-17T18:41:40 burny: hmm? 2010-11-17T18:42:33 Back... the "ack" was because I had to jump up to receive the aforementioned pizza 2010-11-17T18:43:00 but analyzing the play, there seems to be a counting error 2010-11-17T18:43:00 should I attempt to make my bot utilize this error: or give details to someone, so it can be fixed? 2010-11-17T18:43:07 So what do you use to log to a file? Any suggestions? 2010-11-17T18:43:35 burny: the three-way battle code? i thought i told you that was fine 2010-11-17T18:45:24 *** Utkarsh has joined #aichallenge 2010-11-17T18:45:49 zannick: if I was refering to exaclty the thing you are refering to.. is it possible to take advantage of that? :P 2010-11-17T18:46:55 so, if you're okay with reading python, the engine code is here: http://code.google.com/p/ai-contest/source/browse/trunk/planet_wars/backend/engine.py 2010-11-17T18:47:19 you want lines 154-180, fight_battle 2010-11-17T18:47:57 landing fleets and defending fleets are added together by player, and the biggest fleet wins, subtracting the second biggest fleet 2010-11-17T18:48:10 if tied, result is 0 ships and owner doesn't change 2010-11-17T18:48:31 in short, yes, you can take advantage of knowledge of the game engine 2010-11-17T18:48:51 *** nullkuhl has joined #aichallenge 2010-11-17T18:54:21 wow.. my bot lost 10games in 6seconds.. 2010-11-17T18:54:25 that's gota be a new record! 2010-11-17T18:55:03 Anyone here in the game industry? 2010-11-17T18:55:09 * burny is 2010-11-17T18:55:20 *** rebelxt has joined #aichallenge 2010-11-17T18:55:53 Where at? 2010-11-17T18:56:00 canada 2010-11-17T18:56:17 Hmmm... obviously wary of narrowing that down? ;-) 2010-11-17T18:56:27 McLeopold: yo Leopold, are you happy with today's map gen changes? noticed you generated a new set today 2010-11-17T18:56:44 Yes, for the most part. 2010-11-17T18:56:50 it's a pretty small place dave :P 2010-11-17T18:57:01 Canada? 2010-11-17T18:57:16 clearly my sarcasim fails :( 2010-11-17T18:57:17 few hours from toronto 2010-11-17T18:57:34 (I knew what you were talking about... clearly my sarcasm fails,) 2010-11-17T18:57:36 McLeopold: sweet.. are they planning to push these maps to official servers soon? 2010-11-17T18:57:43 Going to GDC? 2010-11-17T18:57:45 janzert said sometime tonight 2010-11-17T18:58:02 no 2010-11-17T18:58:08 cool 2010-11-17T18:58:09 never been to gdc 2010-11-17T18:58:20 Serious bummer. 2010-11-17T18:58:43 whoa, bhasker at #2? 2010-11-17T18:58:53 I'm one of the AI Summit advisors and I'm alternating browser tabs between this and the planning website. 2010-11-17T18:58:55 oh, 1 game 2010-11-17T18:59:23 Gonna be some good stuff again this year. 2010-11-17T19:00:07 cool 2010-11-17T19:00:26 @ranking 2010-11-17T19:00:29 @top10 2010-11-17T19:00:30 pgpaskar: I do not know about 'top10', but I do know about these similar topics: 'tcp' 2010-11-17T19:00:33 pgpaskar: Top 10 players: Raschi(3651), bocsimacko(3636), bix0r4ever(3588), asavis(3483), _Astek_(3453), protocolocon(3447), luksian(3436), dmj111(3431), shangas(3418), george(3402) 2010-11-17T19:00:33 * pgpaskar slaps contestbot 2010-11-17T19:00:38 :] 2010-11-17T19:01:06 *** IADaveMark has quit IRC (Quit: Page closed) 2010-11-17T19:01:34 can we change contestbot's name to ragebot? 2010-11-17T19:02:17 ./cry all my improvements fail 2010-11-17T19:02:32 v23? 2010-11-17T19:02:39 you're winning... 2010-11-17T19:02:50 McLeopold: hmm, idk 2010-11-17T19:02:55 yea 2010-11-17T19:02:55 McLeopold: i already have the subdomain too 2010-11-17T19:02:59 thats my old submission 2010-11-17T19:03:07 it does far better than my new improved version:-\ 2010-11-17T19:03:09 :( 2010-11-17T19:03:17 well theoritically my new one should be better 2010-11-17T19:03:19 i believe 2010-11-17T19:03:31 amstan: what subdomain? 2010-11-17T19:03:38 @topic 2 2010-11-17T19:03:38 amstan: (topic []) -- Returns the topic for . is only necessary if the message isn't sent in the channel itself. 2010-11-17T19:03:39 but 0.23 still beats the crap out of my new version 2010-11-17T19:03:40 ragebot.com? 2010-11-17T19:04:01 McLeopold: contestbot.hypertriangle.com 2010-11-17T19:04:10 oh 2010-11-17T19:04:48 you have to admin it would be funny 2010-11-17T19:04:50 bhasker, i'll run my latest on leopold's server tonight - interested to see how we match up now.. 2010-11-17T19:05:31 @homepage 2010-11-17T19:05:32 pgpaskar: Run as fast as you can and don't look back. 2010-11-17T19:05:39 amstan: you should map that to the homepage 2010-11-17T19:05:43 of contestbot 2010-11-17T19:05:50 or @asl 2010-11-17T19:05:52 lets see how it goes rebel, my new version blows, my last submission is still the best bot i have:-\ 2010-11-17T19:05:53 *** amstan has quit IRC (Quit: Konversation terminated!) 2010-11-17T19:05:55 *** Prillicy_ has joined #aichallenge 2010-11-17T19:05:59 *** amstan has joined #aichallenge 2010-11-17T19:05:59 *** ChanServ sets mode: +o amstan 2010-11-17T19:06:02 *** Prillicy_ has quit IRC (Client Quit) 2010-11-17T19:06:03 to "1/robot/contestbot.hyerptriangle.com" 2010-11-17T19:07:13 *** Prillicy has quit IRC (Ping timeout: 250 seconds) 2010-11-17T19:07:27 *** davidjliu has joined #aichallenge 2010-11-17T19:07:28 bhasker_: i've only managed a very minor improvement over my 1.5-old submission.. better fleet scheduling/cancellation logic 2010-11-17T19:07:47 ah k 2010-11-17T19:08:03 i added cancellation and stuff and scoring in the hopes that it will improve my bots decision making 2010-11-17T19:08:07 but its worse:-\ 2010-11-17T19:08:10 :( 2010-11-17T19:10:58 *** amriedle has joined #aichallenge 2010-11-17T19:11:33 hehe, same. i added a bunch of what-if checks (i.e. can opponent attack a better planet if i attack X?) just to make the bot worse once again ;-) reckless bot still rules! 2010-11-17T19:12:21 looking forward to seeing post mortems from all the top guys soon 2010-11-17T19:12:55 bhasker: http://zeroviz.us:8080/canvas?game_id=6059 2010-11-17T19:15:44 interesting game 2010-11-17T19:15:47 lot of back and forth 2010-11-17T19:15:56 I wonder if I could have pulled it off? 2010-11-17T19:16:01 it was very close 2010-11-17T19:16:18 nice game 2010-11-17T19:16:26 at the end though i had a 6 growth advantage 2010-11-17T19:16:34 i doubt you would have recovered easily from there 2010-11-17T19:16:40 lost by 28 2010-11-17T19:16:50 but you had no major attacks in progress 2010-11-17T19:16:55 so that lead would have grown 2010-11-17T19:17:08 and i don't see any major vulnerable planets of mine 2010-11-17T19:17:17 that you could hit 2010-11-17T19:17:47 Top 10 players: Raschi(3654), bocsimacko(3639), bix0r4ever(3590), asavis(3485), _Astek_(3455), protocolocon(3449), luksian(3439), dmj111(3433), shangas(3422), george(3402) 2010-11-17T19:18:48 its just depressing that i can't seem to improve on this version of my bot:-\ 2010-11-17T19:19:22 what's your official ranking with it? 2010-11-17T19:19:27 so the best bots: aside from superior reaction times.. or on-par with human players? 2010-11-17T19:20:04 47ish 2010-11-17T19:20:16 just fixed one bug and uploaded the version now 2010-11-17T19:20:23 *** narnach has quit IRC (Ping timeout: 264 seconds) 2010-11-17T19:20:30 47 is not bad at all 2010-11-17T19:20:32 been around in the 30s-40's for last few days 2010-11-17T19:20:40 *** Utkarsh has quit IRC (Ping timeout: 240 seconds) 2010-11-17T19:21:11 its topping the mcleopolds server now 2010-11-17T19:21:36 @tcp 2010-11-17T19:21:37 rebelxt: tcp = (#1) http://www.benzedrine.cx/planetwars/, or (#2) http://72.44.46.68/, or (#3) http://zeroviz.us:8080, or (#4) 98.247.248.39:995 (zeroviz.us) 2010-11-17T19:21:54 nice 2010-11-17T19:22:12 though it hasn't gone up against as-z05 yet 2010-11-17T19:22:17 once that happens it will probably drop 2010-11-17T19:22:22 that bot beats mine quite often 2010-11-17T19:23:08 new upload first game lost http://ai-contest.com/profile.php?user_id=5822 :( 2010-11-17T19:23:44 you beat "a" most of the games 2010-11-17T19:24:21 0.23 version yes 2010-11-17T19:25:19 *** Utkarsh has joined #aichallenge 2010-11-17T19:25:25 *** narnach has joined #aichallenge 2010-11-17T19:26:42 bhasker: I'm trying something stupid 2010-11-17T19:27:33 *** amstan has quit IRC (Ping timeout: 245 seconds) 2010-11-17T19:28:08 wow, that was boring 2010-11-17T19:30:32 *** Migi32 has quit IRC (Quit: www.ai-contest.com) 2010-11-17T19:30:38 http://zeroviz.us:8080/canvas?game_id=6122 2010-11-17T19:30:39 heh 2010-11-17T19:30:56 need to find a way to figure out how to swap planets 2010-11-17T19:31:11 there should be some way to make my bot say hey his planet has higher growth 2010-11-17T19:31:13 let me swap 2010-11-17T19:31:13 if growth > my_growth: do it 2010-11-17T19:31:23 yea but sadly the way i decide available ships 2010-11-17T19:31:29 prevents me from launching everything 2010-11-17T19:31:40 when i tinker with that 2010-11-17T19:31:43 it does stupid stuff 2010-11-17T19:31:52 *** Utkarsh has quit IRC (Ping timeout: 240 seconds) 2010-11-17T19:31:56 like there 2010-11-17T19:31:58 I like the one attack to the back line. 2010-11-17T19:32:15 i have enough ships in the middle 2010-11-17T19:32:18 to really take your middle 2010-11-17T19:32:20 for at least a turn 2010-11-17T19:32:37 if i launch from the middle 2010-11-17T19:32:40 *** rebelxt has left #aichallenge 2010-11-17T19:32:41 at all your middle planets 2010-11-17T19:32:44 i can swap 2010-11-17T19:32:52 yeah, you probably could have one 2010-11-17T19:32:53 before you can reinforce 2010-11-17T19:33:09 yea hard to get my bot to do such stuff 2010-11-17T19:33:20 probably need 2 step look ahead 2010-11-17T19:33:23 to do something like that 2010-11-17T19:33:34 you could potentially launch at all my planets 2010-11-17T19:33:42 and i gain 2 and lose all 2010-11-17T19:37:20 *** Utkarsh has joined #aichallenge 2010-11-17T19:38:20 server down? 2010-11-17T19:38:38 *** lavalamp__ has quit IRC (Quit: Page closed) 2010-11-17T19:39:44 just the web part 2010-11-17T19:39:49 it does that :( 2010-11-17T19:41:30 *** Utkarsh has quit IRC (Ping timeout: 240 seconds) 2010-11-17T19:42:10 ah k 2010-11-17T19:42:33 so the main diffrerence between the two versions of my bot seems to be how i decide how many ships are available 2010-11-17T19:42:45 for some reason my inaccurate method in 0.23 does better 2010-11-17T19:42:51 rather unscientific manner 2010-11-17T19:44:52 *** mceier has quit IRC (Quit: leaving) 2010-11-17T19:45:21 time to try a hybrid version 2010-11-17T19:48:02 *** Utkarsh has joined #aichallenge 2010-11-17T19:53:24 *** davidjliu has quit IRC (Ping timeout: 265 seconds) 2010-11-17T19:57:35 *** McLeopold has quit IRC (Quit: Leaving) 2010-11-17T19:57:49 *** amstan has joined #aichallenge 2010-11-17T19:57:49 *** ChanServ sets mode: +o amstan 2010-11-17T19:59:13 *** Utkarsh has quit IRC (Ping timeout: 240 seconds) 2010-11-17T20:00:09 morning 2010-11-17T20:05:14 *** Utkarsh has joined #aichallenge 2010-11-17T20:10:48 *** _Utkarsh <_Utkarsh!~admin@117.201.81.80> has joined #aichallenge 2010-11-17T20:12:08 *** Utkarsh has quit IRC (Ping timeout: 245 seconds) 2010-11-17T20:15:03 *** _Utkarsh <_Utkarsh!~admin@117.201.81.80> has quit IRC (Ping timeout: 245 seconds) 2010-11-17T20:17:48 Top 10 players: bocsimacko(3632), bix0r4ever(3529), dmj111(3439), protocolocon(3426), shangas(3420), george(3412), _Astek_(3402), ChrisH(3400), luksian(3387), davidjliu(3384) 2010-11-17T20:19:55 *** Utkarsh has joined #aichallenge 2010-11-17T20:23:14 *** davidjliu has joined #aichallenge 2010-11-17T20:24:22 @roulette 2010-11-17T20:24:22 *** eburnette was kicked by contestbot (BANG!) 2010-11-17T20:24:23 * contestbot reloads and spins the chambers. 2010-11-17T20:25:19 *** eburnette_ has joined #aichallenge 2010-11-17T20:25:37 One shot, that's unlucky 2010-11-17T20:25:56 @click 2010-11-17T20:25:56 *** eburnette_ has quit IRC (Client Quit) 2010-11-17T20:25:56 antimatroid: No! 2010-11-17T20:26:39 *** eburnette_ has joined #aichallenge 2010-11-17T20:29:20 *** Xanthor[aw] has quit IRC (Quit: :q :q! :wq :w :w! :wq! :quit :quit! :help help helpquit quit quithelp :quitplease :quitnow :leave :shit ^X^C ^C ^D ^Z ^Q QUITDAMMIT ^[:wq) 2010-11-17T20:34:04 that was the best quit message ever 2010-11-17T20:34:05 *** Frontier has quit IRC (Ping timeout: 245 seconds) 2010-11-17T20:34:32 my bot should be popular at parties, it's learned the new dance "the ship shuffle" 2010-11-17T20:35:55 *** eburnette_ has quit IRC (Ping timeout: 265 seconds) 2010-11-17T20:36:45 *** eburnette has joined #aichallenge 2010-11-17T20:37:06 anybody home? 2010-11-17T20:38:15 I am 2010-11-17T20:38:16 I am but I'm going to the other room to eat something 2010-11-17T20:38:41 nope 2010-11-17T20:38:47 :p 2010-11-17T20:40:50 *** bhasker_ has quit IRC (Quit: bhasker_) 2010-11-17T20:43:31 Has anybody tried neural nets or pattern matching for an evaluation function? 2010-11-17T20:46:41 *** superflit has quit IRC (Quit: superflit) 2010-11-17T20:48:47 This was a hard fought game, check it out: http://98.247.248.39:8080/canvas?game_id=5870 2010-11-17T20:49:04 *** McLeopold has joined #aichallenge 2010-11-17T20:49:58 nup sorry, my evaluation function is still based around the change for a given move 2010-11-17T20:50:48 so instead of evaluating a single position you evaluate the difference between two positions? 2010-11-17T20:50:48 *** sigh has quit IRC (Read error: Connection reset by peer) 2010-11-17T20:51:06 *** sigh has joined #aichallenge 2010-11-17T20:55:18 no, i work out my net gain in ships from the move to the planet, so say i take a neutral, all growht is a gain, if i take the enemy, i double all growth i get for the payoff 2010-11-17T20:55:39 but if it was already going to be owned by me at some times, i don't add that to the payoff for the move 2010-11-17T20:55:58 that way i'm picking the moves that give the biggest change in net ships 2010-11-17T20:56:15 it's fairly messy to implement it like that though 2010-11-17T20:59:43 although, when i evaluate the bottom nodes of my tree, i just do that as a simple check of who would win if no more moves were made 2010-11-17T21:00:02 and by how much 2010-11-17T21:00:21 if a win/loss is certain, those get the extremal payoffs obviously 2010-11-17T21:01:22 amstan: did you get the emails re: adding people to the group? 2010-11-17T21:02:12 pgpaskar: no? 2010-11-17T21:02:34 You should be getting them, there are 2 applicants 2010-11-17T21:02:44 oh, saw that 2010-11-17T21:02:45 visit your control panel & go to usergroups -> manage and tell me if you are the group leader 2010-11-17T21:02:48 Ok. 2010-11-17T21:03:04 Just checking if it was working as intended 2010-11-17T21:03:14 I've tried trees of various heights from 1 to 99 and it seems the shorter the better. Maybe because enemy model is flawed. 2010-11-17T21:03:20 *** Utkarsh has quit IRC (Ping timeout: 255 seconds) 2010-11-17T21:03:21 i saw the conversation about using "moderator" for it 2010-11-17T21:03:35 it does seem like quite a few people are going to want in on it, would "contributers" perhaps fit better? 2010-11-17T21:04:01 how do you get a tree of height 99? 2010-11-17T21:04:03 Name is not set in stone, It's the only thing I could think up within 4 seconds. It can be changed to something less verbose 2010-11-17T21:04:10 i would have thought even a binary tree of that size would pretty much be out 2010-11-17T21:08:40 anti: you can easily make a huristic of board position to early-out on most branches 2010-11-17T21:08:51 If I I get everything working, I *might* be able to do a tree of depth 2 2010-11-17T21:08:58 *** Utkarsh has joined #aichallenge 2010-11-17T21:09:58 so if at each stage, you investigate say.. the 50 best paths, and 50 random paths, and investigate the 3best branches from each of those: and then prune back to 50 again.. you'll only have to investigate about 100k branches 2010-11-17T21:12:46 amstan: you guys need to work out a funding shceme 2010-11-17T21:12:57 scheme* 2010-11-17T21:13:04 *** Utkarsh has quit IRC (Ping timeout: 240 seconds) 2010-11-17T21:13:10 *** McLeopold has quit IRC (Ping timeout: 240 seconds) 2010-11-17T21:13:16 antimatroid: i have no idea how that's going or if we even need it 2010-11-17T21:13:28 not now, but for the future 2010-11-17T21:13:55 the ideas of leaving old contests up for new submissions, users submitting their own entire contests etc. 2010-11-17T21:14:00 there needs to be heaps more servers 2010-11-17T21:14:29 the contest could possibly be big enough for someone to give you access to servers for sponsorship now, donations could work, i'd certainly chip some money in 2010-11-17T21:15:08 and, well, some prize money would be nice, even though I'm happy competing anyway :P 2010-11-17T21:17:13 @tcp 2010-11-17T21:17:14 lavalamp_: tcp = (#1) http://www.benzedrine.cx/planetwars/, or (#2) http://72.44.46.68/, or (#3) http://zeroviz.us:8080, or (#4) 98.247.248.39:995 (zeroviz.us) 2010-11-17T21:17:48 Top 10 players: bocsimacko(3640), SUA.khb(3583), bix0r4ever(3523), dmj111(3437), protocolocon(3428), shangas(3420), george(3409), davidjliu(3404), ChrisH(3393), Xrillo(3378) 2010-11-17T21:19:34 some branches had only one possible move 2010-11-17T21:25:10 are you adding "do nothing" moves? 2010-11-17T21:25:23 i find it's kind of hard to avoid it with minimax 2010-11-17T21:26:20 if i just had a decion tree I can add that into the evaluation of the tree, but i need to keep exploring when modelling both players 2010-11-17T21:27:18 *** bhasker has joined #aichallenge 2010-11-17T21:28:14 *** newdive has joined #aichallenge 2010-11-17T21:29:45 *** nullkuhl has quit IRC (Read error: Connection reset by peer) 2010-11-17T21:31:07 *** Utkarsh has joined #aichallenge 2010-11-17T21:41:10 yeah I have lots of do nothing moves. not sure whether that's good or bad. 2010-11-17T21:41:19 this map was made for sniping practice: http://zeroviz.us:8080/zeroviz_538_19.txt.html 2010-11-17T21:41:22 I need a new pokemon name, anybody? 2010-11-17T21:41:45 all I know is pikachu, never watched it 2010-11-17T21:42:50 snorlax 2010-11-17T21:43:37 *** Utkarsh has quit IRC (Ping timeout: 272 seconds) 2010-11-17T21:47:48 *** Utkarsh has joined #aichallenge 2010-11-17T21:53:39 *** mega1 has quit IRC (Read error: Operation timed out) 2010-11-17T21:57:02 *** Utkarsh has quit IRC (Ping timeout: 260 seconds) 2010-11-17T21:58:35 *** McLeopold has joined #aichallenge 2010-11-17T22:03:04 metapod 2010-11-17T22:03:08 the greatest pokemon 2010-11-17T22:09:17 eburnette: http://www.funnyjunk.com/funny_pictures/139073/all+151+pokemon/ 2010-11-17T22:09:48 those are the only pokemon 2010-11-17T22:10:42 if you say so :) 2010-11-17T22:11:02 *** bhasker has quit IRC (Quit: bhasker) 2010-11-17T22:13:07 *** Utkarsh has joined #aichallenge 2010-11-17T22:14:18 *** Naktibalda has quit IRC (Remote host closed the connection) 2010-11-17T22:17:15 *** mega1 has joined #aichallenge 2010-11-17T22:17:48 Top 10 players: bocsimacko(3644), bix0r4ever(3515), SUA.khb(3463), dmj111(3441), protocolocon(3428), shangas(3420), _Astek_(3415), george(3413), davidjliu(3386), Xrillo(3377) 2010-11-17T22:18:11 *** Utkarsh has quit IRC (Ping timeout: 276 seconds) 2010-11-17T22:33:32 *** Utkarsh has joined #aichallenge 2010-11-17T22:35:48 *** newdive has quit IRC (Ping timeout: 265 seconds) 2010-11-17T22:47:46 *** Utkarsh has quit IRC (Read error: Connection reset by peer) 2010-11-17T22:51:05 @ranking 2010-11-17T22:51:06 eburnette: Top 10 players: bocsimacko(3640), bix0r4ever(3517), dmj111(3430), protocolocon(3430), _Astek_(3429), shangas(3418), george(3407), davidjliu(3390), Xrillo(3371), luksian(3370) 2010-11-17T22:52:15 antimatroid: you there? 2010-11-17T22:54:55 I wonder how many games you have to play to reduce the error rate to near zero. Playing 25 or even 100 games gives pretty random results for bots of similar strength imho. 2010-11-17T22:55:22 you can look at the +/- numbers 2010-11-17T22:55:39 *** Utkarsh has joined #aichallenge 2010-11-17T22:57:09 What do +, -, score, and oppo mean? 2010-11-17T22:58:12 it's random, because the maps(or performance) are random 2010-11-17T22:58:21 i am 2010-11-17T22:59:13 antimatroid: I'm trying to do a map analysis to produce a nice even spread of map types. 2010-11-17T22:59:54 eburnette + and - are the confidence interval for the elo score 2010-11-17T23:00:10 so, your elo is the number give or take those 2 numbers 2010-11-17T23:00:20 *** bhasker has joined #aichallenge 2010-11-17T23:00:52 When someone plays 1 game and shoots to the top of the list, you'll see very high +/- numbers, indicating that there isn't high confidence that is the proper elo 2010-11-17T23:01:10 score is percentage of wins 2010-11-17T23:01:16 cool 2010-11-17T23:01:30 oppo might be the average elo of opponents? not sure 2010-11-17T23:01:42 antimatroid: so I've got a math question for you 2010-11-17T23:01:47 *** phxtrack has joined #aichallenge 2010-11-17T23:02:00 if you have random numbers in the range 1-10... 2010-11-17T23:02:18 Ok, but I have a bot with elo72 +/- 34/35, and I play it a few games and watch the elo range from approx 200 to -100 2010-11-17T23:02:23 the set of (2,5,8) would seem more evenly spread than the set (8,9,10) 2010-11-17T23:02:55 they're both equally evenly spread, and equally likely 2010-11-17T23:03:11 Houndour.8a? 2010-11-17T23:03:19 *** Utkarsh has quit IRC (Read error: No route to host) 2010-11-17T23:03:30 burny: I would argue they are not 2010-11-17T23:03:40 then you should play the lottery 2010-11-17T23:03:46 McLeopold: I don't have much idea with that kind of maths 2010-11-17T23:03:52 since you know more about random numbers than I do 2010-11-17T23:03:55 throw me some algebra questions :) 2010-11-17T23:04:15 I'm not saying more probabal 2010-11-17T23:04:17 McLeopold: you haven't asked a question 2010-11-17T23:04:21 Houndour* and Infernape* have been doing that yes 2010-11-17T23:04:25 :) 2010-11-17T23:04:33 if they arne't equaly likely.. then one must be more probably 2010-11-17T23:04:45 * lavalamp_ facepalm 2010-11-17T23:04:47 ugh what a bug 2010-11-17T23:04:48 what's the best way to return a value that indicates eveness? 2010-11-17T23:05:06 Chi squared? 2010-11-17T23:05:17 probably to just shape it, with standard deviation constraints 2010-11-17T23:05:32 stdev doesn't indicate evenness 2010-11-17T23:05:50 eburnette: thx, I'll try that 2010-11-17T23:06:17 once I understand it 2010-11-17T23:06:36 chi square is only a measure of randomness isn't it? 2010-11-17T23:06:43 eburnette: your +/- is 147/138 2010-11-17T23:07:30 I was looking at a page I hadn't refreshed in a few min 2010-11-17T23:07:31 plus, I think map types play so much of a role with so few games, that it is going to be diffucult to pin down an elo 2010-11-17T23:08:00 McL when you say "evenness" what do you mean? 2010-11-17T23:08:07 really though.. if you want to make maps conform to some sort of pattern.. 2010-11-17T23:08:17 well, for planet count, which is 15-30 2010-11-17T23:08:24 things like: the average, and std-dev of planet distances from each other 2010-11-17T23:08:32 McLeopold: what do you want to do with it? 2010-11-17T23:08:33 if we only had 16 maps, I would want 1 map with each number 2010-11-17T23:08:36 and surface area of largest non-planeted region, etc. 2010-11-17T23:09:04 McLeopold look at Infernape.8.1, it says +/- 34/35 but I guarantee you it will vary a lot more than that if I start it up 2010-11-17T23:09:12 burny: those are great ideas for metrics 2010-11-17T23:09:23 the 15-30 are suposed to be equally likely? 2010-11-17T23:09:41 bah, we needed to think about maps way earlier 2010-11-17T23:09:51 eburnette: yeah, like I said, low game count and map types really mess with the math 2010-11-17T23:10:05 eburnette: if this was tic-tac-toe, it would be a little more stable 2010-11-17T23:10:23 if this was tic tac toe everyone would draw :P 2010-11-17T23:10:37 I'm not going to push any maps stuff for the official contest. This is just for my fun server. 2010-11-17T23:10:56 Except RageTicTacToe 2010-11-17T23:11:00 nxn tic tac toe ai might be interesting? 2010-11-17T23:11:12 I played 350 games on 100 maps, if it's not settling for that then how many does it take? 2010-11-17T23:11:49 actually, that'd be pretty cool i think 2010-11-17T23:11:55 eburnette: idk 2010-11-17T23:11:57 sry 2010-11-17T23:11:58 from the way planets are generated.. about 20% of planets are useless 2010-11-17T23:12:11 burny: yeah, often more too 2010-11-17T23:12:19 due to map size vs density vs regrowth 2010-11-17T23:12:19 antimatroid, that game exists, it's called gomoku or five in a row 2010-11-17T23:12:33 would expect more, half the time 2010-11-17T23:12:36 isn't 5 in a row no more general than tic tac toe? 2010-11-17T23:12:39 just a larger board? 2010-11-17T23:12:58 *** Utkarsh has joined #aichallenge 2010-11-17T23:12:59 no, 5 ina row can be defended indefinately 2010-11-17T23:13:15 I thought 5 in a row was a first person win 2010-11-17T23:13:22 it is 2010-11-17T23:13:30 oh.. 2010-11-17T23:13:33 http://www.cacert.at/cgi-bin/rngresults 2010-11-17T23:13:45 I play it that first person goes once, and then they take turns going twice 2010-11-17T23:13:54 *** rebelxt_ has joined #aichallenge 2010-11-17T23:13:55 that's the pie rule 2010-11-17T23:14:01 "you cut I choose" 2010-11-17T23:14:09 oh, yeah no, i mean take an nxn board, you need a line of length n to win 2010-11-17T23:14:15 oh wait I misread 2010-11-17T23:14:38 antimatroid: that would become very hard as the board grows, no? 2010-11-17T23:15:02 if it's nxn it's impossible to win above size2(unless the other perosn screws up), and 2 or less is imposible to lose 2010-11-17T23:15:38 * eburnette went with Pichu->Pikachu->Raichu 2010-11-17T23:16:01 burny: but is it easy enough to not screw up? 2010-11-17T23:16:05 bhasker: can you start 0.23? 2010-11-17T23:16:19 yes 2010-11-17T23:16:23 above size 3 is easier 2010-11-17T23:16:25 i guess that introduces problems with first movers 2010-11-17T23:16:34 simultaneous games are a lot more fun 2010-11-17T23:17:04 *** Utkarsh has quit IRC (Ping timeout: 240 seconds) 2010-11-17T23:17:11 a game where there was no turn limit, players could just execute moves as they wanted would be interesting 2010-11-17T23:17:46 what if they both want the same spot? 2010-11-17T23:17:48 Top 10 players: bocsimacko(3642), bix0r4ever(3519), dmj111(3430), protocolocon(3430), shangas(3416), _Astek_(3408), george(3404), asavis(3383), luksian(3377), Xrillo(3374) 2010-11-17T23:17:51 s/if/when/ 2010-11-17T23:17:58 * eburnette laughs at 303 +/- 548/365 2010-11-17T23:18:24 like in tron? a giant crash 2010-11-17T23:18:46 Multi-threading should be allowed in future game 2010-11-17T23:19:05 multithreading is way harder for system admin 2010-11-17T23:19:24 maybe 2 threads 2010-11-17T23:19:29 and is barely worth while(especially if you only get around 1second to make descision 2010-11-17T23:20:00 they should just setup a bunch of ps3's with yellow dog and we each get 4 processors 2010-11-17T23:20:27 I think it would add another level to the challenge, and besides everybody needs to learn it 2010-11-17T23:20:53 maybe you could register for the level you want to participate at? 2010-11-17T23:21:04 it would add the most challenge: to the people who are setting up the server 2010-11-17T23:21:10 single thread is one competition, multi-thread another 2010-11-17T23:21:46 do you not understand what I'm saying? 2010-11-17T23:21:46 multiple contests would be cool 2010-11-17T23:21:48 how about - games have to be played on a certain sized ec2 machine known in advance 2010-11-17T23:21:55 like one with this where planet count could be in the 100's 2010-11-17T23:22:08 burny: let us dream 2010-11-17T23:22:23 *** Utkarsh has joined #aichallenge 2010-11-17T23:22:40 although having 1 contest does make it pretty cool 2010-11-17T23:23:15 with 2cores, it's really not worth bothering to use them both 2010-11-17T23:23:15 (after 2 games: 375 +/- 485/306) 2010-11-17T23:23:40 not to mention, this game can be played perfect with nearly no processing 2010-11-17T23:24:07 what?!? 2010-11-17T23:24:09 eburnette: just for you, I'm going to modify the ranking calcs... 2010-11-17T23:24:46 imho, if the top bots are hand-coded heuristics then it's not real ai 2010-11-17T23:25:01 there is no real ai 2010-11-17T23:25:20 but there is some truth to that 2010-11-17T23:25:35 real ai is against the rles 2010-11-17T23:25:48 it would be cooler if you just plugged the rules in and the bot told you the best way to play 2010-11-17T23:26:26 there is general game playing competition 2010-11-17T23:26:38 *** virdo has joined #aichallenge 2010-11-17T23:26:40 but it's not as general as you'd hope 2010-11-17T23:26:58 if this was a real ai contest, instead of just hand-made huristics for known map generation.. 2010-11-17T23:27:07 like a minimax framework where you plugin the game rules? 2010-11-17T23:27:12 I'd make it work slightly like that mcleo 2010-11-17T23:27:18 No new fatal bugs found in the map generator? 2010-11-17T23:27:24 what I call a "real ai" system would not rely on heuristics except to optimize (prune trees, etc.), am I wrong? 2010-11-17T23:27:32 mega1: are you going to upload .66 to the main server? 2010-11-17T23:27:33 janzert: just that I suck now 2010-11-17T23:27:43 I agree eburnette 2010-11-17T23:27:43 hehe 2010-11-17T23:28:02 there is no self thinking ai 2010-11-17T23:28:10 i agree with that 2010-11-17T23:28:11 davidjliu: sooner or later, yes. I have a version in local testing that's better. 2010-11-17T23:28:19 it's all about implementing your own strategies 2010-11-17T23:28:27 the primary thing that this contest prevents 'real' AI from.. is you aren't allowed to write to files, or have any persistent data between games 2010-11-17T23:28:29 * eburnette gives mega1 the rest of the month off 2010-11-17T23:28:30 scary 2010-11-17T23:28:47 but that's what I like about these contests, you are competing on strategies (and ability to code them up), not on your ability to "play a game" 2010-11-17T23:28:50 *** Utkarsh has quit IRC (Ping timeout: 240 seconds) 2010-11-17T23:29:01 well, 0.66 is much stronger than it was before 2010-11-17T23:29:11 yeah, i don't quite get why we can't write to files 2010-11-17T23:29:17 mega1: b0.67?! ;-)) 2010-11-17T23:29:19 i guess cause storage would be insane 2010-11-17T23:29:27 mega1 what did you change to make it stronger? 2010-11-17T23:29:39 because storage makes cheating easier 2010-11-17T23:29:44 antimatroid: it would also affect rankings 2010-11-17T23:29:49 because order of games metters 2010-11-17T23:29:53 eburnette: optimized for the current maps 2010-11-17T23:29:58 *matters 2010-11-17T23:30:02 *** McLeopold_ has joined #aichallenge 2010-11-17T23:30:18 it was too conservative 2010-11-17T23:30:30 * antimatroid takes note 2010-11-17T23:30:36 so it was beaten by almost any bot on certain maps 2010-11-17T23:30:46 * eburnette wonders what mega1 would consider an agressive bot 2010-11-17T23:31:26 a suicidal one 2010-11-17T23:31:53 It's just that your bot is already pretty darned agressive 2010-11-17T23:32:12 *** McLeopold_ has quit IRC (Client Quit) 2010-11-17T23:32:29 *** McLeopold_ has joined #aichallenge 2010-11-17T23:32:40 dang malazanbot is going to ruin my elo 2010-11-17T23:32:41 is it? I thought of it as pretty laid back. 2010-11-17T23:33:08 *** McLeopold has quit IRC (Ping timeout: 265 seconds) 2010-11-17T23:33:38 *** Utkarsh has joined #aichallenge 2010-11-17T23:33:43 bocsi has a recognizable 'down your throat' kind of style that not many others have 2010-11-17T23:34:09 an sufficiently aggressive bot with no flaws, should win against most 10-100 ranked bots in around 50 turns 2010-11-17T23:34:29 let me check 2010-11-17T23:35:32 it's about mostly <= 90 2010-11-17T23:35:51 but there are several maps that are split 2010-11-17T23:36:11 and on those maps it's not possible to finish to game early against a good opponent. 2010-11-17T23:36:51 example: http://zeroviz.us:8080/canvas?game_id=2510 2010-11-17T23:37:13 the wider the chasm the longer the game 2010-11-17T23:37:53 actually, I don't really want to give bix04 too much testing time 2010-11-17T23:38:25 although it may be on tcp too under a different name for aIl I know 2010-11-17T23:38:44 well: assuming it's not a trivial draw map :P 2010-11-17T23:38:47 *** Utkarsh has quit IRC (Ping timeout: 276 seconds) 2010-11-17T23:39:35 maps I'm referrig to are not draw maps but two player solitaire game 2010-11-17T23:39:53 those are draws 2010-11-17T23:39:53 two player solitaire game? 2010-11-17T23:40:09 is http://98.247.248.39:8080/canvas?game_id=2602 a counterexample? 2010-11-17T23:40:19 it's all about optimizing your growth rate, without letting population drop too much below opponent 2010-11-17T23:40:33 but f the best way to gain growth, is obvious.. 2010-11-17T23:40:38 then they will both be equal 2010-11-17T23:40:55 antimatroid: there is no contact whatsoever 2010-11-17T23:41:08 oh, decision problems :P 2010-11-17T23:41:30 or http://98.247.248.39:8080/canvas?game_id=989 2010-11-17T23:41:55 if there's a center planet with 5growth and around 6-20cost.. then that planet becomes so critical, that no one can ever take any other planets mostly 2010-11-17T23:42:05 eburnette: i think if both expanded fully, that 100 never would have been takable 2010-11-17T23:42:19 fully == most efficiently/whatever 2010-11-17T23:42:56 so in a way, it was a solitaire game, but whoever won that won before the end in a way 2010-11-17T23:43:18 in a way in a way in a way :) 2010-11-17T23:43:59 hmmm maybe not actually, you also had to not leave the 100 open for them to take it 2010-11-17T23:44:18 if ht e100 is open to take, then so is your base 2010-11-17T23:44:27 not necessarily 2010-11-17T23:44:45 *** Utkarsh has joined #aichallenge 2010-11-17T23:45:47 *** iNeo has joined #aichallenge 2010-11-17T23:47:36 hm, that diff did the opposite of what i wanted it to 2010-11-17T23:47:37 oh well 2010-11-17T23:49:37 *** Utkarsh has quit IRC (Ping timeout: 260 seconds) 2010-11-17T23:50:13 *** iNeo has quit IRC (Ping timeout: 265 seconds) 2010-11-17T23:53:58 *** medrimonia1 has joined #aichallenge 2010-11-17T23:54:11 *** rebelxt_ has quit IRC (Quit: Page closed) 2010-11-17T23:55:22 hahaha janzert you have 108 commits in a row 2010-11-17T23:55:30 and i just broke the streak >.> 2010-11-17T23:55:40 yay :) 2010-11-17T23:55:43 does anybody have a big database of games saved they'd be willing to share? 2010-11-17T23:55:44 109 * 2010-11-17T23:55:50 *** medrimonia has quit IRC (Ping timeout: 255 seconds) 2010-11-17T23:56:03 *** Utkarsh has joined #aichallenge 2010-11-17T23:56:13 janzert: also, i have a php code review for you 2010-11-17T23:56:38 sure, in just a bit. working on updating maps 2010-11-17T23:56:45 no problem, no rush on my end 2010-11-17T23:58:38 *** amriedle has quit IRC (Quit: Lost terminal)