2010-10-04T00:00:41 *** O1athe has joined #aichallenge 2010-10-04T00:00:43 *** Olathe has quit IRC (Ping timeout: 276 seconds) 2010-10-04T00:06:06 *** davidjliu has quit IRC (Quit: Page closed) 2010-10-04T00:06:48 *** delt0r_ has joined #aichallenge 2010-10-04T00:08:08 *** delt0r___ has quit IRC (Ping timeout: 240 seconds) 2010-10-04T00:08:34 *** justin_pdx has quit IRC (Quit: justin_pdx) 2010-10-04T00:11:46 *** sigh has quit IRC (Read error: No route to host) 2010-10-04T00:12:30 evening 2010-10-04T00:13:14 *** sigh has joined #aichallenge 2010-10-04T00:13:56 hey bhasker 2010-10-04T00:14:04 hey 2010-10-04T00:14:08 how is your bot coming along 2010-10-04T00:15:28 my bot gets worse with every iteration i make 2010-10-04T00:15:33 at least i got my logging cleaned up 2010-10-04T00:15:44 bhasker: you mean everyone gets better? 2010-10-04T00:15:58 no my bot gets worse, it loses to itself 2010-10-04T00:16:05 rather newer versions lose to my older ones 2010-10-04T00:16:06 :-\ 2010-10-04T00:16:13 everyone else is also getting better 2010-10-04T00:16:19 bhasker: idk.. that's not really a good measure 2010-10-04T00:16:28 thats a problem to face if i can consistently beat my own older version 2010-10-04T00:19:00 ah perfect balance http://72.44.46.68/canvas?game_id=297068 2010-10-04T00:19:09 I am also finding many new and innovating ways to make my bot worse 2010-10-04T00:24:03 i haven't made a working one in a while, though i'm working on a couple ideas 2010-10-04T00:24:47 and also playing minecraft 2010-10-04T00:25:30 everyone is on minecraft o_O 2010-10-04T00:25:40 but mostly playing with my kid, who won't stop shouting "dada build a tunnel" "dada build something" "dada build a track" 2010-10-04T00:25:49 :) 2010-10-04T00:26:23 we had to get a few more wooden train set parts today so as to maximize the potential of the track we have 2010-10-04T00:27:00 a1k0n: i can has your map? 2010-10-04T00:27:08 minecraft map? 2010-10-04T00:28:02 a1k0n: yeah 2010-10-04T00:28:11 where does that even get stored on the mac version? 2010-10-04T00:28:20 i wish multiplayer worked better 2010-10-04T00:28:24 idk, app data? 2010-10-04T00:28:32 wait, that's windows 2010-10-04T00:28:33 no idea 2010-10-04T00:28:35 ahha. application support. 2010-10-04T00:28:35 lol 2010-10-04T00:28:42 it's .minecraft in linux 2010-10-04T00:28:56 a1k0n: you may also present a map of your map 2010-10-04T00:29:15 wuh? 2010-10-04T00:29:22 a1k0n: hold on, incoming 2010-10-04T00:29:35 dcc is unlikely to work if you're trying to do that 2010-10-04T00:29:42 no 2010-10-04T00:30:25 it's at http://a1k0n.net/temp/minecraft-world.tar.gz 2010-10-04T00:30:30 warning: unzips to World2/ 2010-10-04T00:30:38 np 2010-10-04T00:30:45 and here goes half my night 2010-10-04T00:30:47 lol 2010-10-04T00:30:51 s/here/there 2010-10-04T00:31:07 i left off inside a deep cave.. follow the torches on the left to get out (there's a complex bit with some lava which might be hard to figure out) 2010-10-04T00:31:15 or just... kill yourself 2010-10-04T00:31:28 i have a giant tower on a hill with a zen garden on top, with some trees 2010-10-04T00:31:31 and some grass 2010-10-04T00:31:34 neat.. 2010-10-04T00:31:35 er, wheat 2010-10-04T00:31:48 a1k0n: http://csclub.uwaterloo.ca/~amstan/upload/minecraft/map/ 2010-10-04T00:32:03 blue is caves 2010-10-04T00:32:20 oh shit that's awesome 2010-10-04T00:32:29 can you generate that for mine? 2010-10-04T00:32:31 what program is that? 2010-10-04T00:32:55 i would like to locate some dungeons 2010-10-04T00:34:14 *** JCS^ has quit IRC (Quit: \x00) 2010-10-04T00:35:56 a1k0n: http://csclub.uwaterloo.ca/~amstan/upload/minecraft/map2/ 2010-10-04T00:36:13 i have a makefile for them.. lol 2010-10-04T00:36:16 rad 2010-10-04T00:36:17 go /makefile 2010-10-04T00:36:25 is that my map then? 2010-10-04T00:36:28 y[ep 2010-10-04T00:36:45 you seem to be pretty cave rich 2010-10-04T00:36:46 i just found another cavern above the one i was in 2010-10-04T00:36:51 though.. there's a cave system south of you 2010-10-04T00:36:58 yeah. i wish i knew where i was on that map 2010-10-04T00:37:24 a1k0n: if you have firebug, you could change the opacity of the caves layer 2010-10-04T00:37:42 oh, i don't generally play within a browser 2010-10-04T00:37:51 oh er 2010-10-04T00:38:02 http://csclub.uwaterloo.ca/~amstan/upload/minecraft/map2/map.png 2010-10-04T00:38:08 i see, heh 2010-10-04T00:38:09 the light spots are torches 2010-10-04T00:38:12 this is a night map 2010-10-04T00:38:22 red is flowers.. 2010-10-04T00:39:28 a1k0n: wow.. you have slimes 2010-10-04T00:39:34 so my citadel is pretty easy to see but i can't really see my mines.. they're pretty deep 2010-10-04T00:39:38 yep 2010-10-04T00:39:47 a1k0n: down from your lava.. there's something dropping from the ceiling 2010-10-04T00:39:52 yep 2010-10-04T00:40:00 there's another cavern above there 2010-10-04T00:40:04 by the coal, yes? 2010-10-04T00:40:19 i just had a skeleton walk by after clearing out the coal 2010-10-04T00:40:36 don't make me play minecraft too :| 2010-10-04T00:40:39 *** justin_pdx has joined #aichallenge 2010-10-04T00:40:40 I've avoided it this long 2010-10-04T00:43:55 *** jimmerton_ has joined #aichallenge 2010-10-04T00:44:56 *** justin_pdx has quit IRC (Client Quit) 2010-10-04T00:45:54 *** jimmerton has quit IRC (Ping timeout: 240 seconds) 2010-10-04T00:45:55 *** jimmerton_ is now known as jimmerton 2010-10-04T00:45:58 hmm i should check out this minecraft 2010-10-04T00:47:03 I haven't played minecraft, but it seems like promoting it might be a good strategy 2010-10-04T00:47:21 *** jesionaj` has quit IRC (Read error: Connection reset by peer) 2010-10-04T00:48:45 *** jesionaj has joined #aichallenge 2010-10-04T00:54:36 *** Zeiris_ has joined #aichallenge 2010-10-04T00:58:07 *** Zeiris has quit IRC (Ping timeout: 255 seconds) 2010-10-04T00:58:23 *** medice has quit IRC (Ping timeout: 264 seconds) 2010-10-04T00:58:54 *** medice has joined #aichallenge 2010-10-04T00:59:03 Top 10 players: Hazard(3766), davidjliu(3760), dmj111(3718), cfaftw(3692), rsergio(3665), felixcoto(3638), Polsky(3636), EBraun(3620), sequoh(3613), youwillfail(3597) 2010-10-04T01:00:04 contestbot: whoami 2010-10-04T01:00:04 sigh: I don't recognize you. 2010-10-04T01:00:07 :( 2010-10-04T01:00:43 this yourself bot keeps wiping the floor with my bot:(( 2010-10-04T01:01:56 * sigh resists the temptation to make the obvious "stop playing with yourself" joke 2010-10-04T01:02:32 heh 2010-10-04T01:03:42 i think i am going to stop tinkering unless i can come up with a radical new approach 2010-10-04T01:04:33 the problem with radical approaches is they have high risk and high investment 2010-10-04T01:05:48 heh but my current version doesn't seem to be getting much better 2010-10-04T01:06:07 yup, I know the feeling 2010-10-04T01:06:38 *** amstan has quit IRC (Ping timeout: 245 seconds) 2010-10-04T01:07:21 *** Arthur_ has quit IRC (Read error: Operation timed out) 2010-10-04T01:10:37 *** Arthur has joined #aichallenge 2010-10-04T01:10:55 *** Palmik has joined #aichallenge 2010-10-04T01:12:03 hmm my new version might just be a little better 2010-10-04T01:12:10 http://72.44.46.68/canvas?game_id=297606 2010-10-04T01:14:39 no stalemate this time :) 2010-10-04T01:15:35 heh, more importantly i got the coordinated attack stuff working a bit 2010-10-04T01:15:49 though i am not sure coordinating attack against enemy planets helps much 2010-10-04T01:15:53 cool :) 2010-10-04T01:16:05 it should 2010-10-04T01:16:07 when i try that,ships just pile up while enemy is expanding 2010-10-04T01:16:24 ah, well you need to make sure that you expand yourself too 2010-10-04T01:16:56 if you only attack, then you are just a ragebot variant :P 2010-10-04T01:17:02 :p 2010-10-04T01:17:12 let me try putting the coordinate code back in and see 2010-10-04T01:18:32 oh, I thought that game was with coordinated attacks 2010-10-04T01:19:48 I call it a coordinated attack if multiple sources attack work together to attack a single target 2010-10-04T01:19:59 yea mine does it for neutrals 2010-10-04T01:20:24 for enemy controlled planets i haven't yet got that working since 2010-10-04T01:20:33 i do some prediction on staging planets 2010-10-04T01:20:43 so i tend to send ships ahead of time too 2010-10-04T01:21:37 the general method of attacking in that game doesn't seem to bad 2010-10-04T01:21:47 *too 2010-10-04T01:23:12 omg lol my bot lost to snakebot:( 2010-10-04T01:23:26 http://72.44.46.68/getplayer?player=McLeopold.WormBot.py 2010-10-04T01:23:27 heh 2010-10-04T01:23:39 oh, wormbot is much stronger than snakebot 2010-10-04T01:23:47 *** justin_pdx has joined #aichallenge 2010-10-04T01:23:55 oh yea 2010-10-04T01:23:57 looks like it 2010-10-04T01:24:00 http://72.44.46.68/canvas?game_id=297736 2010-10-04T01:24:01 owned me 2010-10-04T01:24:35 yeah... you are a bit too eager to attack 2010-10-04T01:25:04 *** justin_pdx has quit IRC (Client Quit) 2010-10-04T01:25:07 you could use those ships to expand instead 2010-10-04T01:25:07 hmm 2010-10-04T01:25:21 *** sinan has joined #aichallenge 2010-10-04T01:25:59 I think your planet evaluation function does not respect distance enough 2010-10-04T01:26:44 the longer your ships are in flight, the less flexible you can be 2010-10-04T01:27:06 hmm i am not sure why my bot decided to attack him 2010-10-04T01:27:09 going through logs 2010-10-04T01:30:44 gah, i have more work to do in the next two weeks than a kid in a slave shop 2010-10-04T01:30:52 plus i have to goto sydney this weekend for a family wedding :| 2010-10-04T01:31:28 gah it was a stupid heuristic that screwed me over, it decided taking the neutrals was too expensive, so since it had ships left over 2010-10-04T01:31:32 enemy was the only planets left 2010-10-04T01:31:36 so it sent all of them at the enemy 2010-10-04T01:31:54 bhasker: ah, fixing that up should make it much better 2010-10-04T01:32:12 also not sure why it thought it could acuqire the planet with 50 ships when the planet was halfway across the map 2010-10-04T01:32:31 :) 2010-10-04T01:33:27 oh its a 1 growth planet 2010-10-04T01:33:37 at a distance of 18 2010-10-04T01:33:45 with 6 ships on it 2010-10-04T01:33:46 doh 2010-10-04T01:33:53 haha 2010-10-04T01:38:30 funny game 2010-10-04T01:38:31 http://72.44.46.68/canvas?game_id=297852 2010-10-04T01:38:41 i had a bug but i won that game 2010-10-04T01:38:53 i went for a planet all the way across the map 2010-10-04T01:38:57 but won it 2010-10-04T01:54:48 *** yasith__ has quit IRC (Ping timeout: 252 seconds) 2010-10-04T01:55:19 contestbot: echo hello! 2010-10-04T01:55:19 hello! 2010-10-04T01:55:33 contestbot: echo I am a BANANA 2010-10-04T01:55:33 I am a BANANA 2010-10-04T01:55:50 *** Palmik has quit IRC (Remote host closed the connection) 2010-10-04T01:57:06 !rainbow BANANA 2010-10-04T01:57:06 antimatroid2: BANANA 2010-10-04T01:57:32 !echo $randomNick: You are a BANANA 2010-10-04T01:57:32 MrCeeJ: You are a BANANA 2010-10-04T01:57:35 lo; 2010-10-04T01:57:36 lol 2010-10-04T01:57:50 aha 2010-10-04T01:58:44 !rot13 uryyb 2010-10-04T01:58:44 sigh: hello 2010-10-04T01:58:46 oh dear, i think we should be scared 2010-10-04T01:58:55 top 2 users resubmitted 2010-10-04T01:59:04 Top 10 players: dmj111(3760), davidjliu(3754), cfaftw(3724), rsergio(3674), felixcoto(3652), EBraun(3648), sequoh(3628), Polsky(3615), youwillfail(3607), DocBot(3562) 2010-10-04T01:59:22 haha, yeah 2010-10-04T01:59:27 and dmj is back on top 2010-10-04T01:59:32 did he resubmit? 2010-10-04T01:59:34 only because of that :P 2010-10-04T01:59:35 nope 2010-10-04T01:59:37 just checked 2010-10-04T02:00:08 wow... shows how volatile the top is 2010-10-04T02:00:10 *** Palmik has joined #aichallenge 2010-10-04T02:00:39 yeah, it's been like that this time, but that's probably because no bot is best on a general map 2010-10-04T02:00:48 so it's a bit of luck thrown in as well 2010-10-04T02:01:11 gives me hope... maybe I should upload a slightly different bot and see how it does 2010-10-04T02:01:12 *** amstan has joined #aichallenge 2010-10-04T02:01:12 *** ChanServ sets mode: +o amstan 2010-10-04T02:01:33 hmm, i'm focusing on getting some kind of minimax tree going before uploading anything else 2010-10-04T02:01:55 !rot13 nzfgna vf urer! 2010-10-04T02:01:55 sigh: amstan is here! 2010-10-04T02:02:12 omg! 2010-10-04T02:02:13 yes 2010-10-04T02:02:16 my internet's back 2010-10-04T02:02:22 antimatroid2: I got an idea for a minimax tree, but it will need some work 2010-10-04T02:02:32 a1k0n: sorry about that 2010-10-04T02:02:35 what about your internet's back? 2010-10-04T02:03:04 *my bot will need a lot of work before it is ready for minimax 2010-10-04T02:04:18 *** HakanD has quit IRC (Ping timeout: 240 seconds) 2010-10-04T02:04:27 *** justin_pdx has joined #aichallenge 2010-10-04T02:05:28 !eightball should I try implementing minimax? 2010-10-04T02:05:28 sigh: I doubt it very much. 2010-10-04T02:05:52 well, that settles that 2010-10-04T02:08:23 *** justin_pdx has quit IRC (Client Quit) 2010-10-04T02:10:15 !eightball what is the day tomorrow 2010-10-04T02:10:15 bhasker: It is possible. 2010-10-04T02:10:19 *** medice has left #aichallenge 2010-10-04T02:11:53 *** iFire has quit IRC (Read error: Connection reset by peer) 2010-10-04T02:12:31 haha 2010-10-04T02:12:44 i have started making the files for the game tree 2010-10-04T02:12:53 files? 2010-10-04T02:12:57 but i need to clean planet wars up and add extra functions for more thust 2010-10-04T02:13:03 thust? 2010-10-04T02:13:07 thrust* 2010-10-04T02:13:21 files, as in started setting up gameTree and Node 2010-10-04T02:13:27 ah, I see 2010-10-04T02:13:37 *** iFire has joined #aichallenge 2010-10-04T02:13:37 i don't know whether to use alpha-beta pruning or not 2010-10-04T02:13:47 cause we aren't searching the entire tree 2010-10-04T02:14:01 *** yasith_ has joined #aichallenge 2010-10-04T02:14:43 huh? does alpha-beta stop working in that case? 2010-10-04T02:15:19 well no, but i think it's more likely that you can go down what seems like a bad path with this for a few moves and then it pays off 2010-10-04T02:15:38 given one isn't going to penetrate very far as well, i'm unsure as to whether it might miss that somehow 2010-10-04T02:15:53 but the reason alpha-beta prunes is that because a path is provably worse, no? 2010-10-04T02:16:22 out of the set of moves you are considering 2010-10-04T02:16:27 yeah, alpha-beta should still be better 2010-10-04T02:16:52 I can't see how alpha-beta would change the result in any way 2010-10-04T02:16:54 than pure min-max that is 2010-10-04T02:17:10 I thought it only changes the number of nodes that you lok at 2010-10-04T02:17:17 *look 2010-10-04T02:17:35 sigh: right, as long as you don't get into transposition table issues (which probably don't apply here at all) 2010-10-04T02:17:59 yeah, that's a separate issue :) 2010-10-04T02:18:06 sigh: cause the payoff for a state might look bad, but then you get to make a move 2 or 3 turns later that makes it good? 2010-10-04T02:18:27 but that's an inherent problem with your minimax 2010-10-04T02:18:28 even the payoff function has to be fairly guessy for this 2010-10-04T02:18:36 for tron it was fairly easy to set up a decent payoff function 2010-10-04T02:19:39 that's a problem with minimax, alpha-beta doesn't make it better or worse 2010-10-04T02:20:12 barring bugs, minimax and alpha-beta with a deterministic evaluation function should always return the same result 2010-10-04T02:20:29 if i generate the tree with pruning, then extend those non-dominated nodes, isn't it possible for me to miss a node that became good again? 2010-10-04T02:20:34 when searching to the same depth that is 2010-10-04T02:20:51 antimatroid2: the point of alpha-beta is to prune ONLY when you know that it won't affect the result 2010-10-04T02:21:15 but with the tree extending, that result might change? 2010-10-04T02:21:32 that's a problem with minimax, not alpha-beta 2010-10-04T02:21:52 yes, but if i then extend normal minimax, i see that, if i skip it with alpha-beta, i don't see it? 2010-10-04T02:21:58 hence it's not a problem with minimax? 2010-10-04T02:22:31 what do you mean by extending? 2010-10-04T02:22:33 * Janzert is unsure what you mean by extending the tree 2010-10-04T02:22:39 :) 2010-10-04T02:22:53 you generate to a certain depth, then realsie you have more time, so you keep extending 2010-10-04T02:22:55 are you doing iterative deeping? 2010-10-04T02:23:04 oh you are *storing* the entire tree? o_O 2010-10-04T02:23:18 no, iterative deepening 2010-10-04T02:23:26 if i have time left, i will extend 2010-10-04T02:23:39 then you don't care what alpha-beta pruned last iteration 2010-10-04T02:23:43 this is a new iteration 2010-10-04T02:24:52 when one normally iteratively deepens, do they throw new payoffs back up the tree as they find them? or do a final scan at the end? 2010-10-04T02:25:07 i would normally have tended to go with the former 2010-10-04T02:25:11 *** jesionaj has quit IRC (Ping timeout: 276 seconds) 2010-10-04T02:25:29 in my knowledge yes, usually you keep track of the moves to inform search order but that's about it 2010-10-04T02:25:48 the values from last time are only useful as a heuristic 2010-10-04T02:26:02 normally a true tree structure isn't built at all 2010-10-04T02:26:21 yup ^^ 2010-10-04T02:26:27 last time we could have a tree that just threw out large chunks each turn, but we have to generate from scratch every turn this time 2010-10-04T02:26:36 yes 2010-10-04T02:26:38 so just do them in vectors? 2010-10-04T02:26:46 and not actually generate a tree? i do that for graphs 2010-10-04T02:27:01 vector > myGraph :) 2010-10-04T02:27:11 the point of iterative deepening is that each iteration is exponentially larger, so that the cost of the previous iterations are not significant 2010-10-04T02:27:13 *** Utkarsh has quit IRC (Ping timeout: 265 seconds) 2010-10-04T02:27:26 basically you just do a depth-first search of the tree, you don't store it at all really 2010-10-04T02:27:33 yeah, don't generate the game tree as an actual data structure :) 2010-10-04T02:28:15 but as I said, you can remember the results at the top to inform search order 2010-10-04T02:28:20 the transposition table is an optimization to remember the results of searching a chunk of the tree, but is a simple flat hashmap 2010-10-04T02:28:22 that should help priuning 2010-10-04T02:28:56 Janzert: that only helps within one iteration? 2010-10-04T02:29:22 okay, that would make sense :P 2010-10-04T02:29:25 I can't see how it would carry across iterations, but I have not implemented alpha-beta with transposition tables 2010-10-04T02:29:34 i wont be using anywhere near as much memory this time aha 2010-10-04T02:29:39 haha 2010-10-04T02:29:40 no it's a huge boost for the iterative deepening by remembering the best move for most positions the next iteration runs into 2010-10-04T02:29:52 at least in games like chess 2010-10-04T02:30:08 ah, so just as a heuristic for search order? 2010-10-04T02:30:18 how do you do deepening without storing? i get that you store the states at those end points, do you just store inside that the path of play to get there? 2010-10-04T02:30:39 sigh: basically yes 2010-10-04T02:31:03 antimatroid: you start the search over but with a depth + 1 of the last search 2010-10-04T02:31:25 wouldn't that be very costly time wise? 2010-10-04T02:31:33 *** Utkarsh has joined #aichallenge 2010-10-04T02:31:37 as I said before: the point of iterative deepening is that each iteration is exponentially larger, so that the cost of the previous iterations are not significant 2010-10-04T02:31:51 yeah not significant, but they still exist? 2010-10-04T02:31:56 only constant factor :P 2010-10-04T02:32:01 you still get that many more states searched 2010-10-04T02:32:28 why not just have a vector of states to extend with information about the path that got there? 2010-10-04T02:32:38 memory isn't that constrained 2010-10-04T02:33:27 it really is for most game branching factors and the time limits generally used 2010-10-04T02:33:50 also, you have recalculate the values anyway 2010-10-04T02:33:58 *** keynes has quit IRC (Ping timeout: 240 seconds) 2010-10-04T02:34:25 yeah, there's also a fair bit of work this time to generate moves, run them etc. 2010-10-04T02:34:59 sure, you can cache that sort of stuff 2010-10-04T02:35:10 i'm not expecting the game tree to generate very far, considering you want to consider a large number of moves on each of your frist turns 2010-10-04T02:36:28 if move generation is very expensive, then sure cache that 2010-10-04T02:37:13 i'm even considering having each of my "terminal" nodes playing their own game for a specified number of turns to see how that pans out 2010-10-04T02:37:34 that's an expensive evaluation function 2010-10-04T02:37:44 it's also quite a nice one 2010-10-04T02:37:54 depends 2010-10-04T02:37:56 you could even try to predict some qualities about their bot and have it run against that 2010-10-04T02:38:22 if it was that nice, you would not need minimax :P 2010-10-04T02:38:46 yes, but how are you planning to do payoff otherwise? 2010-10-04T02:38:52 but yeah, I get your point 2010-10-04T02:39:02 yourShips - theirShips + remainingTime(yourGrowth - theirGrowth) is quite a nice idea 2010-10-04T02:39:11 but it doesn't take into account position strength and many other things 2010-10-04T02:39:40 you'd also want that after all current fleets have run their course 2010-10-04T02:40:05 I'm wary about running such simulations because I WILL likely get good results 2010-10-04T02:40:07 *** Cyndre__ has quit IRC (Read error: Connection reset by peer) 2010-10-04T02:40:15 but ultimatly I am optimising against my own bots 2010-10-04T02:40:28 grr, wish i can finish my bot :/ 2010-10-04T02:40:29 so it might suck in the actual contest 2010-10-04T02:40:35 My bot is way beyond that, antimatroid 2010-10-04T02:40:37 yep, hence why one would preferably try to predict characteristics of their bot based off past moves 2010-10-04T02:40:38 :| 2010-10-04T02:40:40 the framework anyway 2010-10-04T02:40:44 *** Cyndre__ has joined #aichallenge 2010-10-04T02:40:52 beyond? 2010-10-04T02:41:11 your ships - their ships, etc are limited to some level 2010-10-04T02:41:24 mine would take into account positions, etc 2010-10-04T02:41:29 or do :/ 2010-10-04T02:41:39 but need to finish my bot first :/ 2010-10-04T02:41:46 already tested the future predictions, at least 2010-10-04T02:42:04 but busy for next week, have to make a game level in udk. :p 2010-10-04T02:42:12 if you look at the graph of most games, it is quite obvious by looking at ship numbers where the game went to one side 2010-10-04T02:42:38 so it's not that bad 2010-10-04T02:43:04 "remainingTime(yourGrowth - theirGrowth)" is important, but I am suspicious 2010-10-04T02:43:34 though predictions is only one part of the game 2010-10-04T02:43:44 yup 2010-10-04T02:43:47 anyway, I'm out 2010-10-04T02:43:47 still need the code to work on it :) 2010-10-04T02:43:49 cya 2010-10-04T02:43:52 *** sigh has quit IRC () 2010-10-04T02:43:54 i'm going to go play gmaes before sleep too 2010-10-04T02:45:59 *** Itkovian has quit IRC (Quit: Itkovian) 2010-10-04T02:47:30 *** ananda_ has joined #aichallenge 2010-10-04T02:47:30 *** otherAntimatroid has joined #aichallenge 2010-10-04T02:48:11 Hi.. I just wanted to make sure that when someone loses by crashing or invalid order, the site doesn't quite tell the right thing, including determining who lost and who won because of that 2010-10-04T02:48:27 Is that true, are you also observing similar things 2010-10-04T02:51:47 *** hellman has joined #aichallenge 2010-10-04T02:53:09 *** hellman has joined #aichallenge 2010-10-04T02:58:49 anada_: people have indeed noted this 2010-10-04T02:59:04 Top 10 players: davidjliu(3787), dmj111(3771), cfaftw(3735), rsergio(3694), felixcoto(3672), EBraun(3661), sequoh(3647), Polsky(3632), youwillfail(3628), DocBot(3576) 2010-10-04T03:00:55 is it same to all? 2010-10-04T03:03:12 @antimatroid2: so what should i assume when I get some crash/invalid order. I occasionally get some, so don't know if it's from my side 2010-10-04T03:03:13 ananda_: I have no idea what you mean. 2010-10-04T03:03:55 @contestbot: When someone loses by crashing or invalid order in the server, the site doesn't quite tell the right thing (who crashed), including determining who lost and who won because of that 2010-10-04T03:03:56 ananda_: You have no gotten any error messages recently, so here's a random one just to let you know that we care. 2010-10-04T03:04:39 ananda_: he's a bot... 2010-10-04T03:04:43 :DD:DD 2010-10-04T03:04:55 wow 2010-10-04T03:05:10 you know.. contest ... bot.. 2010-10-04T03:05:16 lol 2010-10-04T03:05:19 :D 2010-10-04T03:05:36 okay so what abt the erroneous displays 2010-10-04T03:06:03 ananda_: ask j3camero to add you to the mailing list, if you want to contribute in this way 2010-10-04T03:07:35 I'm very new to programming. what do you mean by contribute? 2010-10-04T03:07:48 *** wvdschel has joined #aichallenge 2010-10-04T03:07:57 participate 2010-10-04T03:08:21 rabidus_: participate? - to me? 2010-10-04T03:08:28 y 2010-10-04T03:08:49 im already in the leaderboard.. page 2 2010-10-04T03:11:00 *** Itkovian has joined #aichallenge 2010-10-04T03:12:37 *** justin_pdx has joined #aichallenge 2010-10-04T03:16:55 *** justin_pdx has quit IRC (Ping timeout: 252 seconds) 2010-10-04T03:20:10 *** ananda_ has quit IRC (Quit: Page closed) 2010-10-04T03:21:09 *** jfrank has quit IRC (Ping timeout: 265 seconds) 2010-10-04T03:21:37 *** HakanD has joined #aichallenge 2010-10-04T03:21:53 morning 2010-10-04T03:22:34 *** Palmik has quit IRC (Read error: Operation timed out) 2010-10-04T03:24:11 *** Palmik has joined #aichallenge 2010-10-04T03:24:19 evening :) 2010-10-04T03:24:57 amanda_: thankyou for that bloop above, it was hilarious 2010-10-04T03:25:21 damn, they're gone 2010-10-04T03:30:50 *** Palmik has quit IRC (Ping timeout: 276 seconds) 2010-10-04T03:34:03 *** aerique has joined #aichallenge 2010-10-04T03:38:30 *** Palmik has joined #aichallenge 2010-10-04T03:41:03 *** Palmik has quit IRC (Remote host closed the connection) 2010-10-04T03:43:46 *** Palmik has joined #aichallenge 2010-10-04T03:46:05 finally stabilized over 30 wpm (: 2010-10-04T03:48:14 *** vizier has joined #aichallenge 2010-10-04T03:49:53 wpm ? 2010-10-04T03:49:56 wins per minute? 2010-10-04T03:49:58 :) 2010-10-04T03:51:00 hehe 2010-10-04T03:51:12 i can't understand why is forbidden to use multiple bots 2010-10-04T03:51:30 like one would be aggressive, one would be defensive, one would be mix ... 2010-10-04T03:51:44 dunno 2010-10-04T03:51:47 how can you cheat with those 2010-10-04T03:52:14 say, 5 bots at max 2010-10-04T03:52:26 it would be problem if someone uploads 100000 bots :) 2010-10-04T03:59:05 Top 10 players: davidjliu(3782), dmj111(3767), cfaftw(3744), rsergio(3703), felixcoto(3682), Polsky(3660), EBraun(3658), sequoh(3642), youwillfail(3634), DocBot(3589) 2010-10-04T04:00:22 *** sigh has joined #aichallenge 2010-10-04T04:02:23 rabidus: I believe they didn't expect even 2500 people to join. They had to put something together to make the server go faster. So it really would be a problem for the organizers if we all had two or three bots :) 2010-10-04T04:03:25 Or would you split your time share on your three bots, so each of them plays only every third match? 2010-10-04T04:03:48 *** yasith__ has joined #aichallenge 2010-10-04T04:05:22 seriously, you can test multiple bots on tcp, one on the official server is enough 2010-10-04T04:05:43 *** yasith_ has quit IRC (Ping timeout: 240 seconds) 2010-10-04T04:06:12 how many people joined so far? 2010-10-04T04:06:24 Hey 1631 lines of code. That's less than I expected. After all I have learned a lot now about C++ streams, logging and loading of plugins. 2010-10-04T04:07:42 i'm probably at around 2500 :| yikes 2010-10-04T04:08:07 znutar: You can click on the last page of the ranking to see the number. That are all people who uploaded something working. ~300 of them have crashing/inactive bots though. 2010-10-04T04:08:33 another 2000 or so have extremely shit bots 2010-10-04T04:09:15 Yeah some get bitten by RageBot. 2010-10-04T04:09:34 lol, dmj still gets beaten by ragebot sometimes on tcp 2010-10-04T04:11:53 Somehow I lost interest in writing a bot. It became much more fun to learn new stuff while coding. For example I can now compile bots into the code or load them at runtime from a .so file. That was fun to see in the logs: "Bot not compiled in, checking for ./bots/RageBot.so... bot loaded!" :D 2010-10-04T04:16:16 *** Florian has joined #aichallenge 2010-10-04T04:17:34 I had to learn that in order to get a plug-in to find functions in the main program the main-program must be compiled for dynamic linking. Maybe next year I make the games plug-ins as well. 2010-10-04T04:17:36 Frontier: sry big response delay. but, if some site gets more popular what they expect, then it is good thing, i think. maybe few google ads in front page 2010-10-04T04:17:46 *** Florian95 has joined #aichallenge 2010-10-04T04:17:50 and that way few servers more 2010-10-04T04:18:02 or google would support some mor 2010-10-04T04:18:04 more 2010-10-04T04:18:15 maybe few cluster...... ;) 2010-10-04T04:18:43 Sure, but more servers means that the challenge is split up until they can add a feature that works on a cluster 2010-10-04T04:19:27 Ok.. they could split WEB, DB, "whatever server the action runs on" 2010-10-04T04:19:37 ah, it doesn't work like that now? 2010-10-04T04:19:41 But maybe that is already split. 2010-10-04T04:20:05 Ask amstan, the admin. 2010-10-04T04:20:14 i think most cpu time is taken by challenge-motor 2010-10-04T04:20:19 not web/mysql 2010-10-04T04:20:30 i mean like compiling and the fight-part 2010-10-04T04:21:14 I don't know how much this 'elo' calculation takes and how well it scales with entries and played matches. It's probably a notable part of the CPU load now. 2010-10-04T04:21:37 sounds sick 2010-10-04T04:21:39 .) 2010-10-04T04:22:02 its just an algorithm 2010-10-04T04:22:06 Compiling the C++ entries can also take its time 2010-10-04T04:22:13 like i said 2010-10-04T04:22:24 oh i see ^^ 2010-10-04T04:22:26 :) 2010-10-04T04:23:35 I wonder if some languages are really slow there when they are reloaded over and over. That said I'm confident with C++ ^^ 2010-10-04T04:24:53 wonder how big percent of participants are using c#, c++, java or other language which needs precompiling 2010-10-04T04:24:59 60% ? 2010-10-04T04:25:02 70% ? 2010-10-04T04:25:10 hmm... *looks at ranking* 2010-10-04T04:25:15 doubt its that high 2010-10-04T04:25:21 that says its something 80% 2010-10-04T04:25:35 rankings mean nothing at this point 2010-10-04T04:25:42 the top 10 is still mostly heuristic bots 2010-10-04T04:25:43 618 using python alone 2010-10-04T04:25:57 that's over 20% 2010-10-04T04:26:03 :o 2010-10-04T04:26:12 747 c++ 2010-10-04T04:26:20 c++ <3 2010-10-04T04:26:20 hwo do u know? did you count them? 2010-10-04T04:26:30 http://www.ai-contest.com/language_profile.php?lang=Python&page=7 2010-10-04T04:26:34 :) 2010-10-04T04:26:38 942 java 2010-10-04T04:26:52 Frontier: you can click on languages, countries, organisations etc. to see a leaderboard of that specific item 2010-10-04T04:27:12 they had links to such things last time with rankings across countries etc. 2010-10-04T04:27:13 293 c#, so its something like 2000 are using c#,c++ or java 2010-10-04T04:27:18 i'm unsure as to why they no longer have that 2010-10-04T04:27:41 2000 out of 2742 2010-10-04T04:27:52 70% is pretty close 2010-10-04T04:28:04 hmm... yeah 2010-10-04T04:28:32 anyone that uses java is a terrible person by nature :) 2010-10-04T04:28:32 quite cpu intensive 2010-10-04T04:28:49 tell that to rsergio 2010-10-04T04:28:57 or felixcoto 2010-10-04T04:28:58 :) 2010-10-04T04:29:02 i would if i could :P 2010-10-04T04:29:03 3 java bots in the top 10 2010-10-04T04:29:11 2 2010-10-04T04:29:20 @ranking 2010-10-04T04:29:21 sigh: Top 10 players: davidjliu(3798), dmj111(3777), cfaftw(3749), rsergio(3711), felixcoto(3694), EBraun(3665), sequoh(3650), Polsky(3649), youwillfail(3634), _iouri_(3593) 2010-10-04T04:29:28 I like Java. Compile it once, run it everywhere! And I like the syntax. 2010-10-04T04:29:44 ha, i'm surprised my bot went back in 2010-10-04T04:29:45 object.object.object.object(variable) 2010-10-04T04:29:48 ok, I hadn't refreshed my page 2010-10-04T04:29:50 And the standard library that comes with it. 2010-10-04T04:30:26 rabidus: Have you ever worked with function pointers macros and templates in C++ ^^ ? 2010-10-04T04:30:36 thank god no :D 2010-10-04T04:30:39 well 2010-10-04T04:30:40 Java is so .... clean and easy 2010-10-04T04:30:48 yes i have, but non like in java scale 2010-10-04T04:30:55 even my sister can use it ;) 2010-10-04T04:31:01 Frontier: whats c# then? 2010-10-04T04:31:04 java isn't easy 2010-10-04T04:31:11 i don't understand people thinking that 2010-10-04T04:31:20 c# is M$, nuf said 2010-10-04T04:31:21 python is clean and easy :P 2010-10-04T04:31:27 like you can't even take a string varialbe and just access elemetns like s[10] 2010-10-04T04:31:35 it's something like s.charAt(10); wtf is that 2010-10-04T04:31:43 haha, yeah 2010-10-04T04:32:03 its for the greater good! look how easy it is to parse java code 2010-10-04T04:32:13 lol 2010-10-04T04:32:15 thats not true 2010-10-04T04:32:19 not even overloaded operators anywhere 2010-10-04T04:32:20 python is better on parsing 2010-10-04T04:32:24 compared to c++ maybe 2010-10-04T04:32:24 my only dislikes for c++ so far are not supporting general multidimensional vectors or overloading operator precedence 2010-10-04T04:32:30 but not compared to most 2010-10-04T04:33:33 vector> davector; 2010-10-04T04:33:41 ^^ fails 2010-10-04T04:33:45 why 2010-10-04T04:33:54 vector< vector > davector; 2010-10-04T04:33:57 I use quite some python and Java GUI programs. They do their job :) I think both languages work well with larger projects. 2010-10-04T04:33:58 c++ doesn't like >> 2010-10-04T04:34:04 aah 2010-10-04T04:34:04 :D 2010-10-04T04:34:06 sry 2010-10-04T04:34:13 ok, it was pseudo :) 2010-10-04T04:34:15 haha, nah... just being pedantic 2010-10-04T04:34:19 :P 2010-10-04T04:34:31 sigh: another stupidity that would never have made it into java. 2010-10-04T04:34:47 Frontier: as I said, java is easier to parse comapred to c++ 2010-10-04T04:34:53 but not most other languages 2010-10-04T04:35:09 how is that stupid? 2010-10-04T04:35:21 and really, if your claim to fame is to have cleaner syntax than c++... then I'm sorry for your lots 2010-10-04T04:35:53 Also I have my difficulties with copy constructors now and then. When I have a vector and I add something with .push_back(...) it ends up being copied. 2010-10-04T04:37:36 sigh: Did you find the boost foreach macro yet? It is pretty nice to have and cleans up the code quite a bit when iterating over list<>s 2010-10-04T04:38:07 yeah, but does the server support boost yet? 2010-10-04T04:38:11 I haven't checked 2010-10-04T04:38:24 I will check on my next upload ^^ 2010-10-04T04:38:29 :) 2010-10-04T04:38:32 If not... well... 2010-10-04T04:38:44 work ahead. It is Ubuntu I think 2010-10-04T04:39:21 I remember a while back I uploaded some code with boost smart pointers and that failed 2010-10-04T04:39:35 I also remeber someone asking for boost in the forums, so maybe they put it in 2010-10-04T04:42:04 Server: Apache/2.2.8 (Ubuntu) PHP/5.2.4-2ubuntu5.12 with Suhosin-Patch mod_ssl/2.2.8 OpenSSL/0.9.8g mod_perl/2.0.3 Perl/v5.8.8 2010-10-04T04:44:05 I have a source linux distribution, thats why it was always installed on my box but not on Ubuntu. 2010-10-04T04:46:01 hmm.. only exotic or gui applications depend on it on my installation. Seems I was more lucky than I though. Ok then I just copy the .hpp file from /usr/include/boost into the uploaded zip file ;) 2010-10-04T04:48:04 "the .hpp"? what file are you including that does not depend on 10 other files? 2010-10-04T04:50:01 *** area has joined #aichallenge 2010-10-04T04:50:19 foreach.hpp does hopefully not include so much... does it? :/ 2010-10-04T04:50:42 hahahah 2010-10-04T04:51:16 (btw, 10 was a vast understatement) 2010-10-04T04:51:47 *** perror has joined #aichallenge 2010-10-04T04:52:26 *** HakanD has quit IRC (Read error: Connection reset by peer) 2010-10-04T04:53:11 hmm... I'll add some boost stuff and try uploading again 2010-10-04T04:53:58 cp -a /usr/include/boost ~/projects/planetwars/ 2010-10-04T04:53:58 -.- no comment! 2010-10-04T04:59:06 Top 10 players: davidjliu(3800), dmj111(3773), cfaftw(3751), rsergio(3707), felixcoto(3696), EBraun(3660), sequoh(3652), youwillfail(3640), Polsky(3609), DocBot(3590) 2010-10-04T04:59:08 Frontier: you realise that the upload limit is 2MB compressed :P 2010-10-04T05:08:57 whaaat 2010-10-04T05:09:05 top10 changes dramatically 2010-10-04T05:09:23 wheres dmj111 2010-10-04T05:09:38 oops :D 2010-10-04T05:09:47 whats wrong with my eyes 2010-10-04T05:12:59 *** keynes has joined #aichallenge 2010-10-04T05:24:59 *** O1athe has quit IRC (Ping timeout: 245 seconds) 2010-10-04T05:30:39 Frontier: I uploaded a bot that uses boost and it worked 2010-10-04T05:33:35 sigh: http://www.ai-contest.com/visualizer.php?game_id=5401316 :P 2010-10-04T05:34:11 antimatroid2: yeah, I know your bot owns mine :P 2010-10-04T05:35:14 I'm curious as to how my new upload will do, it's not "better" than my old bot... just different 2010-10-04T05:35:29 i still get beaten by people, i consider that unacceptable but i am yet to manage to do anything about it D: 2010-10-04T05:35:41 haha 2010-10-04T05:36:24 otoh, I just expect to lose and am pleasantly surprised every so often 2010-10-04T05:37:07 well, my bot's current rank is 613 so I don't expect to lose quite so much 2010-10-04T05:37:15 hehe 2010-10-04T05:37:20 you'd want to hope so :P 2010-10-04T05:37:25 :) 2010-10-04T05:39:23 hazard is at 90 and not far off hitting top 10 2010-10-04T05:39:27 just needs to get the right game 2010-10-04T05:39:52 ruilov is at 103 with more wins so not far behind 2010-10-04T05:39:59 neither has lost 2010-10-04T05:40:00 cool 2010-10-04T05:40:46 *** Palmik has quit IRC (Remote host closed the connection) 2010-10-04T05:41:12 *** Palmik has joined #aichallenge 2010-10-04T05:45:29 *** HakanD_ has joined #aichallenge 2010-10-04T05:50:32 woah, my bot massive fail on neutral stealing: http://www.ai-contest.com/visualizer.php?game_id=5404103 2010-10-04T05:50:36 turn ~25 2010-10-04T05:52:10 and quite good at winning (: 2010-10-04T05:53:03 well the opponent was ~500 2010-10-04T05:53:06 :P 2010-10-04T05:53:50 damn... need to fix this, this is a massive bug 2010-10-04T05:56:11 or just hope that stronger player won't allow me to fail at stealing neutrals :D 2010-10-04T05:59:06 Top 10 players: davidjliu(3815), dmj111(3763), cfaftw(3751), rsergio(3720), felixcoto(3706), EBraun(3675), sequoh(3650), jimrogerz(3638), youwillfail(3632), DocBot(3605) 2010-10-04T05:59:29 wait which planet? 2010-10-04T06:00:23 dw i see 2010-10-04T06:00:38 *** Accoun has quit IRC () 2010-10-04T06:01:49 ah, I see... my bot correctly calculated the number ships required to steal the neutral and sent them... then it saw that it would own that planet later, and hey, it's pretty close to the enemy so it should get some reinforcements 2010-10-04T06:01:53 stupid bot 2010-10-04T06:01:59 *** Arthur has quit IRC (Disconnected by services) 2010-10-04T06:02:20 *** Arthur has joined #aichallenge 2010-10-04T06:03:05 aha 2010-10-04T06:03:17 i wont take a neutral near the enemy 2010-10-04T06:03:32 my bot doesn't, it's stealing it FROM the enemy 2010-10-04T06:03:38 i really need to fix that to do it when i can 2010-10-04T06:03:48 opponent sends the 42 to capture the neutral first 2010-10-04T06:04:01 woah, there is a new person in the top ten 2010-10-04T06:04:08 jimrogerz from microsoft 2010-10-04T06:04:12 heh 2010-10-04T06:04:33 using C# :P 2010-10-04T06:05:09 ha, you resubmitted 2010-10-04T06:05:22 didn't seem to help game count 2010-10-04T06:06:38 first resubmission is because of first move loss 2010-10-04T06:06:51 i'm pretty sure the engine would have just got that wrong 2010-10-04T06:06:55 it handed you the win :P 2010-10-04T06:07:07 haha 2010-10-04T06:07:28 but both the engine and the error message agreed that I lost 2010-10-04T06:07:43 so *shrug* 2010-10-04T06:08:08 not the site :P 2010-10-04T06:08:14 i love how unsure it is 2010-10-04T06:08:24 it doesn't even seem to be consistent in the way that it's wrong 2010-10-04T06:11:53 *** otherAntimatroid has quit IRC (Quit: Leaving.) 2010-10-04T06:15:28 *** Haggis has joined #aichallenge 2010-10-04T06:31:38 *** mega1 has joined #aichallenge 2010-10-04T06:33:10 w...t...f... every simple fix I try for that bug makes it play worse 2010-10-04T06:33:47 sounds familiar 2010-10-04T06:35:02 welcome to the party :) 2010-10-04T06:36:30 feels like 'ok, i'll change just this into that and it'll be much better' and results are just opposite 2010-10-04T06:36:39 *** keynes has quit IRC (Ping timeout: 240 seconds) 2010-10-04T06:37:13 *** ratatata has joined #aichallenge 2010-10-04T06:37:18 *** Arthur has quit IRC (Disconnected by services) 2010-10-04T06:37:27 *** ratatata has quit IRC (Client Quit) 2010-10-04T06:37:40 *** Arthur_ has joined #aichallenge 2010-10-04T06:38:14 i always try to put up a bot that seems to beat my older ones 2010-10-04T06:38:23 i have no idea if that means it's actually better or not aha 2010-10-04T06:38:49 *** Accoun has joined #aichallenge 2010-10-04T06:40:38 *** dmj111 has joined #aichallenge 2010-10-04T06:43:49 *** gerard has quit IRC (Read error: Operation timed out) 2010-10-04T06:51:10 *** Zeiris_ has quit IRC (Ping timeout: 240 seconds) 2010-10-04T06:52:02 yay, I fixed it without making my bot worse with one huge ugly line of code! 2010-10-04T06:52:41 *** oPless has joined #aichallenge 2010-10-04T06:53:05 if ( pw.GetPlanet(dest_id)->FutureState( Map::Distance( source_id, dest_id ) ).owner == NEUTRAL ) continue; 2010-10-04T06:56:38 that is not huge xD 2010-10-04T06:57:52 if it doesn't fit in 80 columns I count it as huge :P 2010-10-04T06:58:46 pw => p, getplanets => gp, dest_id => dID, FutureState => FS etc. :) 2010-10-04T06:59:00 hha 2010-10-04T06:59:02 very readable :) 2010-10-04T06:59:06 Top 10 players: davidjliu(3828), dmj111(3777), cfaftw(3752), rsergio(3722), felixcoto(3715), EBraun(3690), sequoh(3651), youwillfail(3631), DocBot(3603), _iouri_(3598) 2010-10-04T06:59:32 if(p.GP(dID)->FS(Map::D(sID,dID)).o == NEUTRAL) 2010-10-04T06:59:49 awe, you didn't even replace NEUTRAL with 0 :P 2010-10-04T06:59:55 *aww 2010-10-04T06:59:58 oh 2010-10-04T07:00:12 if(!p.GP(dID)->FS(Map::D(sID,dID)).o) continue; 2010-10-04T07:00:18 haha, yeah 2010-10-04T07:00:25 :D 2010-10-04T07:00:50 takes just few seconds more to understand :) 2010-10-04T07:02:59 *** BtbN has joined #aichallenge 2010-10-04T07:04:01 wow i'm 7 now 2010-10-04T07:04:06 *** Blkt has joined #aichallenge 2010-10-04T07:04:13 not anymore :P 2010-10-04T07:04:28 @ranking 2010-10-04T07:04:29 sigh: Top 10 players: davidjliu(3834), dmj111(3778), cfaftw(3754), rsergio(3724), felixcoto(3716), EBraun(3684), sequoh(3655), jimrogerz(3652), youwillfail(3632), DocBot(3605) 2010-10-04T07:05:04 sigh: agree on the 80 chars! I have coworkers who think their wide screen monitors means never having to hit return. 2010-10-04T07:05:19 *** jimmerton has quit IRC (Quit: jimmerton) 2010-10-04T07:05:26 ha, my lines get rather wide sometimes 2010-10-04T07:05:55 yeah... in my bot they do sometimes. at work, I try not to. 2010-10-04T07:06:04 _try_ 2010-10-04T07:06:04 dmj111: I try to avoid it, but I don't get too hung up about it 2010-10-04T07:06:23 good thing i don't work as a programmer :P 2010-10-04T07:06:25 if I decide it's more readable as a single line then I keep it that way 2010-10-04T07:06:45 i prefer one big function to do something that lots of incrementally broken up ones, people hate that 2010-10-04T07:06:57 sigh: that is the best point .... do what makes it more readable. 2010-10-04T07:07:07 antimatroid2: you'll come to regret that :P 2010-10-04T07:07:13 *** keynes has joined #aichallenge 2010-10-04T07:07:22 not one function for my whole bot 2010-10-04T07:07:33 but say i want generate future states, i'll do that as one big function 2010-10-04T07:07:37 rather than split it all up 2010-10-04T07:07:45 smaller functions are easier to get your head around, know what variables are in scope, what variables can get modified, etc 2010-10-04T07:08:06 i find it harder to read the actual function doing anything when i need to skip all over the place 2010-10-04T07:08:39 you shouldn't need to skip all over the place, the sub-functions should have obvious names with obvious roles 2010-10-04T07:09:04 yeah, but then when you go back through and look, to see that code you need to skip somewhere else, then go back, or somewhere else 2010-10-04T07:09:15 i find it much less confusing to just trudge through one bigger function 2010-10-04T07:10:22 yeah, again... do what you find more readable 2010-10-04T07:10:27 hi aerique 2010-10-04T07:10:47 but come back after you're maintaining code you haven't touched in 6 months :P 2010-10-04T07:24:52 *** jmpespxoreax has joined #aichallenge 2010-10-04T07:35:24 *** vizier has quit IRC (Ping timeout: 265 seconds) 2010-10-04T07:40:00 *** Arthur_ has quit IRC (Ping timeout: 240 seconds) 2010-10-04T07:48:20 Blkt: hi :) 2010-10-04T07:51:59 @rankings 2010-10-04T07:52:00 HakanD_: Top 10 players: davidjliu(3840), dmj111(3774), cfaftw(3753), rsergio(3723), felixcoto(3712), EBraun(3684), sequoh(3658), youwillfail(3629), DocBot(3611), _iouri_(3604) 2010-10-04T07:52:42 changing fast (: 2010-10-04T07:59:07 Top 10 players: davidjliu(3837), dmj111(3771), cfaftw(3751), rsergio(3720), felixcoto(3709), EBraun(3678), sequoh(3660), youwillfail(3627), DocBot(3610), _iouri_(3602) 2010-10-04T08:02:45 *** iFire has quit IRC (Read error: Connection reset by peer) 2010-10-04T08:04:53 *** iFire has joined #aichallenge 2010-10-04T08:05:56 *** delt0r_ has quit IRC (Read error: Operation timed out) 2010-10-04T08:07:22 *** delt0r_ has joined #aichallenge 2010-10-04T08:08:00 *** Florian95 has quit IRC (Ping timeout: 264 seconds) 2010-10-04T08:08:45 *** jaspervdj has joined #aichallenge 2010-10-04T08:22:12 *** Florian95 has joined #aichallenge 2010-10-04T08:27:19 *** krokokrusa has joined #aichallenge 2010-10-04T08:30:25 *** iFire has quit IRC (Read error: Connection reset by peer) 2010-10-04T08:32:19 *** iFire has joined #aichallenge 2010-10-04T08:36:33 *** BotOwner has joined #aichallenge 2010-10-04T08:40:53 *** sinan has quit IRC (Ping timeout: 252 seconds) 2010-10-04T08:41:08 *** BotOwner has quit IRC (Ping timeout: 265 seconds) 2010-10-04T08:41:35 i put another post on the strategy forum that might be helpful for those who are starting out :) http://www.ai-contest.com/forum/viewtopic.php?f=17&t=858 2010-10-04T08:43:34 mcmillen: nice... although equally important is not getting "sniped" yourself :) 2010-10-04T08:44:29 thats the tricky part (: 2010-10-04T08:45:07 my bot just refuses to attack neutrals that are closer to the enemy 2010-10-04T08:45:44 if you are that strong that you can take and hold a neutral in your enemy's sphere of influence, then you are strong enough to take out the enemy directly 2010-10-04T08:56:05 *** Migi32 has joined #aichallenge 2010-10-04T08:56:21 hi a1k0n are you here? I'd like to ask you something 2010-10-04T08:58:52 sigh: yeah, that's why my "turn 1" tutorial says don't attack planets that are closer or equally distant to your enemy 2010-10-04T08:59:07 Top 10 players: davidjliu(3817), dmj111(3769), cfaftw(3741), rsergio(3716), felixcoto(3695), ruilov(3679), sequoh(3666), EBraun(3661), youwillfail(3625), DocBot(3601) 2010-10-04T08:59:22 of course because it's a game the question is always how you can do a thing without letting your opponent do the same to you :) 2010-10-04T09:00:34 what do you mean? 2010-10-04T09:00:37 *** iFire has quit IRC (Read error: Connection reset by peer) 2010-10-04T09:01:17 i mean, any strategy that you come up with should also make you ask yourself "how do i prevent the opponet from doing this to me?" :) 2010-10-04T09:01:28 *** sinan has joined #aichallenge 2010-10-04T09:01:42 oh, sure 2010-10-04T09:02:35 currently my bot doesn't care too much and just tried to aquire planets 2010-10-04T09:02:37 *** iFire has joined #aichallenge 2010-10-04T09:03:10 *just tries 2010-10-04T09:03:54 anyone uploaded a bot the last hour? if so, did it play any games yet? 2010-10-04T09:04:22 btw. its very difficult to make a score of a planet, like 'how important that planet is' 2010-10-04T09:04:40 few times tried and every time it sucked one way or another 2010-10-04T09:05:45 aerique: I uploaded a bot 3 hours ago and it's played 10 games 2010-10-04T09:06:26 rabidus: yup, good planet selection is key 2010-10-04T09:08:59 what I find well is creating a evaluation function that allows me to compare neutrals to enemy planets 2010-10-04T09:09:05 *find hard 2010-10-04T09:09:10 hmmh 2010-10-04T09:09:22 i make it kinda just opposite 2010-10-04T09:09:31 i check that how many ships i can waste for neutral 2010-10-04T09:09:39 i dont care about enemy :) 2010-10-04T09:09:50 it still is only a enemy 2010-10-04T09:10:02 but, i'm noob 2010-10-04T09:10:16 well, what I mean is that I can compare neutrals to each other, enemy to each other... but hard to determine when to attack and enemy instead of a neutral 2010-10-04T09:10:29 yep, thats true 2010-10-04T09:14:07 I've come up with a kind of kludge which allows me to compare them on the same scale... but I don't like it very much 2010-10-04T09:14:42 it must be made based on production expectations ... or something 2010-10-04T09:14:52 ie. expected value 2010-10-04T09:15:23 similar concept, yes 2010-10-04T09:15:38 plus neutral sniping etc.. 2010-10-04T09:15:43 quite complex 2010-10-04T09:15:44 :) 2010-10-04T09:16:05 :) 2010-10-04T09:18:35 *** sigh has quit IRC (Remote host closed the connection) 2010-10-04T09:19:15 *** delt0r_ has quit IRC (Ping timeout: 276 seconds) 2010-10-04T09:19:51 it's also a good idea to "steal" neutral planets if the enemy attacks it wih just a little more than needed just send a small number of ships that arive 1 turn after your enemy, and the planet is yours 2010-10-04T09:20:11 thats what i call neutral sniping :) 2010-10-04T09:20:18 And it's a good idea, to check neutral planets you plan to attack against this in your AI 2010-10-04T09:24:29 yeah, and to prevent that, if you attack a neutral, send a small fleet to it the turn after as well 2010-10-04T09:24:40 *** Appleman1234 has quit IRC (Ping timeout: 240 seconds) 2010-10-04T09:24:41 if he did the same, keep sending those fleets 2010-10-04T09:25:45 you can destroy a ton of bots this way, cause they think "oh, if I send a small fleet now I can snipe this neutral" 2010-10-04T09:25:48 but they can't 2010-10-04T09:29:26 *** AJC has joined #aichallenge 2010-10-04T09:33:24 true 2010-10-04T09:40:38 *** HakanD_ has quit IRC (Ping timeout: 264 seconds) 2010-10-04T09:40:59 are there any places to look for strategies? 2010-10-04T09:41:22 forum 2010-10-04T09:41:25 :) 2010-10-04T09:42:04 here 2010-10-04T09:42:05 :) 2010-10-04T09:42:38 *** JensTi has joined #aichallenge 2010-10-04T09:43:12 *** holmar has quit IRC (Ping timeout: 272 seconds) 2010-10-04T09:43:59 it's a shame you can't get your submission to challenge a particular bot 2010-10-04T09:45:42 it would be kinda .. wrong 2010-10-04T09:45:53 you could use particular bot weaknesses 2010-10-04T09:46:14 *** BtbN_ has joined #aichallenge 2010-10-04T09:46:47 oh, not in any leaderboard-meaningful way 2010-10-04T09:48:40 but there is a chance 2010-10-04T09:50:00 *** BtbN has quit IRC (Ping timeout: 245 seconds) 2010-10-04T09:50:43 *** BtbN has joined #aichallenge 2010-10-04T09:53:03 *** fawek has joined #aichallenge 2010-10-04T09:53:20 *** BtbN_ has quit IRC (Ping timeout: 272 seconds) 2010-10-04T09:58:49 *** nullkuhl_ has joined #aichallenge 2010-10-04T09:59:07 Top 10 players: davidjliu(3817), ruilov(3813), dmj111(3762), cfaftw(3734), rsergio(3702), felixcoto(3682), sequoh(3661), EBraun(3659), youwillfail(3615), DocBot(3605) 2010-10-04T09:59:21 *** BotLoder has joined #aichallenge 2010-10-04T10:00:09 i have just submitted my first code trial , how often does mybot plays against another ? 2010-10-04T10:01:11 *** BtbN has quit IRC (Ping timeout: 252 seconds) 2010-10-04T10:03:04 About once or twise in an Hour.. 2010-10-04T10:10:05 BotLoder: and once i resubmit, my ranking and score gets reset, yeah ? 2010-10-04T10:10:39 *** optimum has joined #aichallenge 2010-10-04T10:12:10 Yup You get a clean slate Everytime you resub.. 2010-10-04T10:22:44 *** virtualmike has joined #aichallenge 2010-10-04T10:24:32 *** JensTi has quit IRC (Quit: Leaving.) 2010-10-04T10:30:20 *** sinan has quit IRC (Ping timeout: 252 seconds) 2010-10-04T10:30:53 *** fawek has quit IRC (Ping timeout: 252 seconds) 2010-10-04T10:36:23 *** fawek has joined #aichallenge 2010-10-04T10:42:36 *** BtbN has joined #aichallenge 2010-10-04T10:44:27 *** rebelxt has joined #aichallenge 2010-10-04T10:45:31 past few days my bot has been playing something like 3 times / hours 2010-10-04T10:45:32 -s 2010-10-04T10:45:52 hey all 2010-10-04T10:47:54 rabidus, nice. mine gets ~ 2 games per hour. pretty good 2010-10-04T10:56:30 *** rebelxt has quit IRC (Quit: Page closed) 2010-10-04T10:56:45 *** sinan has joined #aichallenge 2010-10-04T10:59:08 Top 10 players: davidjliu(3818), ruilov(3797), dmj111(3759), cfaftw(3733), rsergio(3699), felixcoto(3682), sequoh(3675), EBraun(3666), youwillfail(3615), DocBot(3603) 2010-10-04T11:04:38 *** keyne has joined #aichallenge 2010-10-04T11:05:15 *** keynes has quit IRC (Ping timeout: 264 seconds) 2010-10-04T11:08:22 ah, my bot is working on the official server. it just doesn't play a game right away anymore 2010-10-04T11:08:57 right away after submitting that is 2010-10-04T11:09:56 *** virtualmike has quit IRC (Quit: Page closed) 2010-10-04T11:11:23 *** BtbN has quit IRC (Quit: Verlassend) 2010-10-04T11:11:24 *** virtualmikeai has joined #aichallenge 2010-10-04T11:12:13 bah, my dev bot seems to be breaking the rules 2010-10-04T11:14:25 *** AJC has quit IRC (Ping timeout: 272 seconds) 2010-10-04T11:14:55 Is there anything that prevents an Enemy bot from Sending a ton of Zero sized fleets Just to get you get your bot timed out?.. A Lot of bots have fleet loops.. it would be nice if the game engin filterd zero sized Oders.. 2010-10-04T11:15:52 BotLoder: it does 2010-10-04T11:15:54 *** BtbN has joined #aichallenge 2010-10-04T11:17:24 aerique: the last time i checked. there was nothing stoping me from sending zero sized fleets... 2010-10-04T11:17:44 *** Olathe has joined #aichallenge 2010-10-04T11:20:24 You can send them, but they're just ignored by the server 2010-10-04T11:20:45 *** AJC has joined #aichallenge 2010-10-04T11:20:48 And aren't passed to the other bot 2010-10-04T11:22:14 Hope your right.. I dont want to have to worry about that.. 2010-10-04T11:22:46 *** Olathe has quit IRC (Ping timeout: 276 seconds) 2010-10-04T11:23:56 BotLoder: if you just merge fleets then you don't have to worry anyway 2010-10-04T11:24:21 but yeah, the server ignores 0 size fleets 2010-10-04T11:24:33 *** Itkovian has quit IRC (Quit: Itkovian) 2010-10-04T11:25:00 BotLoder: by "merge fleets" i mean by when and where they arrive (not including where they departed from as that is irrelevant information) 2010-10-04T11:27:23 that could be useful to cut down processing time.. thanks .. 2010-10-04T11:27:39 O_o interesting, my new bot isn't compling on the server, but is fine locally 2010-10-04T11:27:56 *** sinan has quit IRC (Ping timeout: 245 seconds) 2010-10-04T11:29:21 *** aerique has quit IRC (Quit: ...) 2010-10-04T11:31:27 *** yasith_ has joined #aichallenge 2010-10-04T11:33:49 *** yasith__ has quit IRC (Ping timeout: 264 seconds) 2010-10-04T11:36:09 the engine merges fleets, too 2010-10-04T11:44:20 *** McLeopold has joined #aichallenge 2010-10-04T11:48:12 Zannick: it fundamentally can't ignore source planet in that merging though 2010-10-04T11:48:42 Zannick: that is, it can only merge by (src, dst, turn), not simply by (dst, turn) 2010-10-04T11:48:57 ah, i missed that part of the discussion 2010-10-04T11:48:59 since it is required to report the source planet in the game state 2010-10-04T11:51:13 *** nullkuhl_ has quit IRC (Ping timeout: 264 seconds) 2010-10-04T11:52:36 *** mceier has joined #aichallenge 2010-10-04T11:57:25 *** wvdschel has quit IRC (Quit: Leaving) 2010-10-04T11:59:08 Top 10 players: davidjliu(3810), ruilov(3796), dmj111(3761), cfaftw(3731), rsergio(3689), felixcoto(3686), sequoh(3672), EBraun(3671), DocBot(3610), youwillfail(3605) 2010-10-04T12:00:38 *** bhasker has quit IRC (Quit: bhasker) 2010-10-04T12:10:50 *** rebelxt has joined #aichallenge 2010-10-04T12:11:46 *** AJC has quit IRC (Quit: Leaving) 2010-10-04T12:14:07 wow another person at the top today 2010-10-04T12:24:16 it occurs to me that my algorithm could benefit from some map preprocessing time... moreso than most algorithms i have though of. it of course makes sense that we don't really get a chance to do that, but it's still kind of disappointing that i won't get to exploit that 2010-10-04T12:24:26 *thought of 2010-10-04T12:26:05 *** bhasker has joined #aichallenge 2010-10-04T12:29:23 morning 2010-10-04T12:29:48 *** keyne has quit IRC (Remote host closed the connection) 2010-10-04T12:36:38 *** delt0r_ has joined #aichallenge 2010-10-04T12:39:43 *** yasith_ has quit IRC (Ping timeout: 255 seconds) 2010-10-04T12:41:33 *** kozlovsky has joined #aichallenge 2010-10-04T12:43:20 *** BtbN has quit IRC (Quit: Verlassend) 2010-10-04T12:43:48 *** BtbN has joined #aichallenge 2010-10-04T12:46:28 *** lovebob has joined #aichallenge 2010-10-04T12:47:26 *** lovebob has joined #aichallenge 2010-10-04T12:47:50 wow lot of people in here 2010-10-04T12:49:31 WormBot takes down dmj111: http://72.44.46.68/canvas?game_id=304714 2010-10-04T12:49:32 so.. i had a dream, that the len(ppl) i here was decreasing 2010-10-04T12:49:35 *** BotLoder has quit IRC (Ping timeout: 265 seconds) 2010-10-04T12:49:43 there were like 20 2010-10-04T12:50:16 amstan: you're a contest organizer, right? 2010-10-04T12:50:21 yeah 2010-10-04T12:50:43 it was weird though 2010-10-04T12:50:43 amstan: http://ai-contest.com/forum/viewtopic.php?f=18&t=840 2010-10-04T12:50:51 weird how? 2010-10-04T12:51:31 well.. the dream+the reason why ppl were going down 2010-10-04T12:51:35 anyway.. lol 2010-10-04T12:51:48 McLeopold: you sould email jeff if you want the results table 2010-10-04T12:52:12 where do I find his email? 2010-10-04T12:52:44 or do you mean private message on the forum? 2010-10-04T12:52:47 *** virtualmikeai has quit IRC (Quit: Page closed) 2010-10-04T12:52:52 "Jeff Cameron" 2010-10-04T12:52:58 thanks 2010-10-04T12:53:17 *** yasith_ has joined #aichallenge 2010-10-04T12:53:22 i'm still waiting on some tarballs :) 2010-10-04T12:53:42 jmcarthur: which tarballs? 2010-10-04T12:53:59 all game replays, maps, and bot stats 2010-10-04T12:54:08 probably distributed via bittorrent 2010-10-04T12:54:20 updated maybe once a week 2010-10-04T12:54:22 would be ideal 2010-10-04T12:54:32 someone should work up a patch for issue 107, just make a script that dumps the data to a file and can be run from a cron job 2010-10-04T12:54:35 jmcarthur: yeah, that's exactly what I want. 2010-10-04T12:54:50 http://code.google.com/p/ai-contest/issues/detail?id=107#makechanges 2010-10-04T12:57:01 Janzert: problem is that the data is huge.. 2010-10-04T12:57:34 amstan: that's why I'm asking for just a simple table to start with. 2010-10-04T12:57:48 I had a dream that the UN was a tool for american oil politics, also quite weird 2010-10-04T12:57:49 we can probably use the game_info query 2010-10-04T12:58:18 amstan: go on... 2010-10-04T12:58:31 McLeopold: it has everything you need 2010-10-04T12:58:53 i believe danielvf took the actual replays out of game_info in the interest of making the database disk behavior sane 2010-10-04T12:59:12 Top 10 players: davidjliu(3817), dmj111(3761), cfaftw(3719), rsergio(3681), ruilov(3681), felixcoto(3680), EBraun(3674), sequoh(3668), DocBot(3622), youwillfail(3606) 2010-10-04T12:59:19 or that was just games 2010-10-04T12:59:47 I could just automate a webcrawler, but I'd rather not have to do that. 2010-10-04T13:00:10 yeah.. the table size is like 2 GB now 2010-10-04T13:00:34 My last game id is > 5 million 2010-10-04T13:06:23 *** jesionaj has joined #aichallenge 2010-10-04T13:14:24 you could maybe keep that extra info temporarily 2010-10-04T13:14:51 a1k0n, can I ask you a question? 2010-10-04T13:16:12 (also, seems like ruilov's spike to 4000 and dmj's fall were only temporary) 2010-10-04T13:18:24 *** endunio has joined #aichallenge 2010-10-04T13:21:12 hi there, can anyone help me with tcp_server? i dont know how to set apache to parse binary files 2010-10-04T13:22:02 *** fawek has quit IRC (Remote host closed the connection) 2010-10-04T13:22:24 endunio: I do not understand your request 2010-10-04T13:23:31 amstan: just wondering, in your reply about submitting lua as a C++ bot that includes the full lua interpreter, are you saying this is probably ok or not ok? 2010-10-04T13:23:41 ok 2010-10-04T13:23:55 as you can see, our rules are pretty open 2010-10-04T13:24:07 we allow everything as long as it goes nicely with the sandbox 2010-10-04T13:24:07 thanks, that had been what I thought it would be as well but wasn't sure 2010-10-04T13:24:19 the only thing we don't allow is network communications and file writing 2010-10-04T13:24:31 * Janzert nods 2010-10-04T13:24:32 and os threads 2010-10-04T13:24:36 amstan, why no file writing? 2010-10-04T13:24:46 Migi32: because it's hard to keep track of it 2010-10-04T13:24:48 actually 2010-10-04T13:24:54 i think you're allowed file writing 2010-10-04T13:24:57 dwchandler: tcp_server compiles with canvas, getmaps, getplayer and getratings, how to configure apache to work with it? 2010-10-04T13:25:03 but don't expect it to be there after the match 2010-10-04T13:25:15 there is a tight limit on file handles though 2010-10-04T13:25:18 ah ok :) 2010-10-04T13:25:30 well there'd be no way to access it anyway 2010-10-04T13:25:31 jmcarthur: but that's not really an intentional rule, it's just because of the sandbox 2010-10-04T13:25:34 yeah 2010-10-04T13:26:36 endunio: see apachedocs about enabling CGI, ExecCGI directive, etc. 2010-10-04T13:26:55 ok, i'll check it, thanks for directions 2010-10-04T13:28:30 McLeopold, this wormbot, is it a complete rewrite? 2010-10-04T13:28:46 *** krokokrusa has quit IRC (Ping timeout: 255 seconds) 2010-10-04T13:32:03 *** Naktibalda has joined #aichallenge 2010-10-04T13:34:10 Migi32: not completely. I've written my own framework. It's built on that. 2010-10-04T13:35:21 The idea was a rewrite. 2010-10-04T13:35:59 I just tried to build a bot that would look cool when playing, yet be good enough to survive long enough that people would see it being cool. 2010-10-04T13:36:34 lol 2010-10-04T13:36:47 it makes for interesting games 2010-10-04T13:36:47 http://72.44.46.68/canvas?game_id=305806 2010-10-04T13:37:38 but in many of your games there are small planets building up stuff that you don't really use 2010-10-04T13:38:00 The idea is this: each planet picks some good targets, then waits for enough ships to punch through and keeps sending re-enforcements. 2010-10-04T13:38:19 Thus the "worm" part. 2010-10-04T13:38:38 It only sends toward the enemy, not away from. 2010-10-04T13:38:58 It will pause if there is some incoming fleets. 2010-10-04T13:39:07 That's about it. 2010-10-04T13:39:18 Build that and you have a fairly good bot. 2010-10-04T13:39:54 Of course "good targets" is where all the magic is. :) 2010-10-04T13:40:28 *** yasith_ has quit IRC (Ping timeout: 255 seconds) 2010-10-04T13:40:45 that's the secret sauce in any bot 2010-10-04T13:41:25 well, fleet size, too, i guess 2010-10-04T13:41:34 jmcarthur: yes, but it's also interesting how few bots have good defense 2010-10-04T13:41:59 yeah most people are thinking about what tactics to use without considering how others will use those tactics against them 2010-10-04T13:42:30 I figure saving a planet is the same as taking one. 2010-10-04T13:42:33 defense is less "fun" to think about, perhaps 2010-10-04T13:42:59 my bot isn't even really going to see defense and offense as different things 2010-10-04T13:43:11 Yeah, exactly. 2010-10-04T13:43:14 jmcarthur: bingo 2010-10-04T13:43:24 A planet is a planet, who care who owns it. 2010-10-04T13:43:36 well, i do care who owns it 2010-10-04T13:43:47 it's just that i care who owns it in the future more than i care about right now 2010-10-04T13:44:05 more like "this planet is important. just because I own it doesn't mean I stop dealing with it." 2010-10-04T13:44:28 dwchandler: yeah, that is what I was trying to say 2010-10-04T13:44:54 are you paying attention Migi32? 2010-10-04T13:45:14 I was doing something else for 2 seconds 2010-10-04T13:45:16 but I am now :P 2010-10-04T13:45:57 in fact it is usually easier to hold a planet than to get a new one, so if you factor ease of taking a planet into your value function (along with payoff, what your opponent is going to do, etc.) then it should be easy to see when to defend and when to attack 2010-10-04T13:46:02 *** wvdschel has joined #aichallenge 2010-10-04T13:46:03 McLeopold, why? 2010-10-04T13:46:12 http://72.44.46.68/canvas?game_id=305200 2010-10-04T13:46:26 WormBot can punch through lines, but does not have good defense 2010-10-04T13:46:30 or rather, it should be easy to treat it as the same action 2010-10-04T13:46:35 also, losing a planet for even 1 turn leaves you with less ships 2010-10-04T13:46:38 *** kozlovsky has quit IRC (Ping timeout: 265 seconds) 2010-10-04T13:46:58 yeah, every time a planet changes hands even for only one turn, somebody just lost growth ships 2010-10-04T13:47:08 and the other gained 2010-10-04T13:47:29 so noticeable with hotly contested planets 2010-10-04T13:48:03 *** jesionaj has quit IRC (Ping timeout: 240 seconds) 2010-10-04T13:49:07 http://72.44.46.68/canvas?game_id=306062 2010-10-04T13:49:24 what vcsala lose his lead, all to a lack of defense 2010-10-04T13:49:29 *watch 2010-10-04T13:50:41 interesting games, McLeopold 2010-10-04T13:51:34 *** justin_pdx has joined #aichallenge 2010-10-04T13:53:39 *** yasith_ has joined #aichallenge 2010-10-04T13:58:56 McLeopold, your bot is susceptible to neutral stealing 2010-10-04T13:58:59 *** bhasker_ has joined #aichallenge 2010-10-04T13:59:07 watch that last game, turn 26 2010-10-04T13:59:11 Top 10 players: davidjliu(3821), dmj111(3761), cfaftw(3716), felixcoto(3691), rsergio(3675), sequoh(3668), EBraun(3661), ruilov(3630), DocBot(3625), youwillfail(3612) 2010-10-04T14:00:39 well, I believe it was a bad move to try to steal that neutral in the first place 2010-10-04T14:01:38 how do you think this rule would work: never try to capture a neutral if it's closer to the enemy than to you? 2010-10-04T14:01:43 WormBot? yeah, I don't even try sniping or anti-sniping 2010-10-04T14:02:06 *** bhasker has quit IRC (Ping timeout: 245 seconds) 2010-10-04T14:02:16 Migi32: make a bot with that rule and without. See who wins. :) 2010-10-04T14:03:19 *** Florian95 has quit IRC (Ping timeout: 276 seconds) 2010-10-04T14:04:03 McLeopold, it'd be a draw considering my bot plays go rather than PlanetWars ;) 2010-10-04T14:04:13 capture a neutral if you've scored it high 2010-10-04T14:04:26 high enough 2010-10-04T14:05:38 dwchandler, well that rule would translate to altering the scoring function with something like if (closer to enemy than to me) { score -= 10000; } 2010-10-04T14:11:35 *** felixcoto has joined #aichallenge 2010-10-04T14:13:31 migi i have tried both heuristics i think it matters at what stage of the game you are at 2010-10-04T14:13:46 and how much you overpower your opponent by 2010-10-04T14:13:51 *** Naktibalda has quit IRC (Read error: Connection reset by peer) 2010-10-04T14:14:04 if you have more production/ships than enemy it doesn't really matter much since you can usually reinforce the neutral 2010-10-04T14:14:07 but if tis a close tie 2010-10-04T14:14:12 you probably don't want to burn ships on a neutral 2010-10-04T14:14:52 and early/mid game its better to be conservative 2010-10-04T14:15:01 unless you can steal the planet 2010-10-04T14:16:11 sometimes capturing neutrals behind lines can also kill you if you take it at the wrong tme and shift the ships balance to the enemy 2010-10-04T14:16:23 the enemy will just punch through your front line 2010-10-04T14:18:18 oh ok thanks, bhasker_ :D 2010-10-04T14:18:21 Migi32: I'm suggesting making a good enough scoring system so you don't have to tweak in specific rules ;-) 2010-10-04T14:18:43 fleet re-direction (i.e. ability to change fleet's destination planet during flight at any time) would make Planet Wars much more complex. games would be crazy fun to watch though ;-) 2010-10-04T14:18:48 i'd be curious to modify engine and bot to support this after the contest is over. anyone else interested to try this some time in December? 2010-10-04T14:19:32 rebelxt, I think it'd actually make the game easier 2010-10-04T14:19:43 you never commit any ships 2010-10-04T14:19:48 i am still trying to get that right dwc :-\ sadly my heuristics need crutches at times, they make the bot do stupid things otherwise 2010-10-04T14:19:51 yeah, your ships are effectively invulnerable once produced 2010-10-04T14:20:01 you can send them out, as long as they don't reach half-way you risk nothing 2010-10-04T14:20:09 I keep messing with thing in my heuristic based bot, and everything I try is better in some situations and worse in others 2010-10-04T14:20:23 *** perror has quit IRC (Quit: Bye all !) 2010-10-04T14:20:30 pretty much the same problem i have with mine 2010-10-04T14:20:42 all heuristics are an ad hoc, buggy implementation of a good scoring system 2010-10-04T14:20:56 *** coventry has joined #aichallenge 2010-10-04T14:21:05 hmm i am not sure i understand the difference? 2010-10-04T14:21:16 isn't a score calculated using some kind of heuristic? 2010-10-04T14:21:22 migi32, interesting point, at a first glance it feels more complex.. Galcon allows fleet re-direction. really interesting dynamic games 2010-10-04T14:21:33 unless it's calculated based on the underlying uncomputable value function of the game 2010-10-04T14:21:51 *** jesionaj has joined #aichallenge 2010-10-04T14:21:52 which is actually computable i should add, just not in any reasonable amount of time 2010-10-04T14:22:00 right 2010-10-04T14:22:06 Anyone got any examples where deep search would pay off with better tactics than heuristics? 2010-10-04T14:22:23 so heuristics must be used in exploring and pruning the game tree 2010-10-04T14:22:31 *** hellman has quit IRC (Remote host closed the connection) 2010-10-04T14:22:51 coventry, nope since the deepest we've been able to search is 4 turns, which was on a map with 3 planets 2010-10-04T14:23:13 we being me anyway 2010-10-04T14:23:20 dunno what anyone else has achieved 2010-10-04T14:25:03 hey guys, if anyone wants to share a useless metric - what's your bot's LOC (lines of code) at the moment? mine is ~ 400 lines of Python code at this point (not counting some basic utility functions though). 2010-10-04T14:25:25 I've looked at a lot of games, and I can't think of any where I saw a tactically complex situation. Doesn't mean there aren't any, but it sounds like a lot of effort to search for something which might not be there. Has anyone tried to devise some tactically complex situations by hand? 2010-10-04T14:25:59 over 9000! :P 2010-10-04T14:26:06 just kidding, about 1000 2010-10-04T14:26:15 most being utility functions that I don't use 2010-10-04T14:26:16 1219, including all utility functions (c++) 2010-10-04T14:26:35 1848 2010-10-04T14:26:58 a1k0n: what was your LOC? 2010-10-04T14:27:02 for tron? 2010-10-04T14:27:11 coventry: well, there aren't many situations where it's so tactically simple that it's obvious what each player must do 2010-10-04T14:27:14 at least not to me 2010-10-04T14:27:17 official is 541 2010-10-04T14:27:32 amstan: 992 MyTronBot.cc 2010-10-04T14:28:20 Although, there was this game I saw this morning, where davidjliu divided his attention between two target planets between roughly move 20 and move 56. That was a bit complex, but I think he could have gotten the same result by hammering one planet or the other. http://ai-contest.com/visualizer.php?game_id=5424973 2010-10-04T14:28:21 coventry: I test my algorithms against hypothetical simple cases where I know what the optimal move is to see if it gets the same result. 2010-10-04T14:28:22 cool stats! i'm at 1900 counting utilities ("unofficial" python kit is about 1500 lines...) 2010-10-04T14:28:22 i have 540 lines of C++ now, only some of which is used, heh 2010-10-04T14:28:55 a1k0n, 5 versions of scoring functions? ;-) that's what i have heh 2010-10-04T14:28:57 you have to divide by 5 for c++ 2010-10-04T14:29:12 :) 2010-10-04T14:29:16 i'm trying something new but i have no idea whether it's a stupid idea or not 2010-10-04T14:29:38 i'm basically not going to settle for tweaking scoring functions just yet. 2010-10-04T14:29:43 a1k0n: as long as you play tcp against WormBot, it's worth it 2010-10-04T14:29:49 (and that will probably lose) 2010-10-04T14:30:14 McLeopold: is WormBot not as good as your other entry, or did everyone else get better? 2010-10-04T14:30:28 No, it sucks. It's just fun. 2010-10-04T14:30:44 a1k0n, same here - re-writing my bot to simplify it (it was getting out of hand!), and also putting scoring function tweaking on hold 2010-10-04T14:30:50 *** krokokrusa has joined #aichallenge 2010-10-04T14:30:56 What is wormbot? 2010-10-04T14:31:21 a1k0n: there's also deep stochastic search, but that may be hard to apply 2010-10-04T14:31:28 a1k0n: Yes, it's not obvious what to do, but there aren't any obvious deep tactical lines like there are in chess or go, either. I've been trying to imagine what that would look like, and the best I can come up with is splitting your forces between two planets to overload the enemy's defenses and force them to choose which one to pick. But as I said, I haven't been able to think of a specific layout where that would work better th 2010-10-04T14:31:29 just hammering the one planet. 2010-10-04T14:31:32 justin_pdx: http://72.44.46.68/canvas?game_id=306582 2010-10-04T14:31:46 I had a fun version for a bit. It did the normal sucky thing, but when it had higher growth & number of ships it sat tight 2010-10-04T14:32:11 then if it dropped below on either stat then it exploded 2010-10-04T14:32:24 really fun to see which other bots lost to that 2010-10-04T14:32:36 coventry: dmj111 effective launches a giant fleet that lands and counter-launches multiple targets. 2010-10-04T14:32:45 coventry: oh, right. yes, so it may be difficult to find the optimal move, but the "principal variation" is shallow, so to speak 2010-10-04T14:32:49 conventry: I'd say that is "deep" thinking 2010-10-04T14:33:18 (that's kinda what i'm counting on) 2010-10-04T14:33:45 a1k0n: what does it mean for the principal variation to be shallow? 2010-10-04T14:34:01 McLeopold: thanks for the example. I'll think about that. (By "deep" I meant "long" rather than "profound.") 2010-10-04T14:34:01 coventry: i used to hammer one planet but found it's better to attack multiple planets at the same time. even some high level bots will not defend very effectively and send ships on unnecessarily long routes sometimes. 2010-10-04T14:34:07 that although it's deep we cannot compute it? 2010-10-04T14:34:11 mega1: i mean the line of play in which both players are making the best move is only a few moves deep at any point 2010-10-04T14:34:30 conventry: by "deep" I meant he looks many moves ahead 2010-10-04T14:34:41 Wow, you think d 2010-10-04T14:34:42 i guess that doesn't really make sense 2010-10-04T14:35:01 McLeopold: Wow, you think dmj111 searches a long way ahead? 2010-10-04T14:35:02 the deterministic part may be shallow 2010-10-04T14:35:15 that's the thing; it heavily depends on what your opponent does 2010-10-04T14:35:23 but the mixed strategy part should be as long as the game 2010-10-04T14:35:55 coventry: he at least looks far enough to calc for incoming fleets. 2010-10-04T14:36:08 McLeopold: Yep. 2010-10-04T14:36:39 Thanks for the discussion, everyone. 2010-10-04T14:36:56 that doesn' necessarily mean that dmj simulates the game that deeply 2010-10-04T14:37:32 it may be that he has a heuristic or some simple graph algorithm that speculates on the number of ships at various places later in the game without actually simulating much 2010-10-04T14:39:10 it isn't that hard to simulate, and it doesn't take all that much memory. 2010-10-04T14:39:45 anyway i'm going to shut my trap until this bot stops failing assertions 2010-10-04T14:40:24 it depends on if you are branching the simulation 2010-10-04T14:40:32 *** Itkovian has joined #aichallenge 2010-10-04T14:41:01 i suspect that we don't really have any very sophisticated bots playing yet though 2010-10-04T14:41:18 *** Florian has joined #aichallenge 2010-10-04T14:41:31 *** coventry has quit IRC (Ping timeout: 272 seconds) 2010-10-04T14:42:13 *** Naktibalda has joined #aichallenge 2010-10-04T14:42:14 new bot taking 500ms per turn sometimes even with caching everything possible.. might actually time out on the contest server :( 2010-10-04T14:42:35 *** bhasker has joined #aichallenge 2010-10-04T14:43:06 orly. are you searching for something in that time? 2010-10-04T14:43:53 a1k0n, yes experimenting with branching 2010-10-04T14:44:49 a1k0n, but it's losing to my latest "production" bot so there must be a bug.. 2010-10-04T14:45:09 *** Florian has quit IRC (Disconnected by services) 2010-10-04T14:45:17 *** Florian has joined #aichallenge 2010-10-04T14:45:27 *** bhasker_ has quit IRC (Ping timeout: 264 seconds) 2010-10-04T14:45:28 also using python so slower than c++ for sure :( 2010-10-04T14:45:39 rebelxt, how many turns deep are you searching? 2010-10-04T14:45:44 *** Florian has quit IRC (Disconnected by services) 2010-10-04T14:45:46 I will confess that mine is not sophisticated. 2010-10-04T14:46:00 Purely reactionary. 2010-10-04T14:46:04 *** Florian95 has joined #aichallenge 2010-10-04T14:46:41 yeah i assumed from the start i'd want to do some game tree searching of some kind so write it up in C++ 2010-10-04T14:47:04 a1k0n: did you get past your assertions? 2010-10-04T14:47:15 Assertion failed: (planets[src].frames[timestep].owner == 1+player), function undo_move, file MyBot.cc, line 223. 2010-10-04T14:47:37 if i can get away with using GHC HEAD (meaning binary submissions) i can get speeds on par with C. if not, i will probably have to rewrite in C :\ 2010-10-04T14:47:42 hmm that might mean my implementation is forever screwed 2010-10-04T14:48:05 Migi32, i'm not even simulating enemy's future actions. just trying to figure out best possible action plan for each turn is getting computationally intensive (a lot of what-if simulations) 2010-10-04T14:48:44 oh ok, I just asked cause you said "branching" 2010-10-04T14:48:55 and assumed some kind of tree search 2010-10-04T14:49:21 Migi32, yeah sorry about confusion - not game tree search 2010-10-04T14:49:32 no worries :D 2010-10-04T14:50:46 has anyone actually tried search trees? 2010-10-04T14:51:06 looks like a1k0n is, based on that assertion failure he pasted 2010-10-04T14:51:33 i'm doing a search algorithm 2010-10-04T14:51:37 I've got an undo, but it's more for juggling order around finding the best use of resources. 2010-10-04T14:51:53 that still sounds like a (shallow) search tree to me 2010-10-04T14:51:58 *** fawek has joined #aichallenge 2010-10-04T14:52:09 jmcarthur: if it's not too much to ask, how do you decide what moves to consider 2010-10-04T14:52:30 It seems like there would be far too many at every turn to even go 1 ply 2010-10-04T14:52:30 in the context of my bot, that's a tough question to answer directly 2010-10-04T14:53:05 Even if you eliminated partial fleets into neutrals. It's a lot. 2010-10-04T14:53:20 my search is partially stochastic 2010-10-04T14:53:50 Can you beat Rage with it? 2010-10-04T14:54:04 it's not done yet, but i'm pretty sure that it will easily beat rage 2010-10-04T14:54:13 it's taking a long time to code :( 2010-10-04T14:54:13 jmcarthur: is your bot... running at all yet? 2010-10-04T14:54:17 no 2010-10-04T14:54:33 but i thought haskell made everything faster to develop 2010-10-04T14:54:41 *** bhasker_ has joined #aichallenge 2010-10-04T14:54:42 i have actually not been spending as much time on it as it may appear 2010-10-04T14:54:58 i also spent a long time waiting to see if binaries would be supported 2010-10-04T14:54:59 if you spent half the time working on it as you do talking about it it'd probably be in 1st place by now :) 2010-10-04T14:55:13 then i decided to suck it up and start coding with the assumption 2010-10-04T14:55:19 that was actually not long ago 2010-10-04T14:56:10 grr. i don't want to look like one of those annoying "all talk" people 2010-10-04T14:56:22 haha, well, i kinda feel like that myself now 2010-10-04T14:56:39 *** aerique has joined #aichallenge 2010-10-04T14:56:41 *** bhasker has quit IRC (Ping timeout: 245 seconds) 2010-10-04T14:57:52 i actually only have 307 loc right now. that's less than i thought. and even that, a lot of that code is throwaway code 2010-10-04T14:59:10 a1k0n: curious how many hours did you spend working on tron approximately? just trying to see if it was a number which would not overflow 16 bit int ;-) 2010-10-04T14:59:10 Top 10 players: davidjliu(3818), dmj111(3751), cfaftw(3703), felixcoto(3682), rsergio(3665), ruilov(3659), EBraun(3654), sequoh(3644), DocBot(3613), youwillfail(3611) 2010-10-04T14:59:37 i'm getting now lots of 'invalid order' starts 2010-10-04T14:59:48 is it really me, or is server buggy right now 2010-10-04T15:00:13 i think the server should be pretty good about invalid orders, so that one is probably you 2010-10-04T15:00:23 Right now I calculate the planet's scores first and calculate the attack/defend fleet requirements afterwards. I wonder if I should feed back the estimated ressource need into the scoring function until it ... stabilizes ... or so... dunno. 2010-10-04T15:00:29 it's the timeouts that i blame the server on, mainly 2010-10-04T15:00:50 rebelxt: i wasn't really keeping track, but it was just a couple hours a night for three weeks 2010-10-04T15:01:43 rabidus: not getting any invalid orders today. try to replay games locally and check (using dmj's python script or a javascript bookmarklet) 2010-10-04T15:03:29 hmm now i'm projecting i will have -127 ships on my planet. that seems unlikely 2010-10-04T15:03:44 a1k0n: that's good to hear. like you said in your post mortem, it's easy to get behind on "life" a bit when obsessed by challenges like this ;-) 2010-10-04T15:05:22 a1k0n: not if you're playing me. :D 2010-10-04T15:05:28 ho ho ho 2010-10-04T15:10:56 *** Zeiris_ has joined #aichallenge 2010-10-04T15:12:17 *** bhasker_ has quit IRC (Quit: bhasker_) 2010-10-04T15:15:21 *** krokokrusa has quit IRC (Ping timeout: 265 seconds) 2010-10-04T15:15:38 * Frontier plays Metallica - Sad but true 2010-10-04T15:16:23 *** Frontier has left #aichallenge 2010-10-04T15:17:15 should be playing peter gabriel - games without Frontier 2010-10-04T15:17:31 *** Frontier has joined #aichallenge 2010-10-04T15:17:48 then we wouldn't have elite :) 2010-10-04T15:19:06 Yesterday I was in a bot work session with a friend, and we were doing some speculative math about the game and the question "What is the integral of fear over time?" was posed. 2010-10-04T15:20:05 uh, just fear? 2010-10-04T15:20:05 I think we're on the right track. 2010-10-04T15:21:42 *** optimum has quit IRC (Ping timeout: 245 seconds) 2010-10-04T15:22:24 isn't the fear function closely related to the growth rate? 2010-10-04T15:24:43 *** Bobng has joined #aichallenge 2010-10-04T15:25:11 fright-seconds? 2010-10-04T15:25:35 *** bhasker has joined #aichallenge 2010-10-04T15:30:39 *** JensTi has joined #aichallenge 2010-10-04T15:33:30 *** MyHuggingBear has joined #aichallenge 2010-10-04T15:37:21 *** cfaftw has joined #aichallenge 2010-10-04T15:37:27 *** optimum has joined #aichallenge 2010-10-04T15:39:40 *** endunio has quit IRC (Quit: Page closed) 2010-10-04T15:40:32 hello 2010-10-04T15:40:53 anyone here? 2010-10-04T15:41:11 cfaftw: no 2010-10-04T15:41:15 no one is here 2010-10-04T15:41:24 that's a shame :( 2010-10-04T15:41:59 well in truth a lot do lurk 2010-10-04T15:42:16 but just lurk for an answer... 2010-10-04T15:42:25 yeah i sat here for a minute and noticed no chatter 2010-10-04T15:42:33 i just came to talk about this game 2010-10-04T15:42:57 for some reason i have a hunch that there should be a "best" move regardless of what your opponent does during that same turn 2010-10-04T15:43:11 meaning, it seems like there should be a provably optimal bot 2010-10-04T15:43:16 i'm just not smart enough to figure it out 2010-10-04T15:43:50 i think it's possible to build a bot that will stalemate against itself, but win against a bot that does anything differently (given enough turns) 2010-10-04T15:44:00 but i could be wrong 2010-10-04T15:44:50 cfaftw: since there is perfect knowledge -there is a "nash" equilibrium... or a best move i guess... however i belive its perhaps intractable to find it... like chess or go 2010-10-04T15:45:20 yeah i noticed it's a zero-sum game 2010-10-04T15:45:36 I think competative.... 2010-10-04T15:45:55 but i haven't done game theory for a year or two 2010-10-04T15:46:15 my knowledge of game theory is limited to wikipedia perusings :) 2010-10-04T15:46:37 ie your are maxamizing your score... but the total score does not sum to a constant 2010-10-04T15:46:56 ahhh 2010-10-04T15:46:56 ok 2010-10-04T15:47:06 *** sinan has joined #aichallenge 2010-10-04T15:47:07 true 2010-10-04T15:47:24 however, once all planets are possesed by either player, it becomes a zero-sum game, correct? 2010-10-04T15:48:13 no--you have a turn limit..... 2010-10-04T15:48:16 and income 2010-10-04T15:48:19 ships matter 2010-10-04T15:48:20 it's not strictly a zero-sum game unless you are talking about win or loss 2010-10-04T15:49:54 my understanding is that if the criteron that you win on... sums to a constant thrughout the game... its zero sum 2010-10-04T15:50:17 ok 2010-10-04T15:50:23 if you are just tring to maximize a score and the sum of score is not constant over the game --its competiative 2010-10-04T15:50:24 yeah it's not zero sum 2010-10-04T15:50:26 i was mistaken 2010-10-04T15:50:46 some moves are good for you--but not bad for the enemy 2010-10-04T15:51:14 while in zero sum games... any move that imporves your score... must decrease your enemys score 2010-10-04T15:52:06 since this has "elemination" of the enemy --then i can't remember.... 2010-10-04T15:54:35 i'm looking through game theory now trying to find what our modified galcon game would be considered 2010-10-04T15:55:12 *** edcba has quit IRC (Read error: Connection reset by peer) 2010-10-04T15:55:42 cfaftw: I believe some situations have 2 moves that are equally good, and depending on what you opponent does, you may win or lose. 2010-10-04T15:55:47 *** edcba has joined #aichallenge 2010-10-04T15:56:12 The classic rock, paper, scissors game. 2010-10-04T15:57:00 McLeopold: I think that you can always attain a draw unless you're in an inherently imbalanced situation. 2010-10-04T15:57:14 a perfect bot would avoid those situations, of course ;-) 2010-10-04T15:57:18 *** delt0r_ has quit IRC (Remote host closed the connection) 2010-10-04T15:57:54 *** Titankiller has joined #aichallenge 2010-10-04T15:57:59 Imaging 2 staring planets of distance 20, with a 99 growth 5 neutral 20 from each starting planet. 2010-10-04T15:58:04 *starting 2010-10-04T15:58:09 *** delt0r_ has joined #aichallenge 2010-10-04T15:58:17 *** Florian95 has quit IRC (Quit: Leaving...) 2010-10-04T15:58:21 client crashed 2010-10-04T15:58:37 Hmm... make that 19 distance all around. 2010-10-04T15:58:55 I think there is an optimal strat... its just intractable to find it 2010-10-04T15:59:10 Top 10 players: davidjliu(3814), dmj111(3739), cfaftw(3704), ruilov(3690), felixcoto(3680), rsergio(3657), EBraun(3647), sequoh(3636), DocBot(3609), youwillfail(3602) 2010-10-04T15:59:12 McLeopold: can you enter a text string that roughly approximates what you're describing? ( A ) ---- ( N) ----- ( B ) or something with details? 2010-10-04T15:59:20 so optimal is a bit moot 2010-10-04T15:59:27 *** oPless has quit IRC (Quit: This computer has gone to sleep) 2010-10-04T15:59:51 optimal is never moot 2010-10-04T15:59:56 ok... 2010-10-04T16:00:02 What is a "99 growth 5 neutral 20" ? 2010-10-04T16:00:03 know what optimal is, and then build what you can 2010-10-04T16:00:41 99 neutral ships, growth 5, 19 distance from other 2 planets. 2010-10-04T16:01:31 dwchandler: working all day on np-hard and np-complete problems leaves me with a different opinion 2010-10-04T16:01:34 Okay, so they form almost an equilateral triangle. 2010-10-04T16:01:35 The scenario needs tweaking, but I think it could be made to be a 50% 50% chance of winning. 2010-10-04T16:01:38 ;) 2010-10-04T16:01:58 If the system is in balance than a draw is the result of perfect play on both sides. 2010-10-04T16:02:59 mcleopold: the only winning move is not to play... until the opponent plays 2010-10-04T16:03:20 I agree cfaftw. The neutral planet is a poison pill. 2010-10-04T16:03:21 if there's only one neutral planet equidistant from you and your opponent 2010-10-04T16:03:26 *** wh1teside has quit IRC (Ping timeout: 240 seconds) 2010-10-04T16:03:28 cfaftw: yes 2010-10-04T16:03:51 Unless the population of the planet is < growth_rate * travel_time between the starting planets. 2010-10-04T16:03:56 and the opponents are the same distance, meaning no time to react 2010-10-04T16:04:01 In which case both planets should instantly attack it. 2010-10-04T16:06:13 But if you opponent instantly attacks you, your gone. 2010-10-04T16:06:24 someone was saying that he could formulate the optimal strategy in terms of a linear program... that would tend to indicate that it can be found in polynomial time 2010-10-04T16:06:56 which someone? 2010-10-04T16:07:24 *** delt0r___ has joined #aichallenge 2010-10-04T16:07:47 delt0r: point taken 2010-10-04T16:07:55 Migi32 2010-10-04T16:08:00 hi 2010-10-04T16:08:03 i'll keep my eye on him/her/it 2010-10-04T16:08:05 * Migi32 wakes up 2010-10-04T16:08:13 McLeopold: You can spare ships equal to your growth_rate * your distance from the enemy. 2010-10-04T16:08:18 it's foolish to assume a computer hasn't entered itself in the contest 2010-10-04T16:08:39 justin_pdx: yes, but in the scenario, in order to take the neutral, you have to leave yourself open 2010-10-04T16:08:43 *** delt0r_ has quit IRC (Ping timeout: 240 seconds) 2010-10-04T16:08:55 mcleopold: you said the starting planets are 20 apart 2010-10-04T16:08:59 19 2010-10-04T16:09:06 I messed up 2010-10-04T16:09:11 optimum, no i wasn't. I was suggesting a minimax search in which the value of each node would be defined by a linear program 2010-10-04T16:09:11 mcleopold, oh i missed that 2010-10-04T16:09:21 For example in this two planet system A can never capture B: ( A:100 ships, 5 growth ) <--- 10 travel time ---> ( B: 50 ships, 5 growth ) 2010-10-04T16:09:46 A will always have more ships, but it can never capture B. 2010-10-04T16:09:46 *** Palmik has quit IRC (Remote host closed the connection) 2010-10-04T16:10:40 justin_pdx, wrong. due to gravity those 2 planets will attract each other, making the travel time shorter :P 2010-10-04T16:10:58 Migi32: they are a binary system. :) 2010-10-04T16:10:59 eventually :P 2010-10-04T16:11:00 lol, moving planets? 2010-10-04T16:11:20 that would make things difficult 2010-10-04T16:11:22 ahh, but the universe ends (turn limit) before gravity can show its effects 2010-10-04T16:11:39 that would be hilarious 2010-10-04T16:11:40 Actually space is expanding, meaning that the speed of light is reducing and travel time is increasing. 2010-10-04T16:11:44 *** tapwater has joined #aichallenge 2010-10-04T16:14:06 did anyone notice Hazard's resubmission? 2010-10-04T16:14:13 he was king for a while, then resubmitted 2010-10-04T16:14:16 now he's hovering around 20 2010-10-04T16:14:31 he'll be back 2010-10-04T16:14:32 *** felixcoto has quit IRC (Ping timeout: 265 seconds) 2010-10-04T16:14:44 when i saw his resubmission i was afraid that he had made something incredible, something unbeatable 2010-10-04T16:15:19 migi32: i'm positive you're right 2010-10-04T16:15:45 hey guys, i wrote up some basic/intermediate level strategy advice at http://ai-contest.com/forum/viewtopic.php?f=17&t=856 - everyone in TOP 25 already seems to have already implemented most of the strategies covered, so it's probably only useful for people who are trying to get into TOP 50 or so 2010-10-04T16:15:56 hopefully someone finds it useful - feedback is very welcome!.. 2010-10-04T16:16:11 rebelxt, what if i read that and haven't implemented any of those strats? ;) 2010-10-04T16:16:46 *** wh1teside has joined #aichallenge 2010-10-04T16:16:47 cfaftw, haha, that would make you a genius (in my opinion) 2010-10-04T16:16:51 i took randombot... and made it REALLY random 2010-10-04T16:17:03 hi 2010-10-04T16:17:38 hello 2010-10-04T16:17:49 cfaftw, oh i thought you meant a ~TOP50 bot not using any of the strats i wrote. i'd really love to see such a thing ;-) 2010-10-04T16:18:29 rebelxt: i did mean that, but then imagined winning with a ridiculous bot 2010-10-04T16:19:10 cfaftw, it would be incredible for somebody to win the contest then release the source code as being just a few lines of simple code 2010-10-04T16:20:14 why did i just talk to myself 2010-10-04T16:20:25 lol! 2010-10-04T16:20:35 fail 2010-10-04T16:20:49 i expect winning submission will be over 1K LOC 2010-10-04T16:20:55 maybe over 2K.. 2010-10-04T16:21:07 *** Accoun has quit IRC () 2010-10-04T16:21:16 anyone else here have sucky code? so much of my bot is comprised of "ooo, i think this might work..." 2010-10-04T16:21:40 *** Naktibalda has quit IRC (Quit: No Ping reply in 180 seconds.) 2010-10-04T16:21:54 I think that the winning bot will actually be < 1000 LOC. 2010-10-04T16:22:03 cfaftw, you are not alone! i have a few lines with a comment "why the hell does this heuristic make my bot better?!" 2010-10-04T16:22:06 *** Naktibalda has joined #aichallenge 2010-10-04T16:22:38 I think that human concepts, heuristics and approximations increase the LOC count. 2010-10-04T16:22:41 yeah I not only flag my code with // TODO but also // WTF 2010-10-04T16:23:06 The more assumptions and restrictions you try to explain to your bot the more conditionals and naturally LOC you have. 2010-10-04T16:23:15 /WTF is the best comment ever 2010-10-04T16:23:32 But I think the winning bots will have little more than iterators and state analysis algorithms. 2010-10-04T16:24:01 justin_pdx, I believe the winning bot will use a mixture of multiple strategies, so I disagre 2010-10-04T16:24:03 e 2010-10-04T16:24:06 justin_pdx: seems like utilities code is easily in 500-1000 LOC range already 2010-10-04T16:24:20 say you code for neutral sniping, then someone goes meta-neutral sniping, so you go meta-meta-neutral sniping 2010-10-04T16:24:38 Doesn't exist. 2010-10-04T16:25:00 Migi32: i think you are right.. thinking of netflix prize winners who blended 100+ algorithms (!) to win the grand prize 2010-10-04T16:25:14 netflix? 2010-10-04T16:25:18 going meta meta with all the conditionals etc, probably would end up losing to a simpler bot 2010-10-04T16:25:36 like the prisoner dilemma winning bot 2010-10-04T16:25:58 rebelxt: I bet the utilities libraries are useful, but probably are used for deriving not necessarily the most useful information. 2010-10-04T16:26:05 Migi32: yeah 1 mil $ netflix prize contest (predicting movie ratings using collaborative filtering) 2010-10-04T16:26:33 Good for debugging, analysis, etc. But a core math bot can probably ignore most approximations and handle the data raw. 2010-10-04T16:26:53 woah rebelxt I totally missed that 2010-10-04T16:27:33 justin_pdx: well, agreed, i ended up no longer using some of the util code written for early versions of the bot... 2010-10-04T16:28:31 I think it's important to have really good fundamentals / utility functions so you reduce the number of lines of code of your actual bot logic 2010-10-04T16:29:04 Migi32: it was a lot of fun, but got out of hand - never ending algorithm arms race which lasted for 2+ years. blending 100+ algorithms to win the contest was insane... 2010-10-04T16:30:24 what do you expect? $100000 dollars! 2010-10-04T16:31:11 yep $ 1 000 000 2010-10-04T16:31:43 Is this prize in this current contest just bragging rights? 2010-10-04T16:32:12 currently and in the forseeable future yes 2010-10-04T16:32:46 my goal is to finish in the top 100. 2010-10-04T16:33:06 *** Accoun has joined #aichallenge 2010-10-04T16:33:14 ONE MILLION DOLLARS?!?!? 2010-10-04T16:33:19 I respect all these other guys too much to think I could come out on top. 2010-10-04T16:33:20 jk 2010-10-04T16:33:27 i do it because i'm addicted 2010-10-04T16:33:41 as soon as my bot got its first win, i couldn't stop 2010-10-04T16:33:42 Yeah, it's a personal challenge. 2010-10-04T16:34:44 did any of you also join this contest thinking "pfff, this game is simple. I can find a perfect solution, no problem." 2010-10-04T16:35:08 * foucist raises hand 2010-10-04T16:35:09 migi32, haha, and i still think that 2010-10-04T16:35:16 weeks later 2010-10-04T16:35:26 cfaftw, haha me too actually :P 2010-10-04T16:35:27 it's always "I'm SO close!!!" 2010-10-04T16:35:38 "I almost got it" 2010-10-04T16:35:58 Migi32: haha great question.. having played Galcon for 2 years (30K games played total against human players), I did know exactly what I was getting into!.. no regrets though! 2010-10-04T16:36:29 no chance to win, still fun to participate 2010-10-04T16:36:36 honestly, though... when i entered i did not expect to get to the rank i'm at right now 2010-10-04T16:37:30 cfaftw: you are doing great! 2010-10-04T16:37:43 rebelxt, there's definitely a chance to win 2010-10-04T16:38:23 if I get "error while compiling submission", will I get an email with the error output? 2010-10-04T16:38:30 well, maybe for you ;-) i seem to be always 1-2 steps behind you top10 guys 2010-10-04T16:38:44 because it compiles fine here 2010-10-04T16:38:51 even the best bots fail on certain maps... i'm still waiting for The One Bot to appear that will have a perfect record 2010-10-04T16:39:13 wvdschel: 20-30% chance to get error e-mail. it's still a bit buggy. what language? 2010-10-04T16:39:14 wvdschel, you use C#? 2010-10-04T16:39:14 rebelxt, you could get lucky and happen to only play on maps against opponents such that the maps are the opponents' weakest maps 2010-10-04T16:39:27 rebelxt, mietek: Lisp 2010-10-04T16:39:48 didn't know it was possible to use Lisp :P 2010-10-04T16:39:56 they have a bunch of language packs now 2010-10-04T16:40:03 it is: http://ai-contest.com/language_profile.php?lang=Lisp 2010-10-04T16:40:09 anyone make a brainf*ck language pack yet? 2010-10-04T16:40:39 haha :D 2010-10-04T16:40:44 wvdschel: sorry - can't help with Lisp :( if there is no e-mail, trey re-submitting again, you may get it upon next compile failure 2010-10-04T16:41:26 actually thinkig about it, bf-like language could be ideal for representing the concepts in a topographical way to make the winning bot 2010-10-04T16:41:26 I bet brainf*ck would time out before it could calculate 1 planet-planet distance :P 2010-10-04T16:41:39 Piet! 2010-10-04T16:42:01 can i see some code in brainf*ck?... 2010-10-04T16:42:17 http://en.wikipedia.org/wiki/Brainfuck 2010-10-04T16:42:31 ++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>. 2010-10-04T16:42:36 that was hello world 2010-10-04T16:42:40 just remember... some things you can't unsee 2010-10-04T16:42:48 yeah 2010-10-04T16:42:52 warning: major timesink 2010-10-04T16:43:19 migi32: that's the warning they should put on the ai-contest page 2010-10-04T16:43:24 br0nfuk 2010-10-04T16:43:46 ;-) 2010-10-04T16:44:44 wvdschel: hurp 2010-10-04T16:45:07 hurp? 2010-10-04T16:45:24 durp? 2010-10-04T16:46:27 now that we're offtopic anyway, does anyone here play starcraft 2? 2010-10-04T16:46:43 wait, it crashes on errors? 2010-10-04T16:46:50 erm 2010-10-04T16:46:53 warnings 2010-10-04T16:47:00 sorry, errors would have made sense :) 2010-10-04T16:47:09 ; caught STYLE-WARNING: 2010-10-04T16:47:09 ; The variable UNIVERSE is defined but never used. 2010-10-04T16:47:43 strange 2010-10-04T16:47:53 might be Lisp-specific 2010-10-04T16:48:22 i have starcraft 2 but haven't had much time to play it yet 2010-10-04T16:48:40 i just started coding my bot =) 2010-10-04T16:48:48 in reference to earlier conversation about optimal play, i'm going to stick with my assertion that two optimal bots will usually not draw 2010-10-04T16:49:46 optimum, wait you're that #ai regular that joined here because I asked a question there yesterday, right? If so, we're all doomed! 2010-10-04T16:50:49 jmcarthur, can you remind us the reason of that assertion? 2010-10-04T16:50:51 no way, optimium mentioned ai-contest in #ai so i joined! 2010-10-04T16:51:10 oh ok 2010-10-04T16:51:21 Migi32: why talk about sc2 when you could talk about minecraft? irc is buzzing with minecraft :P 2010-10-04T16:51:26 not that i've played it yet sadly 2010-10-04T16:51:32 however, my difficulty in coming up with a simple enough scenario to demonstrate that might be a sign that i am incorrect 2010-10-04T16:51:40 i bought it yesterday 2010-10-04T16:51:44 just as a warning 2010-10-04T16:51:50 DON'T BUY MINECRAFT 2010-10-04T16:51:53 a warning to whom 2010-10-04T16:51:55 wvdschel: too late :( 2010-10-04T16:51:57 foucist, ah yes Minecraft. Perhaps the biggest timesink in the entire internet :P 2010-10-04T16:51:59 it's cybercrack 2010-10-04T16:52:04 how about dwarf fortress? 2010-10-04T16:52:06 anyway, I'm off 2010-10-04T16:52:06 minecraft has destroyed me 2010-10-04T16:52:17 not going to play minecraft! 2010-10-04T16:52:19 irc was buzzing about dwarrf fortress mere days before minecraft took over 2010-10-04T16:52:28 I've been clean for a whole 4 hours now 2010-10-04T16:52:40 *** wvdschel has quit IRC (Quit: Leaving) 2010-10-04T16:52:49 jmcarthur, what makes you think two optimal bots will usually not draw? 2010-10-04T16:53:01 are most of you using your bot names as your nicks? 2010-10-04T16:53:18 i'm still working on whittling my thoughts down to a simple scenario 2010-10-04T16:53:28 cfaftw: i will answer you once i've done that ^^ 2010-10-04T16:54:00 Minecraft is just so evil. You just have some fun doing random stuff on your first day, quickly build a shelter for your first night, you like your shelter and BAM, you're hooked. From that point on your shelter will only get bigger and bigger. 2010-10-04T16:54:03 jmcarthur, oh... i was just looking for you to explain your hunch 2010-10-04T16:54:14 cfaftw: i'd rather just prove it ;) 2010-10-04T16:54:21 jmcarthur, good call, haha 2010-10-04T16:54:25 i've explained it before and failed to convince many people 2010-10-04T16:59:12 Top 10 players: davidjliu(3805), dmj111(3739), cfaftw(3701), ruilov(3691), felixcoto(3675), rsergio(3659), EBraun(3642), sequoh(3629), DocBot(3615), youwillfail(3598) 2010-10-04T16:59:35 cfaftw, there are some rock-paper-scissor mechanics in the game. Very subtle, but they're there, so draw between optimal bots is unlikely 2010-10-04T16:59:54 jmcarthur: have you ever played diplomacy? 2010-10-04T16:59:56 i'm still having trouble finding a RPS scenario though 2010-10-04T17:00:02 McLeopold: no 2010-10-04T17:00:21 hmm.. think risk, but better 2010-10-04T17:00:30 I've got a good article for you... 2010-10-04T17:01:07 migi32, gotcha. 2010-10-04T17:02:28 jmcarthur, RageBot made me see: RageBot < expanding with such a number of ships that you can hold against RageBot < expanding with all ships < RageBot 2010-10-04T17:03:46 jmcarthur: http://www.diplom.org/Zine/F1995R/Dreier/Endgame2.html 2010-10-04T17:03:51 that's the second part 2010-10-04T17:03:58 I'm looking for the first 2010-10-04T17:04:57 jmcarthur: http://www.diplom.org/Zine/F1995M/Dreier/Endgame.html 2010-10-04T17:05:03 i'll check it out 2010-10-04T17:05:24 jmcarthur: okay, the summary is, sometimes you look even, but your win prob is about 1 2010-10-04T17:05:43 I'm starting to think planet wars will be this way 2010-10-04T17:06:01 you mean the probability that it's not a draw? 2010-10-04T17:06:06 Yes 2010-10-04T17:06:09 I played Diplomacy once, as Austria, I was dead on the fifth or so turn XD 2010-10-04T17:06:14 yeah, that is also my suspicion 2010-10-04T17:06:18 Migi32: seeing how i studied political science in college, i'm sure i won't win the contest 2010-10-04T17:06:35 but i still can't seem to prove it! 2010-10-04T17:06:55 Migi32: austria's hard, you usually have to ally with italy to have a chance 2010-10-04T17:07:23 well my genious move was to backstab italy on the first move :P 2010-10-04T17:07:30 didn't turn out great :P 2010-10-04T17:07:52 Migi32: no, backstabbing can't happen until you know they will die, and only after the neutral are gone. 2010-10-04T17:08:41 hmm, it was a long time ago, I don't remember much of it 2010-10-04T17:08:53 you seem to play diplomacy quite a lot, McLeopold 2010-10-04T17:09:02 I have. I'm quite good. 2010-10-04T17:09:31 alright, it's quittin time! meaning time to go home and work on my bot at some point. 2010-10-04T17:09:36 Planet Wars would be similar, if it was multi-player. 2010-10-04T17:09:42 later 2010-10-04T17:10:03 *** cfaftw has quit IRC (Quit: Leaving) 2010-10-04T17:10:36 It would be interesting and fun if you could signal an opponent to help you defeat a third opponent. 2010-10-04T17:10:43 McLeopold, GalCon is multi-play PlanetWars. It's in fact also a kind of diplomacy, but without typing/discussing, just provoking and (not) reacting 2010-10-04T17:10:49 Then you'd have to worry about the backstab. 2010-10-04T17:11:31 I've always preferred no-press diplomacy. 2010-10-04T17:15:26 god this game is so complicated 2010-10-04T17:16:10 i'm trying to map out some moves for just three planets with only 1 ship on the starting planets, no neutral ships, and low growth rates 2010-10-04T17:16:10 planet wars? 2010-10-04T17:16:27 what's your map? 2010-10-04T17:16:32 complex to analyze, i mean 2010-10-04T17:16:48 Hello, approximately after how many games, my bot would reach it's fair ELO Score ? 2010-10-04T17:17:13 Starting planets 2 turns from each other with 1 ship and 1 growth. A neutral planet 2 turns from both starting planets with no ships and 2 growth. 2010-10-04T17:17:18 nullkuhl: I think about 10-15 2010-10-04T17:17:21 nullkuhl: figure somewhere around 20 usually 2010-10-04T17:17:53 nullkuhl: that would be easier to say with certainty if the server would just show confidence intervals :( 2010-10-04T17:19:22 jmcarthur: I think I'd go for the neutral and abandon my home. 2010-10-04T17:19:56 jmcarthur: are you doing a game tree? 2010-10-04T17:21:13 jmcarthur: it seems like a game of chicken, first person to stop going for the neutral losses 2010-10-04T17:21:37 McLeopold: sorry to interrupt but what do you mean by a game tree here... 2010-10-04T17:22:10 nulkuhl: at the opening, each player has 3 moves, hold, neutral or enemy. so 9 opening combos 2010-10-04T17:22:24 nullhuhl: the next turn is more complicated 2010-10-04T17:23:17 McLeopold: is your main server bot using a tree or still heuristic based? 2010-10-04T17:23:34 i've started the construction of the code for a tree, but still working out the finer details with a comb 2010-10-04T17:23:37 no trees, pure calculation 2010-10-04T17:23:42 McLeopold: i also think it looks like a game of chicken. it was originally a more complicated scenario. every time i distill it down to look for a game requiring a mixed strategy (not resulting in a draw every time), i end up with the game just being a draw 2010-10-04T17:24:15 i am certain intuitively that there exists a simple scenario that will not end up with a draw with optimal play on part of both players, but i can't find it :\ 2010-10-04T17:24:29 jmcarthur: i agree with that 2010-10-04T17:24:30 I'm looking for it too 2010-10-04T17:25:22 what about 1 ship on the neutral? 2010-10-04T17:26:05 that would probably mean wait 1 turn, then go for neutral. 2010-10-04T17:26:08 antimatroid2: main server ? is there a secondary server that people try their bots on ? 2010-10-04T17:26:09 i'm guessing that would end up being both wait for a turn and then keep going for neutral 2010-10-04T17:26:10 still draw... 2010-10-04T17:26:14 haha 2010-10-04T17:26:30 we think alike 2010-10-04T17:26:32 the problem is, if you're using a symmetric map, then by symmetry a nonlosing strategy for one player has to also be a nonlosing strategy for the other player 2010-10-04T17:27:18 this is assuming deterministic play, of course 2010-10-04T17:27:24 Zannick: but if it's a mixed strategy then... yeah that :) 2010-10-04T17:27:49 optimal does not necessarily imply nonlosing, is what i'm attempting to demonstrate 2010-10-04T17:27:58 it will be on average even 2010-10-04T17:28:22 *** mega1 has quit IRC (Ping timeout: 245 seconds) 2010-10-04T17:28:38 jmcarthur, 1 ship on neutral, growth 1 2010-10-04T17:28:43 McLeopold: yes, i can see you tree , starting by a move where u attack an enemy, neutral or u hold, but what if you attack neutral and enemy at the same turn 2010-10-04T17:29:20 nullkuhl: all combinations of moves are considered at 1 ply 2010-10-04T17:30:16 yes so how does that make a tree, i mean for example in the first turn, you could hold, or attack 1 enemy or attack 1 neutral or attack n neutrals 2010-10-04T17:31:14 nullkuhl: the tree is made after you pick a move combination, then generate the next set of move combinations from that board 2010-10-04T17:32:16 nullkuhl: http://www.sifflez.org/misc/tronbot/ 2010-10-04T17:32:26 (by analogy) 2010-10-04T17:32:58 a1k0n: am familiar with minmax algorithm however i cant imagine how will planet wars map into a tree or so 2010-10-04T17:33:17 planet wars in general maps very badly into a tree 2010-10-04T17:33:19 McLeopold: ok ( the next set of move combinations ) will be the sub nodes of the current node i am at in the tree ? 2010-10-04T17:33:23 well, instead of moving in the four directions you send your n fleets to m places. it's a very, very, very large branching factor. 2010-10-04T17:33:30 yes 2010-10-04T17:33:37 but the tree will be HUGE 2010-10-04T17:33:44 in a real game 2010-10-04T17:33:59 well i think that each node then could have more than 1000000 sub node 2010-10-04T17:34:03 If I'm looking at it right the simplest r-p-s scenario is three planets 2 distance apart, starter planets have growth 2, ships 3 and neutral has growth 3 ships 2 2010-10-04T17:34:11 yep. you don't explicitly construct such a tree in memory. 2010-10-04T17:34:35 122 choose 22 is the number of possible sets of orders on the first turn, when there are 23 planets and you start with 100 ships 2010-10-04T17:34:48 Janzert: wow that is already more complex than i was hoping 2010-10-04T17:34:52 *** Titankiller has quit IRC (Ping timeout: 252 seconds) 2010-10-04T17:34:53 a1k0n: so what do i do then or how do i map the game into states 2010-10-04T17:34:56 9.4 x 10^23 2010-10-04T17:35:07 Janzert: I think that's a draw as well 2010-10-04T17:35:13 and yeah, your move depends heavily on what your opponent does, so if you implement standard minimax where player 2 gets to pick a move given player 1's move, so player 2 comes out ahead 2010-10-04T17:35:31 i haven't come up with a good way to resolve that in the standard minimax search 2010-10-04T17:35:37 nullkuhl: you give up. nobody else is doing it that way. :) 2010-10-04T17:35:38 Janzert: neutral is easily snipped with no threat to home planet 2010-10-04T17:35:58 *** Titankiller has joined #aichallenge 2010-10-04T17:36:05 but i have implemented it using a backtracking search. 2010-10-04T17:36:15 basically do a move, recurse, undo it 2010-10-04T17:36:23 aha 2010-10-04T17:36:36 a1kon: average of board positions only including your move? 2010-10-04T17:36:50 a1k0n is writing prolog in c++ 2010-10-04T17:36:50 McLeopold: you're right 2010-10-04T17:37:02 bleh, I give up for the moment then :) 2010-10-04T17:37:16 McLeopold: hmm? 2010-10-04T17:37:24 you want to know how many? 2010-10-04T17:37:41 okay, the 3 planets are all distance 1 2010-10-04T17:37:41 my best idea of even trying minimax/alphabeta is to determine what the worst thing my opponent could do right now is and determine what the best way to get around that is 2010-10-04T17:38:09 yeah you can probably at least find a countermove to every move without too much poking around 2010-10-04T17:38:11 maximin? 2010-10-04T17:38:40 i think i implemented that and it basically thought the enemy was invulnerable and just sat there 2010-10-04T17:38:42 Zannick: you have to pick the thing that doesn't lose, but leave you opponent with the biggest chance of making a mistake. 2010-10-04T17:38:59 McLeopold: heh 2010-10-04T17:38:59 (but i probably also had bugs) 2010-10-04T17:39:12 like tic tac toe 2010-10-04T17:39:19 *** jaspervdj has quit IRC (Quit: NEVER GONNA GIVE YOU UP NEVER GONNA LET YOU DOWN) 2010-10-04T17:39:30 my first attempt is going to be moderately heuristic based at first 2010-10-04T17:40:43 McLeopold: hmm. that's actually quite profound 2010-10-04T17:40:51 a1k0n: McLeopold: so what is the general approach anyway to approach this game, its my first step in AI, and so far is what i have been doing is some adhoc approaches,.. 2010-10-04T17:41:03 i've been trying to figure out the right way to implement that in general. 2010-10-04T17:42:00 nullkuhl: most people are doing for each of my planets i, for each other planet j, what happens if i try to send ships to j? and then repeatedly do that until they're out of ships to send from i 2010-10-04T17:42:00 a1k0n: it's why I think over time, all bots with tend to draw, with the better ones not making mistakes and winning over the course of more moves 2010-10-04T17:42:36 nullkuhl: I'd barely call what I'm going AI. 2010-10-04T17:42:40 *doing 2010-10-04T17:42:55 er, insert "pick the best of those options" in there 2010-10-04T17:43:01 I'm not learning at all, it's against the rules. :( 2010-10-04T17:43:07 *** davidjliu has joined #aichallenge 2010-10-04T17:43:25 well if you doing some minmax and having a game tree, i would call that AI since you are some how forming a smart agent/bot based on a search tree and some states 2010-10-04T17:43:37 McLeopold: you can train locally, then upload your current learnings 2010-10-04T17:44:04 *** aerique has quit IRC (Quit: ...) 2010-10-04T17:44:23 i am increasingly believing that my hypothetical scenario doesn't exist 2010-10-04T17:44:36 my intentions for a game tree aren't to search the whole things breadth wise 2010-10-04T17:44:36 McLeopold: a1k0n;: do you think i should read some AI book (some book discussing minmax or so) first that would help me find my way to a good solution ? 2010-10-04T17:44:38 jmcarthur: yeah, so do i 2010-10-04T17:44:45 but pick put possible reasonable moves from each player at each step 2010-10-04T17:44:48 and just see how that does 2010-10-04T17:44:59 the absense of such a scenario might have a deep meaning itself 2010-10-04T17:45:00 nullkuhl: no, write a stupid bot, then start improving it 2010-10-04T17:45:31 nullkuhl: make sure it can beat the previous version 2010-10-04T17:45:42 McLeopold: thats what am doing but the problem is , iam not having any kind of data structure for the game to map its states or a game tree like you said or so.. 2010-10-04T17:45:58 use wikipedia 2010-10-04T17:46:00 nullkuhl: I'm not using a tree. 2010-10-04T17:46:01 it has lots of information 2010-10-04T17:46:14 a tree doesn't have to actually be a data structure 2010-10-04T17:46:23 antimatroid2: what kind of topics should i read about, i kind of have no idea where to start from? :) 2010-10-04T17:46:24 a recursive algorithm is also a tree 2010-10-04T17:46:29 the whole tree thing is more of a hypothetical discussion rather than an a strategy anybody expects to work 2010-10-04T17:46:50 *** amjka has joined #aichallenge 2010-10-04T17:46:54 i expect it to do reasnably, if ingrained with decent enough heuristics 2010-10-04T17:47:05 under the assumption that everyone else is having as much trouble working out good moves 2010-10-04T17:47:06 jmcarthur: FINISH YOUR BOT 2010-10-04T17:47:16 we need to know if your on to something 2010-10-04T17:47:24 jmcarthur: yes my algorithm isnt recursive, i mean that i dont have some form of game state planning,or some sort of a backtracking, etc.. 2010-10-04T17:47:26 but i'm trying to solve this scenario in my head :( 2010-10-04T17:47:37 or rather come up with a scenario 2010-10-04T17:47:56 jmcarthur: the problem is, there is always time to defend 2010-10-04T17:48:30 are there any external websites that could act as useful tools to this contest ? ( like some bots to challenge, more maps etc.. ) 2010-10-04T17:48:54 nullkuhl: have you played on the tcp server? 2010-10-04T17:49:00 http://72.44.46.68/getplayer?player=McLeopold.WormBot.py 2010-10-04T17:49:27 watch a1k0n lose: http://72.44.46.68/canvas?game_id=308608 2010-10-04T17:49:39 :| 2010-10-04T17:49:51 :) 2010-10-04T17:50:15 if i could stop this thing from crashing i might have a response to that 2010-10-04T17:50:52 i'm pretty sure player 1051804 isn't going to own planet 4 2010-10-04T17:51:00 ? 2010-10-04T17:51:07 oh 2010-10-04T17:51:18 ...but just in case... 2010-10-04T17:51:20 ==78462== Conditional jump or move depends on uninitialised value(s) 2010-10-04T17:51:22 whoops 2010-10-04T17:52:40 a1k0n: just to make you feel better... 2010-10-04T17:52:43 AttributeError: 'NoneType' object has no attribute 'planets' 2010-10-04T17:52:44 *** Appleman1234 has joined #aichallenge 2010-10-04T17:52:57 McLeopold: tcp server seems quite useful, do i use the same account as that of ai-contest.com ? 2010-10-04T17:53:08 no, you make up any name 2010-10-04T17:53:17 I'd go with nullkuhl 2010-10-04T17:53:20 for instance i'm playing as McLeopold-WormBot 2010-10-04T17:53:29 heh 2010-10-04T17:53:31 ...and you name sounds German. 2010-10-04T17:53:47 it's the german version of zero cool from hackers 2010-10-04T17:53:48 >:| 2010-10-04T17:54:01 lol , a1k0n got it right 2010-10-04T17:54:19 wow, wild guess 2010-10-04T17:54:51 make your bot Verätzungen then 2010-10-04T17:55:21 nullkuhl: also, you can play multiple versions of you bot on tcp server 2010-10-04T17:55:38 McLeopold: is this documented anywhere ? 2010-10-04T17:55:39 vortex has 39. O_o 2010-10-04T17:55:47 http://www.benzedrine.cx/planetwars/ 2010-10-04T17:55:55 I hope you can compile c code 2010-10-04T17:56:08 not realy 2010-10-04T17:56:15 fc 2010-10-04T17:56:27 McLeopold: i can but i think someone provided a jar file already ? 2010-10-04T17:56:27 for this example you guys are looking for, have you considering 4 planets, 2 players starting a fair distance away, then each of the other two planets being cheap and valuable, 1 move away from each centre? 2010-10-04T17:56:34 so each player has to go for one of the middle planets 2010-10-04T17:56:38 maybe, i don't use it 2010-10-04T17:56:47 "Unpack TCP.jar, from the ZIP file just below, into the tools directory of your starter package." PlanetWars tester TCP edition 2010-10-04T17:56:48 i gues you bot then go for the other players planet? 2010-10-04T17:56:52 both* 2010-10-04T17:57:10 McLeopold: can i play against a bot of a user that i specify on the tcp server ? 2010-10-04T17:57:17 other players neutral planet* that is 2010-10-04T17:57:26 nullkuhl: it matches you up randomly with players ranked near you 2010-10-04T17:57:42 ...or whoever else is playing, no matter what their rank is 2010-10-04T17:58:20 wait, no that wouldn't work, you'd both wait? 2010-10-04T17:58:21 *** bhasker has quit IRC (Quit: bhasker) 2010-10-04T17:58:35 antimatroid2: no, you wouldn't wait 2010-10-04T17:58:35 a1k0n: ok, do i start the matches manually , or its automated like the ai-contest.com server 2010-10-04T17:58:36 unless the growth rate is high enough, then you'd have to attack 2010-10-04T17:58:51 McLeopold: i think it depends on growth 2010-10-04T17:59:10 nullkuhl: you run your bot when you want, from your own machine 2010-10-04T17:59:12 Top 10 players: davidjliu(3793), dmj111(3720), ruilov(3707), cfaftw(3688), felixcoto(3662), rsergio(3646), sequoh(3626), EBraun(3623), Hazard(3617), DocBot(3614) 2010-10-04T17:59:12 *** BtbN has quit IRC (Quit: Verlassend) 2010-10-04T17:59:20 antimatroid2: if they are closer to the enemy than your home planet, then no 2010-10-04T17:59:31 i'll draw some examples 2010-10-04T17:59:42 dwchandler: so how does it match with another player then ? i mean there has to be another player trying to run his bot as well at the same time, yeah ? 2010-10-04T17:59:56 nullkuhl: yep 2010-10-04T18:00:03 yes. fortunately that's almost always the case 2010-10-04T18:00:36 tcp connects and waits for a match up. only a second or three, usually 2010-10-04T18:00:41 44 connections, 21 games running currently 2010-10-04T18:02:04 nullkuhl: it's really easy. if you can compile C then just try it and see 2010-10-04T18:02:30 why does compiling C sound like compiling kernel today 2010-10-04T18:02:49 some people don't use C. ever. 2010-10-04T18:03:02 maybe that's you. maybe not 2010-10-04T18:03:16 *** fawek has quit IRC (Remote host closed the connection) 2010-10-04T18:03:23 isnt it always (Compiler|Interpreter) filename.ext 2010-10-04T18:04:02 tcp.c has instructions in comments at the top 2010-10-04T18:04:07 *** sigh has joined #aichallenge 2010-10-04T18:04:29 anyway, so the idea of reading more about games ai or game theory wont be as helpful as trial and error in this game ? 2010-10-04T18:04:37 *** otherAntimatroid has joined #aichallenge 2010-10-04T18:05:14 it's fine. just don't go complete your education before starting to write code 2010-10-04T18:05:42 and making a bot might give you some hints what areas of game theory to look into 2010-10-04T18:05:45 I'm pretty sure that AI and Game Theory knowledge will get you father than pure trial and error. 2010-10-04T18:05:47 I havn't been able to apply any of the "traditional" ai techniques to this problem 2010-10-04T18:05:58 but having a knowledge of them certainly helps 2010-10-04T18:06:08 After all the point is to learn how to solve a generalized set of problems, not just play Galcon. 2010-10-04T18:06:14 i think being a intuitionist superman would get you farthest 2010-10-04T18:06:31 an* 2010-10-04T18:06:32 okay, what would you guys do for these two situations? 2010-10-04T18:06:51 http://img28.imageshack.us/img28/9880/egmap.png 2010-10-04T18:06:55 sorry, imgur seems to be down 2010-10-04T18:07:07 I would say that game theory and discrete math would be better than AI studies. It's not a vague enough game for complex AI, imho. 2010-10-04T18:07:28 i'm a huge game theory fan and i think people will be struggling to use that as well 2010-10-04T18:08:02 justin_pdx: antimatroid2: do you know any good books explaining/discussing game thoery ? 2010-10-04T18:08:12 how much maths do you know? 2010-10-04T18:08:20 regardless, you need to become comfortable with the heuristics that work for this game, before you can apply more sophisticated ideas 2010-10-04T18:08:25 if not much, martin osbornes, if a lot then myerson or fudenberg 2010-10-04T18:08:35 Lots of information on the web, nullkuhl. Use wikipedia and its references as a good starting point. 2010-10-04T18:09:07 you would want to have done some analysis before picking up those later two books and expecting to understand them though 2010-10-04T18:09:53 McLeopold: any first attempts at optimal strategies in those two situations? 2010-10-04T18:09:55 i'm not positive 2010-10-04T18:10:38 antimatroid2: Can you explain these images? 2010-10-04T18:10:50 What are the greater distances? They're not labeled. 2010-10-04T18:10:56 there are 4 planets for each "map", each player starts on one of the outer planets, how do you play? 2010-10-04T18:10:56 *** amjka has quit IRC (Quit: Page closed) 2010-10-04T18:11:04 greater distances aren't important imo 2010-10-04T18:11:10 but say 4 if you must 2010-10-04T18:11:20 Are the two in the middle neutral? 2010-10-04T18:11:21 I would send all 100 units to the closest planet 2010-10-04T18:11:24 yes 2010-10-04T18:11:28 in both cases 2010-10-04T18:11:30 are all planets distance 4 2010-10-04T18:11:40 no, middle 2 are distanced 1 2010-10-04T18:11:41 The greater distances are very important, antimatroid2. 2010-10-04T18:11:50 not if they're equal 2010-10-04T18:11:52 what sigh said 2010-10-04T18:11:59 i agree with sigh as well 2010-10-04T18:12:10 if the other player waited in 1 or send to that neutral, you lose 2010-10-04T18:12:36 whereas in 2 they only beat you if they sent to your neutral straight away 2010-10-04T18:12:46 hmmm 2010-10-04T18:12:58 i would probably wait in 1 and send to your neutral in 2 2010-10-04T18:13:00 it depends on the strength of the neutral 2010-10-04T18:13:11 s = 4 in the first and 9 in the second 2010-10-04T18:13:13 I would send ships+1 to each neutral, and then if the enemy sent 100 ships at anything, I'd attack it the next turn. 2010-10-04T18:13:58 okay, what if ship sizes were adjusted that you couldn't attempt to claim both, but situation was otherwise isomprphic? 2010-10-04T18:14:09 i think it's possible to have situations with no clear draw strategy 2010-10-04T18:14:15 I'm not sure about s=9 anymore, for s=4 it is easy to defend by streaming full growth to the planet 2010-10-04T18:14:17 antimatroid2: you can always snipe, so just wait 2010-10-04T18:14:34 until, you don't have to wait, then send 2010-10-04T18:14:55 either way, the optimal strategy even for a draw switches quite a lot 2010-10-04T18:14:56 It makes no difference how many planets are in the center 2010-10-04T18:15:04 i don't think anyone has that kind of detail yet? 2010-10-04T18:16:03 *** Blkt has quit IRC (Ping timeout: 265 seconds) 2010-10-04T18:16:43 s=9 has no optimal strategy I think 2010-10-04T18:17:02 in pure strategies? 2010-10-04T18:17:07 yeah 2010-10-04T18:17:35 I'm not that familiar with analysing mixed strategies, so I'll leave that to you 2010-10-04T18:17:39 :) 2010-10-04T18:18:15 it's not much different, just find probability distributions for each player over their moves such that one can't can't theirs to increase their expected payoff 2010-10-04T18:18:21 the maths just gets harder :P 2010-10-04T18:19:06 what if you send all your ships from home planet (including future growth) to your enemies neutral 2010-10-04T18:19:31 i guess they still beat you if they do the same to their neutral 2010-10-04T18:19:32 hmmmm 2010-10-04T18:19:46 anyway, i've gotta run, i have a lecture at 10 2010-10-04T18:19:51 cya 2010-10-04T18:20:53 i think for 2 you really want to go for your neutral and back it up with any growth you get 2010-10-04T18:21:01 you can only stay in a draw position from doing that? 2010-10-04T18:21:05 otherwise win 2010-10-04T18:21:52 i guess it depends on whether you can do return trip to your planet in the time it takes them to get to you and if you had less starting ships at home 2010-10-04T18:22:26 i don't see how anyone plans to do well without some kind of gametree, even when you are only trying to guess at reasonable moves 2010-10-04T18:24:20 *** bhasker has joined #aichallenge 2010-10-04T18:25:32 I don't think you need a game tree if your positional analysis is strong enough. 2010-10-04T18:26:48 ultimately you need to be able to take into account your opponent's moves, but that's easier said than done 2010-10-04T18:29:37 *** McLeopold has quit IRC (Quit: Leaving) 2010-10-04T18:30:57 *** dmj111` has joined #aichallenge 2010-10-04T18:32:32 *** dmj111 has quit IRC (Ping timeout: 245 seconds) 2010-10-04T18:34:02 After my first submission, my bot played 22 games and got a rank of 1165, is this good or bad ? 2010-10-04T18:34:48 rank 1 is good, everything else is bad ;-) 2010-10-04T18:35:03 heh 2010-10-04T18:35:04 * dwchandler <-- bad 2010-10-04T18:35:17 means you're better than ~1620 and worse than ~1164 :) 2010-10-04T18:37:22 maybe 2010-10-04T18:39:09 *** otherAntimatroid has quit IRC (Quit: Page closed) 2010-10-04T18:39:15 gah! my last submission is down to 502 from a high of ~300 2010-10-04T18:39:25 proof that code rot exists 2010-10-04T18:39:51 200 bots improved since your last submission 2010-10-04T18:39:53 time to do the big rewrite 2010-10-04T18:40:12 Naktibalda: indeed ;-) 2010-10-04T18:41:59 *** Migi32 has quit IRC (Quit: http://boinc.berkeley.edu) 2010-10-04T18:44:27 *** MyHuggingBear has quit IRC () 2010-10-04T18:44:37 *** dmj111` has quit IRC (Ping timeout: 245 seconds) 2010-10-04T18:48:29 cool, found a neat bug in basically all versions of my bot 2010-10-04T18:48:45 why is the battle resolution so hard to get right? it's so dumb 2010-10-04T18:49:45 i've said before that i regret the battle resolution method that i was pushing for 2010-10-04T18:50:04 a1k0n: agree 2010-10-04T18:50:27 i have a situation where there's 3 neutrals, 5 player 1, 5 player 2 arriving. winner? neutral with -2 2010-10-04T18:50:29 it would have been simpler in practice to have the players fight before the winner fights neutral since that players fighting could have been integrated into fleet merging 2010-10-04T18:51:30 wait what the hell? i handle that situation?! 2010-10-04T18:51:42 *** optimum has quit IRC (Ping timeout: 245 seconds) 2010-10-04T18:51:47 you handle what? 2010-10-04T18:52:01 137 if(bs[1] > bs[0] && bs[1] >= bs[2]) { winner = 1; if(bs[2] > bs[0]) second = 2; } 2010-10-04T18:52:14 that should do it 2010-10-04T18:52:46 ohh duh, haha 2010-10-04T18:52:51 140 if(bs[winner] == bs[second]) 2010-10-04T18:52:52 141 winner = frames[i].owner; // no change in owner if it's a tie 2010-10-04T18:52:52 o_O 2010-10-04T18:53:05 hehe 2010-10-04T18:53:17 *** jmpespxoreax has quit IRC (Ping timeout: 264 seconds) 2010-10-04T18:53:32 *** jmpespxoreax has joined #aichallenge 2010-10-04T18:56:14 *** optimum has joined #aichallenge 2010-10-04T18:56:39 *** rebelxt has left #aichallenge 2010-10-04T18:59:12 Top 10 players: davidjliu(3799), ruilov(3728), dmj111(3725), cfaftw(3693), felixcoto(3663), rsergio(3646), EBraun(3634), sequoh(3624), DocBot(3622), Hazard(3599) 2010-10-04T19:04:28 *** Zeiris_ has quit IRC (Ping timeout: 265 seconds) 2010-10-04T19:04:52 *** hornairs has joined #aichallenge 2010-10-04T19:05:08 *** hornairs has quit IRC (Remote host closed the connection) 2010-10-04T19:21:37 *** cfaftw has joined #aichallenge 2010-10-04T19:36:37 *** foucist has quit IRC (Remote host closed the connection) 2010-10-04T19:47:58 *** Frontier has left #aichallenge 2010-10-04T19:51:14 the maximum number of turns is 200 ? or is that only for now ( not for final game standings,etc.. ) 2010-10-04T19:52:18 nullkuhl: I think it may change. Or not. But I would not program my bot to take this into account 2010-10-04T19:55:14 Itkovian: it is marginally useful. 2010-10-04T19:55:45 justin_pdx: if you could project until that turn, maybe, yes. Or am I mistaken? 2010-10-04T19:55:56 Yes. 2010-10-04T19:55:56 For example, say there was a growth 1 neutral planet with five ships on it that you could capture and the game has 10 turns left. 2010-10-04T19:56:32 Then you'd net 5. If it had 15 ships on it you couldn't recoup your costs within the time remaining, so it's useful to know. 2010-10-04T19:57:34 There is a slightly more complicated function for determining if it's worth it against an enemy planet/ 2010-10-04T19:59:13 Top 10 players: davidjliu(3792), ruilov(3758), dmj111(3728), cfaftw(3699), felixcoto(3655), rsergio(3648), sequoh(3638), DocBot(3630), EBraun(3625), Hazard(3600) 2010-10-04T20:04:06 *** jesionaj has quit IRC (Ping timeout: 240 seconds) 2010-10-04T20:08:05 *** jesionaj has joined #aichallenge 2010-10-04T20:11:43 *** optimum has quit IRC (Ping timeout: 245 seconds) 2010-10-04T20:12:29 *** optimum has joined #aichallenge 2010-10-04T20:13:23 *** nullkuhl has quit IRC (Ping timeout: 252 seconds) 2010-10-04T20:25:47 *** mceier has quit IRC (Quit: leaving) 2010-10-04T20:29:18 *** moriarty has joined #aichallenge 2010-10-04T20:30:00 blegh, uploading is broken :( 2010-10-04T20:31:46 *** justin_pdx has quit IRC (Quit: justin_pdx) 2010-10-04T20:32:13 *** Itkovian has quit IRC (Quit: Itkovian) 2010-10-04T20:35:47 i was wondering.. i just made a simple weighted average model and wanted to see how it would do 2010-10-04T20:39:32 any prizes? 2010-10-04T20:39:34 :) 2010-10-04T20:39:39 like 1mil, maybe? 2010-10-04T20:44:14 *** Bobng has quit IRC (Quit: Leaving) 2010-10-04T20:47:52 * amstan yawns 2010-10-04T20:49:33 yawn? 2010-10-04T20:50:17 heh, yawning is contagious over the internet too 2010-10-04T20:55:08 does anyone know if uploading is working yet? 2010-10-04T20:56:39 optimum: how is it broken? 2010-10-04T20:59:15 Top 10 players: davidjliu(3818), ruilov(3794), dmj111(3765), cfaftw(3724), felixcoto(3688), rsergio(3681), sequoh(3660), DocBot(3649), EBraun(3647), Hazard(3620) 2010-10-04T20:59:25 *** optimum has quit IRC (Remote host closed the connection) 2010-10-04T21:00:08 *** optimum has joined #aichallenge 2010-10-04T21:00:52 amstan: i'm not certain if it is my internet connection or the contest server, but if i try to upload, it seems to take forever. justin_pdx said it was broken 2010-10-04T21:02:07 *** Cyndre__ has quit IRC (Quit: Leaving) 2010-10-04T21:04:30 *** moriarty has left #aichallenge 2010-10-04T21:06:48 amstan: http://ai-contest.com/forum/viewtopic.php?f=18&t=865 2010-10-04T21:09:16 *** amram99 has joined #aichallenge 2010-10-04T21:12:52 uploading is broken for me.... if anyone is looking into it, i can help test 2010-10-04T21:14:55 DanielVF: ping 2010-10-04T21:15:00 j3camero: ping 2010-10-04T21:15:28 *** Meatkat has joined #aichallenge 2010-10-04T21:16:38 *** bhasker has quit IRC (Ping timeout: 252 seconds) 2010-10-04T21:16:45 Hi all, is the server down? 2010-10-04T21:17:29 oh fuck.. 2010-10-04T21:17:40 i think we hit the upper limit for entries in a folder 2010-10-04T21:17:49 *** krokokrusa has joined #aichallenge 2010-10-04T21:17:57 ls hangs.. 2010-10-04T21:17:58 lol 2010-10-04T21:18:04 :) 2010-10-04T21:18:07 lol 2010-10-04T21:18:28 yeah.. this is jeff's design feature... 2010-10-04T21:18:29 what's the upper limit? 2010-10-04T21:18:32 "keep all the entries" 2010-10-04T21:18:42 even if they're 3 months old 2010-10-04T21:18:44 PLEASE scratch my earlier entries 2010-10-04T21:18:46 and there's a million others 2010-10-04T21:18:47 they are embarassing 2010-10-04T21:19:04 haha 2010-10-04T21:19:10 yeah... problem is that there's no easy way for determining what the old entries are 2010-10-04T21:19:19 ls -altr 2010-10-04T21:19:32 there would have to be an agreed upon standard for "old entry" 2010-10-04T21:19:58 amstan@ltfs1110:/home/contest/ai-contest/planet_wars/submissions$ ls -altr|wc 2010-10-04T21:20:00 32001 256002 2111994 2010-10-04T21:20:06 yep.. there's a 32k there 2010-10-04T21:20:08 lol 2010-10-04T21:20:17 32k files? 2010-10-04T21:20:30 try a df 2010-10-04T21:20:50 nice 2010-10-04T21:20:56 That doesn't sound fun... 2010-10-04T21:20:59 nope 2010-10-04T21:21:01 lol 2010-10-04T21:22:51 i don't think 32k is a filesystem limit - http://stackoverflow.com/questions/466521/how-many-files-in-a-directory-is-too-many 2010-10-04T21:23:54 *** Accoun has quit IRC () 2010-10-04T21:24:07 *** Haggis has quit IRC (Quit: Page closed) 2010-10-04T21:24:48 is there a subfolder for every submission? 2010-10-04T21:25:00 *** Accoun has joined #aichallenge 2010-10-04T21:25:27 maybe it's not the filesystem's problem 2010-10-04T21:25:32 maybe it's just the php script that errors on that 2010-10-04T21:25:36 i successfuly touched a test 2010-10-04T21:25:56 if there's a subfolder per submission, the 32000 makes more sense to me 2010-10-04T21:26:05 yes, there is 2010-10-04T21:26:32 http://serverfault.com/questions/16757/is-there-any-limit-on-subfolders-within-folder 2010-10-04T21:26:32 ahh 2010-10-04T21:26:37 yeah, that's your problem then 2010-10-04T21:30:26 hmm. any submission that isn't the latest for a given user can be deleted, so how do you determine that? 2010-10-04T21:30:40 ok! so.. 2010-10-04T21:30:43 i can touch new files 2010-10-04T21:30:48 but i can't mkdir more folders 2010-10-04T21:32:15 *** adam__ has joined #aichallenge 2010-10-04T21:32:39 it looks like the userId is in the folder name, perhaps that's a start to removing the old folders 2010-10-04T21:32:41 not sure if you guys already know about this, but there are php errors when someone tries to upload, and it doesn't work 2010-10-04T21:32:48 yep ok 2010-10-04T21:33:00 too many symlinks or something in the directory 2010-10-04T21:34:04 http://gist.github.com/610796 2010-10-04T21:34:43 adam__: it's being looked at 2010-10-04T21:35:02 adam__: the submissions directory apparently hit an OS limit 2010-10-04T21:35:03 thanks, because I want to upload my code, again, and win.. :) 2010-10-04T21:35:07 yes 2010-10-04T21:35:10 http://ryan.ifupdown.com/2008/08/17/warning-mkdir-too-many-links/ 2010-10-04T21:35:10 FS limit 2010-10-04T21:35:20 yep 2010-10-04T21:35:20 right, FS limit, my bad 2010-10-04T21:35:25 amram99: we got 32000 folders.. 2010-10-04T21:35:28 woot! lol 2010-10-04T21:35:35 awesome! 2010-10-04T21:35:59 i got 32000 folders but your submission ain't one 2010-10-04T21:36:36 * jmcarthur shakes his head slowly 2010-10-04T21:36:53 lol 2010-10-04T21:37:03 maybe too many files in /tmp? 2010-10-04T21:37:30 adam__: huh? 2010-10-04T21:37:33 DanielVF: ping 2010-10-04T21:37:44 a1k0n pong 2010-10-04T21:37:50 DanielVF: yes! ok 2010-10-04T21:37:53 DanielVF: http://gist.github.com/610796 2010-10-04T21:37:59 (gotta in in a couple secs though. :( 2010-10-04T21:38:01 DanielVF: we have too many folders in submissions/ 2010-10-04T21:38:10 DanielVF: we have to delete the inactive ones 2010-10-04T21:38:10 hahaha 2010-10-04T21:38:12 ouch 2010-10-04T21:38:17 the contest site needs to delete old submissions; but for now someone needs to generate a list of active submissions and nuke the others 2010-10-04T21:38:20 that's pretty hilarious 2010-10-04T21:38:22 amstan We need to just store them 2010-10-04T21:38:29 or, hash the subdirectories 2010-10-04T21:38:34 like 1/2/4/1234675 2010-10-04T21:38:36 yeah 2010-10-04T21:38:53 DanielVF: let's dump a few so we can get it back online first 2010-10-04T21:39:00 you have 32000 entries? When I look at the leader boards there are like 2800? 2010-10-04T21:39:10 32000 uploads 2010-10-04T21:39:16 ahhh... 2010-10-04T21:39:17 so if someone submitted an average of 12 times 2010-10-04T21:39:23 so each upload is a new directory.. 2010-10-04T21:39:24 k 2010-10-04T21:39:33 *** Naktibalda has quit IRC (Remote host closed the connection) 2010-10-04T21:40:24 What if we toss all the submissions that failed to compile 2010-10-04T21:40:32 I have to go now, can do it in the morning 2010-10-04T21:40:43 or you can find dstuffed or j3camero 2010-10-04T21:40:55 DanielVF: darn 2010-10-04T21:41:01 * DanielVF out. :( 2010-10-04T21:41:01 well.. yeah.. ok 2010-10-04T21:41:15 give me root access, i'll do it :-) lol 2010-10-04T21:41:27 amram99: you forfeit? 2010-10-04T21:41:31 is Daniel the only one that could do it? 2010-10-04T21:41:34 haha 2010-10-04T21:41:38 hold on, j3camero's here, i think.. 2010-10-04T21:42:20 yeah, ok 2010-10-04T21:42:29 (09:34:04 PM) Alex: we have to delete old submissions 2010-10-04T21:42:30 (09:34:17 PM) Jeff Cameron: Okay it's cool. I have a script for that somewhere. 2010-10-04T21:42:37 doh! 2010-10-04T21:42:37 and the world rejoices! 2010-10-04T21:42:56 the existence of such a script not being a cron job or something baffles me 2010-10-04T21:43:09 find /home/contest/ai-contest/planet_wars/submissions/ -type d -ctime +20 -exec echo rm -rf {} \; 2010-10-04T21:43:21 *** arik181 has quit IRC (Ping timeout: 255 seconds) 2010-10-04T21:43:21 *** lovebob has quit IRC (Quit: Page closed) 2010-10-04T21:43:28 adam__: bad idea, some old those submissions might still be active 2010-10-04T21:43:33 remove the echo if you really want to delete directories older than 20 days 2010-10-04T21:43:33 you need to ask sql first 2010-10-04T21:43:34 i'm late in the convo, but why not have each user have its own submission dir? 2010-10-04T21:43:52 hmmm.. 2010-10-04T21:43:57 cfaftw: hindsight is 20/20 2010-10-04T21:44:00 that would be nice 2010-10-04T21:44:13 can you raise the fs limit easily? 2010-10-04T21:44:20 yeah sorry... i just wasn't sure if it was an option :) 2010-10-04T21:44:23 not unless you change the fs 2010-10-04T21:44:38 why not? 2010-10-04T21:44:41 oh the limit 2010-10-04T21:44:44 i misread 2010-10-04T21:44:53 *** krokokrusa has quit IRC (Ping timeout: 264 seconds) 2010-10-04T21:44:58 certain filesystems are more dynamically configurable than others 2010-10-04T21:45:11 ext3 has a hard limit of 32000 subfolders from the brief info I read 2010-10-04T21:45:11 Imma gonna clean up some old bots to fix the problem. Gimme ten minutes. 2010-10-04T21:45:52 yes, I just read ext3 is limited to 32000 per sub directory.. 2010-10-04T21:45:57 dang, no easy fixes.. 2010-10-04T21:46:00 *** arik181 has joined #aichallenge 2010-10-04T21:46:00 ok I have to run.. 2010-10-04T21:46:04 I'll check again later.. 2010-10-04T21:46:05 ext4? 2010-10-04T21:46:06 lol 2010-10-04T21:46:24 *** adam__ has left #aichallenge 2010-10-04T21:50:22 *** computerwiz_222 has joined #aichallenge 2010-10-04T21:55:53 *** Utkarsh has quit IRC (Ping timeout: 245 seconds) 2010-10-04T21:57:01 *** dmj111 has joined #aichallenge 2010-10-04T21:59:14 Top 10 players: davidjliu(3828), ruilov(3828), dmj111(3770), cfaftw(3730), felixcoto(3687), rsergio(3680), sequoh(3659), EBraun(3659), DocBot(3644), Hazard(3633) 2010-10-04T22:00:43 *** Utkarsh has joined #aichallenge 2010-10-04T22:02:03 let us know when we can submit again 2010-10-04T22:04:03 there's 31991 now 2010-10-04T22:04:11 idk where 10 folders dissapeared 2010-10-04T22:04:13 *** dmj111 has quit IRC (Ping timeout: 245 seconds) 2010-10-04T22:04:15 j3camero: ? 2010-10-04T22:05:02 deleting one folder every two minutes? :P 2010-10-04T22:05:08 cool 2010-10-04T22:05:11 i just submitted :) 2010-10-04T22:05:15 make that 31992 2010-10-04T22:05:30 96 2010-10-04T22:06:11 06 2010-10-04T22:06:12 amstan: oh chit where are all the folders gone? 2010-10-04T22:06:28 j3camero: ? 2010-10-04T22:06:49 its 06 + 31900 :) 2010-10-04T22:06:54 Yeah I made a little script like the one we had during tron that wakes up every few minutes and nukes a bunch of old submissions. 2010-10-04T22:07:13 Hopefully that's problem solved. 2010-10-04T22:07:13 j3camero: yeah.. i remember, we had space issues on that one 2010-10-04T22:07:17 i'm upset i missed the tron contest 2010-10-04T22:07:18 Someone try submitting now. 2010-10-04T22:07:24 it works! 2010-10-04T22:07:26 i didn't hear about the ai contest until i saw this one posted on reddit 2010-10-04T22:07:27 Sick. 2010-10-04T22:07:34 muchos gracias 2010-10-04T22:07:43 amstan saves the day by tracking me down, hehe. 2010-10-04T22:07:45 i heard about the tron contest as the results came out : 2010-10-04T22:07:46 :P 2010-10-04T22:07:52 *sigh* 2010-10-04T22:08:03 *** sinan has quit IRC (Ping timeout: 264 seconds) 2010-10-04T22:08:09 2SIGH 2010-10-04T22:08:13 @sigh 2010-10-04T22:08:13 i know... 2010-10-04T22:08:31 @yay 2010-10-04T22:08:31 Double rainbow all the way! 2010-10-04T22:08:37 cfaftw you just pwn3d my face... :( 2010-10-04T22:09:24 lol 2010-10-04T22:09:32 we just submitted at the same time, it seems :) 2010-10-04T22:09:37 So I really hope that my script works properly, and I didn't irreversibly nuke a bunch of people's bots. 2010-10-04T22:09:58 j3camero: pastebin the query that you're running 2010-10-04T22:10:21 might be best to move them to a different folder rather than deleting them 2010-10-04T22:10:31 Yeah... what place was your old bot in? 2010-10-04T22:10:35 amram99: you get the same problem then 2010-10-04T22:10:51 4th 2010-10-04T22:11:02 j3camero: I think nuking them is perfectly acceptable 2010-10-04T22:11:13 Meatkat, i have high hopes for this current submission 2010-10-04T22:11:35 So there's been quite an upset in the rankings lately. dmj has lost his grip on power. 2010-10-04T22:11:37 *** znutar has quit IRC (Quit: alice.) 2010-10-04T22:11:52 j3camero: dmj lost first a while back when Hazard came on the scene 2010-10-04T22:12:11 @rankings 2010-10-04T22:12:13 amstan: Top 10 players: ruilov(3864), davidjliu(3838), dmj111(3761), felixcoto(3694), rsergio(3666), Hazard(3657), DocBot(3653), sequoh(3644), EBraun(3641), adumlah(3609) 2010-10-04T22:12:29 i like how i don't even bother checking the main site anymore.. just use the bot 2010-10-04T22:12:31 hazard uploaded a new bot, and it seems to be worse 2010-10-04T22:12:45 sigh, indeed 2010-10-04T22:12:49 sigh: or other ppl got better 2010-10-04T22:12:57 sigh, i was afraid that his latest bot would be The One bot 2010-10-04T22:13:13 actually.... ruilov came out of nowhere and stole 1st 2010-10-04T22:13:18 cfaftw: haha, yeah 2010-10-04T22:13:28 man... I need to do some work to get back into the top 10 2010-10-04T22:13:40 @rankings 20 2010-10-04T22:13:41 amstan: Top 20 players: ruilov(3864), davidjliu(3838), dmj111(3761), felixcoto(3694), rsergio(3666), Hazard(3657), DocBot(3653), sequoh(3644), EBraun(3641), adumlah(3609), youwillfail(3607), _iouri_(3591), goldman(3591), fglider(3587), Arekku(3586), Manwe56(3556), ofrias(3555), Mistmanovx(3553), McLeopold(3541), bix0r4ever(3536) 2010-10-04T22:13:42 I mean, I honestly think that Ruilov's old bot at least was a couple notches above everybody else 2010-10-04T22:13:50 I uploaded a bot last night and I dropped form 12 to 22 :P 2010-10-04T22:13:56 davidjliu, definitely 2010-10-04T22:13:59 sigh: make that top 20, lol 2010-10-04T22:14:03 How do I scroll up and down in irssi? 2010-10-04T22:14:10 j3camero: page up and down 2010-10-04T22:14:10 amstan: I'll be in the top 20 if I upload my old bot 2010-10-04T22:14:48 j3camero: oh yeah.. at first i tought it was a php problem, so i went in ##php 2010-10-04T22:14:52 I just wanted to try this one out, even though it wasn't better, just different 2010-10-04T22:14:56 they started lolling at me 2010-10-04T22:15:05 I'm actually suprised it's at 21 2010-10-04T22:15:43 We need somebody to write a genetic algorithm that auto-submits bots once every hour to test different parameters. That would be ballin. 2010-10-04T22:15:54 Anyways, I am gonna get me some sweet sweet sleep. 2010-10-04T22:16:03 j3camero: not enough games for good sample 2010-10-04T22:16:04 night 2010-10-04T22:16:14 why are genetic algorithms the solution to everything? lol 2010-10-04T22:16:35 j3camero: oh yeah 2010-10-04T22:16:41 j3camero: i got a google interview, woot 2010-10-04T22:16:43 *** optimum has quit IRC (Ping timeout: 245 seconds) 2010-10-04T22:17:54 amstan: remember when I got you a job at Google? :-P 2010-10-04T22:18:12 amstan: blow it away, I know you will. 2010-10-04T22:18:18 lol 2010-10-04T22:18:39 amstan: just study up the 2nd and 3rd year algorithms like crazy. They will love you.. Mention that you run the GAC with Jeff Cameron. 2010-10-04T22:18:54 j3camero: of course 2010-10-04T22:19:17 amstan: remind me later this week and I'll drop a line to the recruiter so he knows to look for you. 2010-10-04T22:19:34 going to sleep now. Yessss. 2010-10-04T22:19:35 j3camero: who, zach? 2010-10-04T22:19:39 My precious. 2010-10-04T22:19:43 j3camero: will do 2010-10-04T22:19:45 amstan: yes, but no names, haha. 2010-10-04T22:20:02 i'll md5 hash him next time 2010-10-04T22:20:15 anyway, good night 2010-10-04T22:20:40 j3camero, you work at google? 2010-10-04T22:20:43 *** iFire has quit IRC (Ping timeout: 272 seconds) 2010-10-04T22:20:48 he used to 2010-10-04T22:20:52 ah 2010-10-04T22:21:00 but he's running this contest now? 2010-10-04T22:21:05 yes 2010-10-04T22:21:07 is he affiliated with waterloo? 2010-10-04T22:21:12 used to 2010-10-04T22:21:17 i am too 2010-10-04T22:21:22 haha 2010-10-04T22:21:24 haha 2010-10-04T22:21:45 technically this contest started as a project of csclub.uwaterloo.ca 2010-10-04T22:22:07 it used to have like 20 ppl in it 2010-10-04T22:22:11 compeeting 2010-10-04T22:22:18 *** iFire has joined #aichallenge 2010-10-04T22:23:27 *** optimum has joined #aichallenge 2010-10-04T22:23:44 come on.. questions and answers ppl, nobody interested? 2010-10-04T22:24:16 well.. this is boring 2010-10-04T22:24:23 * amstan goes to watch that 70s show 2010-10-04T22:25:01 amstan, sorry... my attention is split between here and Bones 2010-10-04T22:25:08 bones? 2010-10-04T22:25:11 amstan, didn't know you opened up an AMA, haha 2010-10-04T22:25:13 the show 2010-10-04T22:25:21 ama? 2010-10-04T22:25:26 "Ask Me Anything" 2010-10-04T22:25:28 it's a reddit thing 2010-10-04T22:25:37 i'm always an ama 2010-10-04T22:25:48 *** Naith has joined #aichallenge 2010-10-04T22:26:00 *** Meatkat has quit IRC () 2010-10-04T22:26:03 technically most people are... on reddit though, sometimes people post some revealing comment and people want to know more 2010-10-04T22:26:09 so they ask them to do an AMA thread 2010-10-04T22:28:35 amstan, are you with u of waterloo? 2010-10-04T22:28:49 i'm a student at uwaterloo 2010-10-04T22:28:59 sweet... how is your CS club so large? 2010-10-04T22:29:06 where is the thread? 2010-10-04T22:29:16 sigh: here 2010-10-04T22:29:32 haha 2010-10-04T22:30:44 cfaftw: no idea.. lots of advertisment i guess 2010-10-04T22:33:55 i don't think there were even that many students in my CS department! i went to U of Maryland. 2010-10-04T22:34:13 we have 300 ppl 2010-10-04T22:34:47 yeah... not sure there were 300 cs students at UMD :) 2010-10-04T22:34:58 there probably were, but it wasn't a large department 2010-10-04T22:43:47 *** iFire has quit IRC (Ping timeout: 240 seconds) 2010-10-04T22:45:35 *** amram99 has left #aichallenge 2010-10-04T22:46:07 *** iFire has joined #aichallenge 2010-10-04T22:49:50 time for sleep 2010-10-04T22:49:54 night everyone 2010-10-04T22:49:57 *** cfaftw has quit IRC (Quit: Leaving) 2010-10-04T22:51:03 *** JCS^ has joined #aichallenge 2010-10-04T22:55:51 *** iFire has quit IRC (Read error: Connection reset by peer) 2010-10-04T22:56:31 *** iFire has joined #aichallenge 2010-10-04T22:56:36 *** iFire has joined #aichallenge 2010-10-04T22:59:15 Top 10 players: ruilov(3877), davidjliu(3843), dmj111(3770), felixcoto(3708), rsergio(3689), sequoh(3657), DocBot(3653), EBraun(3648), Hazard(3640), adumlah(3613) 2010-10-04T22:59:51 *** genericbob has joined #aichallenge 2010-10-04T23:01:56 *** Titankiller has quit IRC (Quit: Nettalk6 - www.ntalk.de) 2010-10-04T23:02:14 *** iFire has quit IRC (Read error: Connection reset by peer) 2010-10-04T23:04:10 *** iFire has joined #aichallenge 2010-10-04T23:10:02 *** hexsprite has joined #aichallenge 2010-10-04T23:19:08 *** omorashi has joined #aichallenge 2010-10-04T23:20:40 amstan: it finally uploaded, and i got an error that it couldn't copy a temp file 2010-10-04T23:21:17 *** Zeiris_ has joined #aichallenge 2010-10-04T23:21:57 optimum: pastebin it 2010-10-04T23:24:02 amstan: unfortunately, i already tried uploading again, but i will if the error reproduces itself 2010-10-04T23:24:22 optimum: ... 2010-10-04T23:24:31 so is it a problem or not? 2010-10-04T23:25:00 *** omorashi has left #aichallenge 2010-10-04T23:25:03 *** bhasker has joined #aichallenge 2010-10-04T23:25:17 i'll let you know 2010-10-04T23:25:37 *** McLeopold has joined #aichallenge 2010-10-04T23:30:18 *** computerwiz_222 has quit IRC () 2010-10-04T23:33:14 *** Naith has quit IRC (Quit: Naith) 2010-10-04T23:33:59 *** Naith has joined #aichallenge 2010-10-04T23:35:39 is there a version of showgame that shows the planet ID of each planet? 2010-10-04T23:41:27 hexsprite: you could make one... 2010-10-04T23:41:38 hexsprite: easiest to change would be the canvas one 2010-10-04T23:41:47 *** yasith_ has quit IRC (Ping timeout: 240 seconds) 2010-10-04T23:41:51 ya... wheres the code for it? 2010-10-04T23:42:08 contestbot: where visualizer 2010-10-04T23:42:08 amstan: visualizer = (#1) http://github.com/DanielVF/Planet-Wars-Canvas-Visualizer, or (#2) http://csclub.uwaterloo.ca/~amstan/aichallenge/canvas-visualizer/ 2010-10-04T23:42:19 sweet tnx 2010-10-04T23:53:45 *** Frontier has joined #aichallenge 2010-10-04T23:54:11 *** davidjliu has quit IRC (Quit: Page closed) 2010-10-04T23:55:04 *** yasith_ has joined #aichallenge 2010-10-04T23:59:16 Top 10 players: ruilov(3872), davidjliu(3852), dmj111(3752), felixcoto(3701), rsergio(3690), Hazard(3681), sequoh(3660), DocBot(3640), EBraun(3630), adumlah(3619)