2011-06-06T00:00:03 jmcarthur: no idea 2011-06-06T00:00:11 jmcarthur: people should use later tell more often 2011-06-06T00:00:15 agreed 2011-06-06T00:00:16 or email 2011-06-06T00:00:42 people should just idle 2011-06-06T00:00:59 i do, but then people just ping me instead of saying what they want to say 2011-06-06T00:01:14 or maybe you mean McLeopold 2011-06-06T00:01:19 yep 2011-06-06T00:25:31 To be clear, in trueskill the sigma does not always go down - if there is a match which is an upset or something similar, the sigma may grow 2011-06-06T00:26:04 And when a player is ranked first, there is no upper bound, so the sigma may grow as well, because their max skill level is uncertain 2011-06-06T00:26:28 (you are using trueskill, right?) 2011-06-06T00:28:06 we are using trueskill 2011-06-06T00:28:14 they talk about it in the trueskill paper, the behavior you are seeing is not certain. and also, i'm starting to think trueskill is not a good ranking system to determine 1st, 2nd, 3rd 2011-06-06T00:28:20 but more of a good ranking system for matchmaking 2011-06-06T00:28:35 what else would you use for multiplayer ranking? 2011-06-06T00:28:38 bayeslo is out 2011-06-06T00:28:48 i don't really understand any of them tbh 2011-06-06T00:28:56 i don't know, i'm wondering the same thing myself 2011-06-06T00:29:16 i think that having a tournament might be better than using trueskill 2011-06-06T00:29:34 what do you mean by tournament? 2011-06-06T00:29:40 like, have one of those tournaments where everyone plays X games and then calculate standings from that 2011-06-06T00:29:46 something resembling a round robbin is not possible and incredibly inefficient 2011-06-06T00:29:47 like they do for chess 2011-06-06T00:29:54 there's too many bots 2011-06-06T00:30:00 i didn't say round robin 2011-06-06T00:30:01 you're looking at 4000+ bots 2011-06-06T00:31:04 i still like the idea of incrementally partitioning each section of the leaderboard and restarting a new tournament for each section 2011-06-06T00:31:12 with final rankings determined in the final round 2011-06-06T00:31:28 it means the final rank only really matters about how well you play against similarly ranked bots though 2011-06-06T00:33:17 http://en.wikipedia.org/wiki/Swiss-system_tournament 2011-06-06T00:36:16 contestbot: later tell mcleopold please see above(in the logs) 2011-06-06T00:36:16 amstan: Ready to serve, my lord. 2011-06-06T00:37:46 I'd rather stick with a ranking system than a tournament system 2011-06-06T00:37:53 it gives us a lot more flexibility 2011-06-06T00:38:15 and so far seems to work quite well 2011-06-06T00:38:55 in either case I definitely do not want to change what we do between the open and finals portion 2011-06-06T00:39:03 i don't personally think trueskill is at all appropriate for determining a winner at the end. bayeselo might be better for that, or maybe a real tournament. for the bulk of the contest, trueskill should be good though 2011-06-06T00:39:10 *** Cyndre has joined #aichallenge 2011-06-06T00:39:24 we'll also do the pairings cutoff probably 2011-06-06T00:39:34 so, it should get more accurate at the end 2011-06-06T00:40:17 trueskill is basically a modified glicko which has been shown to out perform elo pretty well 2011-06-06T00:40:31 bayeselo is not elo 2011-06-06T00:40:41 admittedly bayeselo may do better because it can operate on a global basis 2011-06-06T00:40:52 but we pretty much pushed the limits of that the last contest 2011-06-06T00:41:09 and yes I was typing that before your response ;) 2011-06-06T00:41:38 and generalizing bayeselo to a multiplayer game is certainly non-trivial 2011-06-06T00:41:44 definitely :( 2011-06-06T00:42:05 as is generalizing trueskill to bayestrueskill 2011-06-06T00:42:25 ;) 2011-06-06T00:42:29 :) 2011-06-06T00:42:58 and assuming the contest grows again I really don't think we can handle a non-incremental system of any kind 2011-06-06T00:43:54 *** delt0r___ has joined #aichallenge 2011-06-06T00:44:02 *** delt0r_ has quit IRC (Read error: Operation timed out) 2011-06-06T00:44:05 or at least it needs to have better than the quadratic time in number of players that bayeselo has 2011-06-06T00:44:09 we might get to the point that we just run trueskill for the dev period, give them a day extra or so to stabilize, then pick the top n bots and run a tournament with them 2011-06-06T00:44:29 might seem unfair, but probably the best we can do 2011-06-06T00:44:41 which will be completely untested :/ 2011-06-06T00:44:58 in a live situation at least 2011-06-06T00:45:11 yeah 2011-06-06T00:45:36 that's the main reason I really don't want to change the way the tournament runs for the finals 2011-06-06T00:46:23 it's just a tough situation 2011-06-06T00:46:54 it means we have to make hard tradeoffs between fairness/correctness and stability and time/cost 2011-06-06T00:48:09 yes, I just believe that so far using a ranking system with some sort of progressive cutoff has worked really well at balancing the tradeoffs 2011-06-06T00:48:55 i can't argue with that. i'm just speaking of possibilities and ideals 2011-06-06T00:49:06 of course it's been nice that the number 1 player was so dominant the last two contest too :) 2011-06-06T00:49:40 it's hard for anyone to argue with the result that way 2011-06-06T00:50:06 that might be less likely with ants :\ 2011-06-06T00:50:53 yes, I think so 2011-06-06T00:51:17 maybe if it's a bit ambiguous we could just round-robin the top n (for a pretty small n, like 10 or so) 2011-06-06T00:52:00 *** yasith has quit IRC (Ping timeout: 240 seconds) 2011-06-06T00:52:03 would be simple enough and small enough that we can error check it pretty easily 2011-06-06T00:54:24 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has quit IRC (Quit: Lost terminal) 2011-06-06T00:55:01 oh hmm, i forgot it's a >2 player game this time 2011-06-06T00:55:04 it just seems to open up too much room for arguing about the results 2011-06-06T00:55:42 also by narrowing the cutoff enough at some point it will basically become a roundrobin 2011-06-06T00:55:47 if we have to go that far 2011-06-06T00:57:03 we hit a technical limitation last time on the way pairing worked and how fast it ran that we couldn't lower the cutoff any further, but I think with the way pairing works now we should be able to bring the cutoff down to the point of a round robin if we needed to 2011-06-06T01:05:07 *** yasith has joined #aichallenge 2011-06-06T01:26:13 *** UncleVasya has joined #aichallenge 2011-06-06T01:31:04 wow, something very strange happened with the rankings. 2011-06-06T01:40:47 *** Palmik has joined #aichallenge 2011-06-06T01:54:38 what happened? 2011-06-06T01:56:50 my skill is about 50 now, it's definitely a bug. 2011-06-06T01:59:42 in this profile you can see many zeroes as outcome from the fights: http://aichallengebeta.hypertriangle.com/profile_games.php?user=70&page=1. I have them to and think everybody has. 2011-06-06T02:02:51 *** Ice_Harley has joined #aichallenge 2011-06-06T02:03:25 how are there 10 players on a map? 2011-06-06T02:06:39 *** UncleVasya has quit IRC (Ping timeout: 260 seconds) 2011-06-06T02:08:59 *** UncleVasya has joined #aichallenge 2011-06-06T02:25:45 aichallenge: janzert epsilon * r6ba8ba2 / manager/manager.py : Correct variable used - http://bit.ly/kto4Zi 2011-06-06T02:32:31 *** xQuasar has joined #aichallenge 2011-06-06T02:44:37 *** amstan has quit IRC (Remote host closed the connection) 2011-06-06T02:45:21 contestbot: later tell McLeopold setting the draw_probability to 0 is causing division by zero errors when updating the trueskill values for a game 2011-06-06T02:45:21 janzert1: Aye, aye, sir 2011-06-06T02:46:03 later tell McLeopold it also doesn't seem like it is correct at all since draws obviously happen at a much greater than 0% rate 2011-06-06T02:46:14 contestbot: later tell McLeopold it also doesn't seem like it is correct at all since draws obviously happen at a much greater than 0% rate 2011-06-06T02:46:14 janzert1: I come to serve. 2011-06-06T02:50:11 *** berak has joined #aichallenge 2011-06-06T03:24:40 *** ltriant has quit IRC (Quit: Computer has gone to sleep) 2011-06-06T03:25:27 *** mcstar has joined #aichallenge 2011-06-06T03:25:50 *** locutus2 has joined #aichallenge 2011-06-06T03:32:49 So has anyone taken a look at the fact that the jvm starting is often bugged? 2011-06-06T03:33:31 xQuasar: yes, as I replied to your earlier query about it ;) , it is a known problem on the beta server because it doesn't have much ram 2011-06-06T03:35:29 Sorry I'm being unreasonable. Time for bed I'm obviously getting much too irritable 2011-06-06T03:36:38 janzert1: sorry then, must've missed your reply 2011-06-06T03:37:11 yeah, not your fault. I think my reply was a good 15 minutes after you asked 2011-06-06T03:38:06 should know better than to let my irritation from trying to track down bugs bleed out into chat 2011-06-06T03:54:58 *** Zaphus has joined #aichallenge 2011-06-06T04:00:07 *** Zaphus has quit IRC (Quit: Page closed) 2011-06-06T04:02:40 *** qacek has quit IRC (Quit: WeeChat 0.3.5) 2011-06-06T04:36:37 *** Naktibalda has joined #aichallenge 2011-06-06T04:52:51 *** ulope has joined #aichallenge 2011-06-06T04:53:54 *** ulope has left #aichallenge 2011-06-06T05:03:44 *** locutus2 has quit IRC (Ping timeout: 240 seconds) 2011-06-06T05:11:49 *** locutus2 has joined #aichallenge 2011-06-06T05:20:19 *** UncleVasya has quit IRC (Ping timeout: 260 seconds) 2011-06-06T05:29:45 *** berak has quit IRC (Ping timeout: 255 seconds) 2011-06-06T05:35:38 *** berak has joined #aichallenge 2011-06-06T05:36:46 *** Zaphus has joined #aichallenge 2011-06-06T05:36:59 *** Frontier has joined #aichallenge 2011-06-06T05:37:52 anyone here ? 2011-06-06T05:39:31 *** olexs has joined #aichallenge 2011-06-06T05:43:27 *** locutus2 has quit IRC (Ping timeout: 246 seconds) 2011-06-06T05:54:34 *** yasith_ has joined #aichallenge 2011-06-06T05:54:53 *** Eruonen has joined #aichallenge 2011-06-06T05:56:41 *** yasith has quit IRC (Ping timeout: 240 seconds) 2011-06-06T06:01:39 *** olexs has quit IRC (Quit: Leaving.) 2011-06-06T06:09:05 *** locutus2 has joined #aichallenge 2011-06-06T06:10:54 *** UncleVasya has joined #aichallenge 2011-06-06T06:13:58 Zaphus: hello :) 2011-06-06T06:14:07 Hi 2011-06-06T06:14:12 it's been a while :-) 2011-06-06T06:14:26 erm, it must have been :P 2011-06-06T06:14:30 i don't remember you sorry :P 2011-06-06T06:14:58 your username seemed familiar though 2011-06-06T06:15:07 lol, I was around for the last comp - was doing OK but life got in the way (company folded, had to worry about how to pay for food!) 2011-06-06T06:15:08 but i think that was from a zaphyr on another site a few years ag 2011-06-06T06:15:17 yeah fair enough 2011-06-06T06:15:24 things going better now? 2011-06-06T06:15:50 yep, been having a look at the Ants comp, trying to find bugs 2011-06-06T06:15:57 no pun intended 2011-06-06T06:16:03 which language are you using? 2011-06-06T06:16:09 the pun should always be intended ;) 2011-06-06T06:21:34 *** UncleVasya has quit IRC (Ping timeout: 260 seconds) 2011-06-06T06:21:57 *** UncleVasya has joined #aichallenge 2011-06-06T06:26:05 I'm trying to replicate this bug: https://github.com/aichallenge/aichallenge/issues/159 2011-06-06T06:29:54 why is there red around water? 2011-06-06T06:33:14 i think the reason non-neon green ants are dying on turn 590 is colliding ants no? 2011-06-06T06:34:30 does anyone know what the deal is with the water square with red behind it? 2011-06-06T06:35:27 oh, i think that's where the link is going to? 2011-06-06T06:35:45 it could be colliding ants 2011-06-06T06:36:45 pretty sure it's all colliding ants 2011-06-06T06:36:48 and there's no bug :P 2011-06-06T06:37:10 what language are you using? c++ <3 2011-06-06T06:38:52 one thing to check would be that collided ants aren't participating in battles 2011-06-06T06:39:07 Zaphus: want to check that one out :P 2011-06-06T06:39:25 just make a map where you can cause it to happen after one turn 2011-06-06T06:40:36 yeah, c++ 2011-06-06T06:40:55 i'm an awful person, have you stripped out the 'r' line information? 2011-06-06T06:41:16 ? 2011-06-06T06:41:33 oh, the removed food info 2011-06-06T06:41:40 in the >> function in State.h, remove the processing of 'r', and in clear() add food.clear() 2011-06-06T06:41:49 are you on windows? 2011-06-06T06:41:53 yep 2011-06-06T06:42:00 what did you do about the timer? 2011-06-06T06:42:12 grabbed the code I used in the last comp 2011-06-06T06:42:12 or are you using mingw like a sane person? 2011-06-06T06:42:25 is it os agnostic? 2011-06-06T06:42:28 or windows specific? 2011-06-06T06:42:38 if it's os agnostic, would you be able to pastebin it for me? 2011-06-06T06:42:43 can I cut and paste to here ? 2011-06-06T06:42:50 pastebin.com 2011-06-06T06:42:58 you paste it into a text box then give me the link 2011-06-06T06:43:01 might be .org 2011-06-06T06:43:07 google is your friend :P 2011-06-06T06:43:34 i'm technically in the middle of an exam till wednesday, but i'll fix the cpp bot up once that's finished 2011-06-06T06:43:38 http://pastebin.com/aMe6866X 2011-06-06T06:43:58 ignore the conio include, thats for something else 2011-06-06T06:44:29 gross :P, i'll see if i can use that in a few days 2011-06-06T06:44:33 i hate timer code 2011-06-06T06:47:15 I might have old c++ code, there is no clear() and there is no 'food' - it uses isFood 2011-06-06T06:47:24 unless you just mean set it to false/0 2011-06-06T06:51:46 *** Cyndre has quit IRC (Ping timeout: 260 seconds) 2011-06-06T06:57:37 *** xQuasar has quit IRC (Ping timeout: 252 seconds) 2011-06-06T06:59:05 Zaphus: https://github.com/aichallenge/aichallenge/blob/epsilon/ants/dist/starter_bots/cpp/State.cc 2011-06-06T06:59:08 line 211 on that page 2011-06-06T06:59:35 oh, and then it'd be reset() where you want to clear food, 2011-06-06T06:59:55 yeah, square.reset() 2011-06-06T07:00:35 except... I kind of like what you did, as a strategy - if I saw two bits of food, went to get one, it would remember the other and go look for it, thats kind of neat 2011-06-06T07:01:02 lol, that starter bot is really broken 2011-06-06T07:01:08 *** xQuasar has joined #aichallenge 2011-06-06T07:01:23 i'm going to upload a different version now 2011-06-06T07:01:42 Does the server send all the information about what your ants can see, every turn? 2011-06-06T07:01:47 whats the deal with the r anyway, if you can see the square you can see it has no food, otherwise you cant see it and won't get the r - is it purely to let the bots optimize a bit ? 2011-06-06T07:01:51 If so, wouldn't it be more efficient to only send the changes? 2011-06-06T07:02:11 'r' line has been removed from the engine 2011-06-06T07:02:14 ignore it's existence 2011-06-06T07:03:03 thunderstorm again bb 2011-06-06T07:03:06 *** mcstar has quit IRC (Quit: WeeChat 0.3.5) 2011-06-06T07:04:28 xQuasar: Sending full vision is good, better than requiring everyone to write a bot that can track all the changes. might be a bit more efficient to send changes but it's not the bottleneck 2011-06-06T07:04:43 *** Cyndre has joined #aichallenge 2011-06-06T07:05:28 antimatroid: if 'r' was removed from the engine, then your bot code would keep thinking there was food all over the place, right ? ahhh 2011-06-06T07:06:12 Zaphus: 3... 2... 1... 2011-06-06T07:06:46 impressive how high I got my bot with that in there :-) 2011-06-06T07:06:48 * antimatroid kicks CIA 2011-06-06T07:08:44 ah crud, something messed up 2011-06-06T07:11:21 the last contest we have the ugly thing of climbing up the ranking for each new bot 2011-06-06T07:11:55 what do you mean? 2011-06-06T07:11:58 i see this time there are the ghost entry of last submission for better comparision 2011-06-06T07:12:00 pw didn't have that which was terrible 2011-06-06T07:12:06 i don't like that idea 2011-06-06T07:12:24 it's good to make a new submission work it's way up the leaderboard 2011-06-06T07:12:38 yeah, I like the ghosts... but what happens when we have 4,000 ghosts ? need a ranking page that has no ghosts for easy viewing 2011-06-06T07:12:40 it encourages people not to resubmit as often and forces people to see how their bot does against lower ranked bots 2011-06-06T07:12:55 *** xQuasar has quit IRC (Ping timeout: 252 seconds) 2011-06-06T07:13:01 i have read the papers to trueskill and think we should for each new version start mit the last mean but the starter sigma 2011-06-06T07:14:12 because the bot has probably a similar strenght 2011-06-06T07:14:19 that's better 2011-06-06T07:14:41 not your comments sorry, my gitfu is poor 2011-06-06T07:14:44 but if this wrong through the high sigma for new bots this corrected very fast 2011-06-06T07:15:44 * antimatroid kicks CIA-26 again 2011-06-06T07:15:57 problem with starting at the bottom is it discourages changes in the final weeks, since you may never make it back up to the top in time 2011-06-06T07:16:13 the rankings are all cleared at the start of the contest 2011-06-06T07:16:38 i submitted my bot for the first time in a month or so for pw an hour or so before the contest 2011-06-06T07:17:21 the main server is a crap place to test your bot, it's just for bragging rights imo 2011-06-06T07:17:24 tcp is where it's at 2011-06-06T07:18:20 also if we start a new version with old mean the trueskill updates for the other players in the same match a more realistic and not biased to the 'wrong' strength of the new version 2011-06-06T07:20:57 apparently cia is dead :\ 2011-06-06T07:21:39 the problem with TCP is that I can't watch while at work. Plus timeout / earlyout can only be tested properly on the main server 2011-06-06T07:22:00 why can't you watch while at work? 2011-06-06T07:22:34 it's going to be tough going back to non-live tcp games 2011-06-06T07:22:37 <3 jbotmanager 2011-06-06T07:23:17 I guess I could leave my PC on all day, playing TCP 2011-06-06T07:25:31 does anyone have a TCP test server going yet ? 2011-06-06T07:27:22 *** xQuasar has joined #aichallenge 2011-06-06T07:27:53 aichallenge: Nick Ham epsilon * r49170e8 / (5 files): removed 'r' lines - http://bit.ly/jpsJCy 2011-06-06T07:30:09 ... that wasn't delayed 2011-06-06T07:30:19 no tcp server yet 2011-06-06T07:39:36 I've made those same changes to my bot (with a few personal tweaks) - still behaves properly, which is nice. 2011-06-06T07:40:00 'r' line has been removed from the engine <- are you serious or joking D: 2011-06-06T07:40:40 *** Quasar has joined #aichallenge 2011-06-06T07:41:06 *** Quasar is now known as Guest98744 2011-06-06T07:44:11 *** xQuasar has quit IRC (Ping timeout: 240 seconds) 2011-06-06T07:44:53 ah, the version of the playgame tools I have still have the 'r' support in them, I think - so it cleans up after itself 2011-06-06T07:45:38 xQuasar: that happened AGES ago 2011-06-06T07:46:27 i had been holding off to change it till i fixed up the timer, but hadn't got around to it, so i finally did then 2011-06-06T07:46:29 hmm - ok, then my code is really smart and it didnt get upset with the invisible food - because it makes zero difference to the behaviour of my (local) bot 2011-06-06T07:46:31 it's a very simple fix though 2011-06-06T07:47:05 Zaphud: you probably also want to add isFood = 0 to clear in Square.h 2011-06-06T07:47:12 so 3 changes 2011-06-06T07:47:27 well this way we have to refresh ALL of our data on our bot every turn D: 2011-06-06T07:47:31 so inefficient ;-; 2011-06-06T07:47:38 *** Guest98744 is now known as xQuasar 2011-06-06T07:48:03 it barely costs anything resource wise 2011-06-06T07:48:12 What if the map's really big :o 2011-06-06T07:48:21 *** mcstar has joined #aichallenge 2011-06-06T07:48:29 the real problem with no 'r' lines is the implicit information 2011-06-06T07:48:33 but i lost that argument 2011-06-06T07:59:15 *** olexs has joined #aichallenge 2011-06-06T08:00:03 ok, I grabbed the latest game engine from git, now I see a difference between before and after (not much, but a difference none the less) - the old tools must have still had the r in them 2011-06-06T08:03:48 *** mcstar has quit IRC (Ping timeout: 246 seconds) 2011-06-06T08:07:52 *** Frontier has joined #aichallenge 2011-06-06T08:39:49 *** locutus2 has quit IRC (Ping timeout: 260 seconds) 2011-06-06T08:43:15 *** Zaphus has quit IRC (Quit: Page closed) 2011-06-06T08:43:53 *** delt0r_ has joined #aichallenge 2011-06-06T08:44:13 *** xQuasar has quit IRC () 2011-06-06T08:45:28 *** delt0r___ has quit IRC (Ping timeout: 244 seconds) 2011-06-06T08:55:39 what is 'yesterday' in chat slang? 2011-06-06T08:56:34 *** sigh has joined #aichallenge 2011-06-06T08:56:55 *** FireFly has joined #aichallenge 2011-06-06T09:03:14 *** locutus2 has joined #aichallenge 2011-06-06T09:07:32 *** Ice_Harley has quit IRC (Ping timeout: 252 seconds) 2011-06-06T09:09:27 Frontier: what's the definition of "the"? 2011-06-06T09:09:29 no looking it up either 2011-06-06T09:09:41 i can't think of an answer to your question 2011-06-06T09:09:47 and it reminded me how hard defining the is 2011-06-06T09:25:28 *** berak has quit IRC (Quit: berak) 2011-06-06T09:26:52 *** amstan has joined #aichallenge 2011-06-06T09:26:52 *** ChanServ sets mode: +o amstan 2011-06-06T09:28:21 *** locutus2 has quit IRC (Ping timeout: 244 seconds) 2011-06-06T09:29:14 hello 2011-06-06T09:29:18 *** yasith_ has quit IRC (Quit: The Sniper Got Me) 2011-06-06T09:35:10 hello 2011-06-06T09:40:48 *** jugde has joined #aichallenge 2011-06-06T09:41:53 evening 2011-06-06T09:56:40 *** mcstar has joined #aichallenge 2011-06-06T10:02:19 @topic 2011-06-06T10:02:19 antimatroid: Official Google AI Challenge: http://ai-contest.com/ || Channel Logs: http://contestbot.hypertriangle.com/ || Code Repo: http://github.com/aichallenge/aichallenge || Beta testers needed: http://aichallengebeta.hypertriangle.com/ (amstan) 2011-06-06T10:07:13 *** Eruonen has quit IRC () 2011-06-06T10:10:30 *** sigh has quit IRC (Remote host closed the connection) 2011-06-06T10:19:51 *** InsaneMalkavian has joined #aichallenge 2011-06-06T10:30:00 *** Kingpin13 has joined #aichallenge 2011-06-06T10:31:54 *** smiley1983 has joined #aichallenge 2011-06-06T10:42:51 *** Torquemada has joined #aichallenge 2011-06-06T10:43:27 hi guys 2011-06-06T10:43:42 *** berak has joined #aichallenge 2011-06-06T10:45:40 amstan, could you check, why I didn't recieved a registration email? I've registered for beta testing on prev week. 2011-06-06T10:45:51 there is no registration email 2011-06-06T10:46:07 there's an activation link after you register 2011-06-06T10:46:10 you have to click it 2011-06-06T10:46:52 Torquemada: did you miss it? 2011-06-06T10:47:19 Yeah i clicked it. Ok I'll try to enter now. 2011-06-06T10:48:55 I enter, thanks. 2011-06-06T10:55:23 *** amstan has quit IRC (Ping timeout: 250 seconds) 2011-06-06T11:01:25 *** berak has quit IRC (Write error: Broken pipe) 2011-06-06T11:10:38 *** amstan has joined #aichallenge 2011-06-06T11:10:38 *** ChanServ sets mode: +o amstan 2011-06-06T11:11:49 *** locutus2 has joined #aichallenge 2011-06-06T11:33:10 *** amstan has quit IRC (Read error: Operation timed out) 2011-06-06T11:44:50 *** amstan has joined #aichallenge 2011-06-06T11:44:50 *** ChanServ sets mode: +o amstan 2011-06-06T11:48:01 *** Eruonen has joined #aichallenge 2011-06-06T11:48:33 *** amstan has quit IRC (Client Quit) 2011-06-06T11:50:10 *** amstan has joined #aichallenge 2011-06-06T11:50:10 *** ChanServ sets mode: +o amstan 2011-06-06T11:54:05 *** BtbN has quit IRC (Remote host closed the connection) 2011-06-06T12:05:58 *** jmcarthur is now known as Gregamahol 2011-06-06T12:06:38 *** Gregamahol is now known as jmcarthur 2011-06-06T12:08:19 *** Torquemada has quit IRC (Quit: Page closed) 2011-06-06T12:12:00 jmcarthur: greg am ahole :P 2011-06-06T12:13:27 heh 2011-06-06T12:13:47 somebody in another channel lost their home internet connection so about ten people stole variations of his nick 2011-06-06T12:14:43 *** okayzed has quit IRC (Ping timeout: 246 seconds) 2011-06-06T12:18:49 haha :P 2011-06-06T12:18:50 sounds fair 2011-06-06T12:19:01 *** computerwiz_222 has joined #aichallenge 2011-06-06T12:19:12 hello 2011-06-06T12:19:46 *** bhasker has quit IRC (Quit: bhasker) 2011-06-06T12:20:19 *** BtbN has joined #aichallenge 2011-06-06T12:35:26 *** guyverthree has joined #aichallenge 2011-06-06T12:35:32 hello all 2011-06-06T12:37:48 i'm having a problem with running the client in a browser 2011-06-06T12:39:51 hi. what browser do you use? 2011-06-06T12:41:07 what client? 2011-06-06T12:42:15 *** computerwiz_222 has quit IRC (Remote host closed the connection) 2011-06-06T12:42:50 *** locutus2 has quit IRC (Ping timeout: 250 seconds) 2011-06-06T12:43:33 *** _0x47 <_0x47!~slash@p57B3FB61.dip.t-dialin.net> has quit IRC (Ping timeout: 255 seconds) 2011-06-06T12:48:42 *** amstan has quit IRC (Quit: Konversation terminated!) 2011-06-06T12:49:06 *** janzert1 is now known as janzert 2011-06-06T12:49:16 internet explorer 2011-06-06T12:49:38 version? 2011-06-06T12:49:45 i agree to use the active X control and it gives back a typeError 2011-06-06T12:50:05 IE 8 2011-06-06T12:50:34 interesting, IE8 should work fine actually 2011-06-06T12:50:45 object doesn't support this property or method 2011-06-06T12:51:03 you are just trying to view a game, right? 2011-06-06T12:51:19 yep 2011-06-06T12:51:31 just starting to get the tools running 2011-06-06T12:51:47 the planetwars one worked fine straight fgrom the box 2011-06-06T12:51:50 :) 2011-06-06T12:52:37 hm, I have IE9 on my machine already, can't really test it. but activeX already sounds wrong, game viewer is html5 canvas + javascript, not activeX 2011-06-06T12:53:22 add ?debug=true to your URL (&debug=true if the URL already has a ? in it) and see if anything changes 2011-06-06T12:54:14 C:\Documents and Settings\jamesd\Desktop\AI\Ants\game_logs\replay.html 2011-06-06T12:54:19 is the user 2011-06-06T12:54:22 url 2011-06-06T12:54:26 *** smiley1983 has quit IRC (Quit: leaving) 2011-06-06T12:54:37 try C:\Documents and Settings\jamesd\Desktop\AI\Ants\game_logs\replay.html?debug=true 2011-06-06T12:54:40 however it does not have a ? in it for arguments 2011-06-06T12:55:03 windows cannot find the url 2011-06-06T12:56:05 okay, doesn't work with IE and local content, it seems 2011-06-06T12:56:53 what if you open a game from the beta server, like http://aichallengebeta.hypertriangle.com/visualizer.php?game=15644 ? does that run properly? 2011-06-06T12:57:12 just gonna open in in firefox from now on 2011-06-06T12:57:47 that should do it, as well :) chrome has best performance in html5, firefox and IE can get slow on large games 2011-06-06T12:58:15 as for a sidenote: when i tried to view a game on win 7 with ie whichever.version it complained that a dont have Java installed and it didnt show anything but a nice errorbox 2011-06-06T12:58:25 *** _0x47 <_0x47!~slash@p5B12EF8E.dip.t-dialin.net> has joined #aichallenge 2011-06-06T12:58:32 *** InsaneMalkavian has quit IRC (Ping timeout: 252 seconds) 2011-06-06T12:58:42 a->I 2011-06-06T12:58:51 ie9 works for me on win7 x64 2011-06-06T12:59:06 but why does it need java? 2011-06-06T12:59:19 no idea :) 2011-06-06T12:59:30 on chromium it worked 2011-06-06T12:59:34 in 2011-06-06T12:59:41 silly IE 2011-06-06T12:59:49 not that i care a bit im just saying 2011-06-06T13:00:14 Frontier: do you know anything about this? 2011-06-06T13:00:58 older IE versions don't support canvas so it falls back to a java applet 2011-06-06T13:01:06 not really sure what it does in IE9 2011-06-06T13:01:14 it cant be that old 2011-06-06T13:01:25 its from a win7 enterprise 2011-06-06T13:01:37 IE9 was released just a few months ago 2011-06-06T13:01:39 win7 came with ie8, ie9 came as an optional update 2011-06-06T13:01:39 so i guess it must be at least ie8 right? 2011-06-06T13:01:50 *** guyverthree has quit IRC (Ping timeout: 252 seconds) 2011-06-06T13:01:53 yes, but IE8 is old enough :) 2011-06-06T13:01:57 ok 2011-06-06T13:05:34 *** aeter has joined #aichallenge 2011-06-06T13:07:18 *** amstan has joined #aichallenge 2011-06-06T13:07:18 *** ChanServ sets mode: +o amstan 2011-06-06T13:07:50 *** olexs has quit IRC (Quit: Leaving.) 2011-06-06T13:17:54 *** Naktibalda has quit IRC (Quit: ChatZilla 0.9.86.1 [Firefox 4.0.1/20110413222027]) 2011-06-06T13:20:14 *** bhasker has joined #aichallenge 2011-06-06T13:41:43 *** okayzed has joined #aichallenge 2011-06-06T13:51:18 *** itsme has joined #aichallenge 2011-06-06T13:51:22 mcstar: You need at least IE9. Previous versions of IE don't support enough web standards 2011-06-06T13:52:24 *** guyverthree has joined #aichallenge 2011-06-06T13:52:29 hey guys 2011-06-06T13:52:31 mcstar: Especially the html5 canvas is missing for the visualizer. Therefor I added a Java version, that emulates the canvas. But some of the CSS3 formating needed work arounds in IE as well 2011-06-06T13:52:39 another small problem 2011-06-06T13:54:03 *** komeedipoeg has joined #aichallenge 2011-06-06T13:54:12 mcstar: there is a static check for IE < 8 in the php. In this case Java is used. 2011-06-06T13:54:22 sorry, IE <= 8 2011-06-06T13:54:26 Hi 2011-06-06T13:55:07 In the previous contest we had practically the same setup 2011-06-06T13:56:12 I cannot seem to get a java bot to run a game 2011-06-06T13:56:24 it's only the start bot one 2011-06-06T13:56:33 it keeps crashing 2011-06-06T13:56:43 i think i'm running the command incorrectly 2011-06-06T13:57:00 i'm using java MyBot/build/classes/MyBot 2011-06-06T13:57:04 where can i find the python module Queue? ImportError: No module named Queue 2011-06-06T13:57:06 Hi 2011-06-06T13:57:08 Frontier: thx for the clarification, however this doesnt concern me at all(not using windows), but might be of interest to some other ppl, cant you throw an error message like: the problem would be solved if you lazy bastard had installed IE9 :D 2011-06-06T13:58:10 I'm using windows and I try to run the example on the tools package 2011-06-06T13:58:15 $ ./play_one_game.sh running for 100 turns turn 0 stats: ant_count: [1, 1, 1, 1] score 1 1 1 1 status timeout crashed crashed timeout waiting 0.25 seconds for bots to process end turn 2011-06-06T13:58:19 hehe when i coded that ie9 wasn't even ready 2011-06-06T13:58:29 any clue?? 2011-06-06T13:58:34 itsme 2011-06-06T13:58:39 i have the same thing 2011-06-06T13:58:42 i had 2011-06-06T13:58:55 you need tyo change the paths of the sample bots slightly 2011-06-06T13:59:21 change the paths to this 2011-06-06T13:59:30 python sample_bots/python/LeftyBot.py 2011-06-06T13:59:40 basically remove the dst from the start of the tools thing 2011-06-06T13:59:44 o it's true 2011-06-06T14:00:33 mcstar: The problem can be solved by either installing IE9 OR Java. I'm not sure IE9 is supported on Windows 2000/XP. 2011-06-06T14:00:54 now it gave timeouts 2011-06-06T14:01:29 itsme: turn on debugging and see if the bots spit out some useful info 2011-06-06T14:01:29 at least on second turn 2011-06-06T14:01:42 -E 2011-06-06T14:02:00 *** guyverthree has quit IRC (Quit: Page closed) 2011-06-06T14:02:33 ./play_one_game.sh -E ???? 2011-06-06T14:02:48 give it to playgame.py 2011-06-06T14:02:59 in play_one_game.sh 2011-06-06T14:03:18 -E or --E 2011-06-06T14:03:24 doestn matter 2011-06-06T14:04:28 --E -> playgame.py: error: no such option: --E 2011-06-06T14:04:41 but -E works, where the debug is stored?? 2011-06-06T14:05:35 *** Frontier has quit IRC (Quit: Leaving.) 2011-06-06T14:06:25 i think -E should make the bot errors appear in the console 2011-06-06T14:06:51 the out it gave is: 2011-06-06T14:06:52 $ ./play_one_game.sh running for 100 turns turn 0 stats: ant_count: [1, 1, 1, 1] turn 1 stats: ant_count: [1, 1, 1, 1] score 1 1 1 1 status timeout timeout timeout timeout waiting 2.25 seconds for bots to process end turn 2011-06-06T14:06:57 theres another one, -e which spits them out into files in a user given directory 2011-06-06T14:07:12 thats not it 2011-06-06T14:07:38 if the bot uses the stderr to print error messages, -E makes them visible 2011-06-06T14:07:55 otherwise, the engine doesnt show them 2011-06-06T14:08:09 ok i'm using the defaults bots 2011-06-06T14:08:23 it's the first time i have download the tools.zip 2011-06-06T14:09:12 are you trying this in cygwin? 2011-06-06T14:09:25 yes 2011-06-06T14:09:29 (why cant a windows person answer his question anyway?) 2011-06-06T14:09:46 because in windows native console the are problems with python 2011-06-06T14:09:49 ok, sure the path is correct to the bots? 2011-06-06T14:10:30 well theres a play_one_game.cmd which is supposed to be run from cmd 2011-06-06T14:11:19 C:\ai_contest\ants\tools>play_one_game.cmd File "C:\ai_contest\ants\tools\playgame.py", line 159 print("Running profile and outputting to %s" % (prof_file,), file=sys.stderr) ^ SyntaxError: invalid syntax 2011-06-06T14:12:19 maybe your python is simply old 2011-06-06T14:12:29 is there any windows guy who have managed to run the examples?????? 2011-06-06T14:12:37 you need at least 2.6 but i think 2.7 is recommended 2011-06-06T14:12:46 yes, sure many 2011-06-06T14:13:17 ok i have 2.5.2 2011-06-06T14:13:26 this may be the motive, thanks 2011-06-06T14:13:37 np 2011-06-06T14:19:18 don't even try with 2.5, lol 2011-06-06T14:19:38 Traceback (most recent call last): File "E:\ants\playgame.py", line 16, in from engine import run_game ImportError: No module named engine 2011-06-06T14:19:57 python 2.7.1 2011-06-06T14:22:57 ls ..\worker 2011-06-06T14:23:14 do you have engine.py in there? 2011-06-06T14:23:32 *** aeter has left #aichallenge 2011-06-06T14:25:44 wwooohh, itworks, great 2011-06-06T14:25:56 this nigth iwillbeginmy bot 2011-06-06T14:26:08 it works :D 2011-06-06T14:27:09 and i will have to review my space key 2011-06-06T14:28:04 many thanks to all they guys that have worked so hard to have ready this constest 2011-06-06T14:28:54 *** McLeopold has joined #aichallenge 2011-06-06T14:29:35 anyone need help with playgame syntax? 2011-06-06T14:31:20 @later tell jmcarthur I'm worried about the accuracy of our TrueSkill implementation. It seems odd that nobody is above 60. I lowered the draw margin and reset submissions to see if it makes a difference. Janzert noticed I did it wrong, so I need to correct. 2011-06-06T14:31:20 McLeopold: Ready to serve. 2011-06-06T14:31:21 *** itsme has quit IRC (Quit: Page closed) 2011-06-06T14:33:55 McLeopold: I did reset the draw margin to 10% last night 2011-06-06T14:34:05 just to stop the division errors 2011-06-06T14:34:07 ok, I'm going to set it to 1% 2011-06-06T14:34:15 why are you lowering it? 2011-06-06T14:34:20 I'm checking the database for any actual draws 2011-06-06T14:34:22 the actual rate seems higher 2011-06-06T14:34:43 at least there was a divide by zero pretty much every other game 2011-06-06T14:35:52 A higher draw margin makes bigger updates for a win 2011-06-06T14:35:55 also the my range and average looked fairly healthy to me last night 2011-06-06T14:36:24 err, mu not my :P 2011-06-06T14:36:41 hmm... yeah, the rankings look a lot better 2011-06-06T14:37:02 I think there is a slight error somewhere that makes skill drop over time 2011-06-06T14:37:57 The draws are from games with only starter bots getting stuck in the same location. 2011-06-06T14:38:11 I think the draw percentage will actually be lower, close to zero. 2011-06-06T14:38:15 hmm, could be I see the average my is now 48.47, last night it was 49.xxx 2011-06-06T14:38:43 I check some trueskill updates against the online calculator, and we are off 2011-06-06T14:39:20 we'll probably always have a fairly good sized group of starter bots so I don't think the draw percentage will ever approach 0 2011-06-06T14:39:30 I did see that someone said sigma may rise, so I was wrong there 2011-06-06T14:39:58 But I'm not tuning the system for starters, I'm tuning for actual bots 2011-06-06T14:40:11 I was wondering how much floating point rounding error we're getting with from some of the normal distribution functions 2011-06-06T14:40:21 A win doesn't mean as much with a lower draw percentage, which is good 2011-06-06T14:40:31 but the overall system has to account for the group of starter bots there will be 2011-06-06T14:40:44 That might be the issues, the rounding. 2011-06-06T14:41:12 adjust beta if you're trying to adjust how much a win means 2011-06-06T14:41:15 that's what it's for 2011-06-06T14:41:40 I was thinking about that. I would like a feature where a bot is only active for maybe a week or so, then deactivated. You'd have to log in and request more games to get reactivated. That would drop out abandonded accounts. 2011-06-06T14:42:07 yeah, I've thought about that before. I have mixed feelings about it 2011-06-06T14:42:08 More games for people who are actually participating 2011-06-06T14:42:22 what's the downside? 2011-06-06T14:42:44 people get mad because there submission is dropped out of the contest? 2011-06-06T14:42:52 why dont you request users periodically to renew their bot status? 2011-06-06T14:43:00 maybe with a clear explanation and an email when the deactivation occurs 2011-06-06T14:43:37 It wouldn't drop out, just not be scheduled for games. Plus, they can click one button to reactivate. 2011-06-06T14:43:51 and there have certainly been top 10 bots that go for long stretches without updates 2011-06-06T14:44:06 We can have a countdown timer for them, they don't have to wait to reactivate it, they can click everyday to keep it going. 2011-06-06T14:44:22 but are you skewing the rest of the field by not then playing against them 2011-06-06T14:44:40 okay, that could be a downside... 2011-06-06T14:45:13 we could just take them out of seed rotation, but they could still be matched against 2011-06-06T14:45:26 that might be work out 2011-06-06T14:45:47 In the last contest I saw some bots from september with rank near 300. 2011-06-06T14:46:00 McLeopold: oh yes, have you seen that convo? 2011-06-06T14:46:04 I think any potential minor downside is greatly outweighed by the benefit of removing starter bots from playing 2011-06-06T14:46:39 UncleVasya: I think my bot won the "longest stretch in the top 100" award :p 2011-06-06T14:46:57 amstan: yes 2011-06-06T14:47:05 I kinda like the idea of removing them from being seeds 2011-06-06T14:47:33 It's already an issue, so I do want to do it. I just also want the automatic removal so you have to be active. 2011-06-06T14:48:04 so, back to my original point, draws aren't really happening among non-starter bots. 2011-06-06T14:48:22 I think you are supposed to adjust epsilon, not beta. 2011-06-06T14:48:29 beta is calculated from that 2011-06-06T14:48:54 but the overall rankings parameters should be adjusted for the overall results, not just the high end results 2011-06-06T14:49:20 beta is a measure of how 'random' the game outcome is 2011-06-06T14:49:21 epsilon is supposed to be tuned for the game, not the contestants 2011-06-06T14:49:28 I think there should be a line (for example 15%+) after which bots can be deactivated after some time inactive. 2011-06-06T14:50:02 I mean that first 15% of the rankings will play even if their masters abandoned them. 2011-06-06T14:50:37 *** Accoun has quit IRC () 2011-06-06T14:50:44 UncleVasya: if we take them out of seed rotation, they would still get games against newer bots. 2011-06-06T14:50:52 i think rank should be proportional with their sigma 2011-06-06T14:51:10 amstan: do you mean mu? how? 2011-06-06T14:51:18 given that ants seems to be quite sensitive to initial conditions (why non-symmetric food is a problem) I think beta is going to be fairly high 2011-06-06T14:51:21 is the trueskill implementation using floating point? 2011-06-06T14:51:26 as in.. it doesn't matter if the guy near the end is uncertain 2011-06-06T14:51:51 jmcarthur: it should, it's python 2011-06-06T14:52:03 jmcarthur: yes it is currently at least 2011-06-06T14:52:04 btw, about sigma: is problem with it's lowering already fixed? 2011-06-06T14:52:23 McLeopold: there are other fractional number representations available in python 2011-06-06T14:53:21 janzert: what would be the most acurate type if we don't want rounding errors? 2011-06-06T14:53:34 i don't know what all the intermediate values in trueskill look like, but if we had the time to fiddle with it i would be tempted to look into a fixed point representation 2011-06-06T14:54:03 jmcarthur: I'll see what I can do and see if it then matches the online calculator 2011-06-06T14:54:10 McLeopold: depends on what you want do, anything else will probably be much slower of course 2011-06-06T14:54:23 but i doubt anything like that is built in to python. we'd have to find a library or roll our own. also, to use it well, you need to always be aware of your precision and adjust depending on context 2011-06-06T14:54:29 we want accuracy above speed 2011-06-06T14:54:32 Decimal is probably the logical one for better accuracy 2011-06-06T14:55:20 is Decimal dynamic precision (and is it adjustible or automatic)? 2011-06-06T14:55:32 jmcarthur: http://docs.python.org/library/decimal.html 2011-06-06T14:55:52 user specified precision 2011-06-06T14:56:04 I don't believe there is any automatic adjustment 2011-06-06T14:56:18 I'll max it out and see what happends 2011-06-06T14:57:19 so it's a global precision 2011-06-06T14:57:55 hopefully we don't use any libs that change it then... 2011-06-06T14:58:16 score 1 for functional programming... 2011-06-06T14:58:19 each number has a context specifying the precision among other things 2011-06-06T14:58:28 each *number*? 2011-06-06T14:59:04 ah you can create new contexts 2011-06-06T14:59:09 oops, each thread 2011-06-06T14:59:15 and make it active/inactive 2011-06-06T14:59:20 yes 2011-06-06T14:59:34 *** UncleVasya has quit IRC (Ping timeout: 260 seconds) 2011-06-06T14:59:42 so as long as we always activate our own context whenever the trueskill stuff is entered then we should be fine 2011-06-06T14:59:52 and to play nice we would deactivate it when we are done 2011-06-06T15:00:04 I doubt anything else we use, uses Decimals 2011-06-06T15:00:08 * jmcarthur shrugs 2011-06-06T15:00:39 manager only gets called once for one update, then exits 2011-06-06T15:01:01 that's right 2011-06-06T15:01:08 *** amstan has quit IRC (Ping timeout: 240 seconds) 2011-06-06T15:02:21 janzert: float("inf")? How does that become a decimal? 2011-06-06T15:02:49 nvm 2011-06-06T15:03:12 Decimal("Infinity") should do it I believe 2011-06-06T15:03:21 but why are we starting with an infinity? 2011-06-06T15:03:25 * janzert goes to look 2011-06-06T15:03:41 janzert: line 93 2011-06-06T15:04:00 *** Accoun has joined #aichallenge 2011-06-06T15:04:36 *** UncleVasya has joined #aichallenge 2011-06-06T15:04:46 i good fixed point library will provide operations that adjust the precision to retain as much information as possible. i have one written in haskell that provides operations that give more claimed precision after multiplication and less claimed precision after division, for example. rounding (which by contract lowers precision) is always explicit then and you aren't caught by surprise 2011-06-06T15:05:35 jmcarthur: talk to the IEEE and get it in the decimal specification and it will end in python ;) 2011-06-06T15:05:44 eh 2011-06-06T15:07:28 hm, come on 64bit floating point precision is enough for most applications, an bug is more likely to cause imperfections than precision problems 2011-06-06T15:07:48 it just depends 2011-06-06T15:08:20 the more operations you have, the large the error 2011-06-06T15:08:27 trueskill has quite a few 2011-06-06T15:08:39 i dont think that trueskill ranking involves numerical integration 2011-06-06T15:08:39 a 64 bit binary fixed point number can express any point from the sun to pluto to submicrometer precision 2011-06-06T15:08:53 with uniform precision 2011-06-06T15:09:05 what about alpha centari? 2011-06-06T15:09:07 mcstar: what is sum(0.1 for r in range(10))? 2011-06-06T15:09:19 McLeopold: for submicrometer precision you might need more bits ;) 2011-06-06T15:09:22 haha 2011-06-06T15:09:33 but what is the precision? 2011-06-06T15:09:52 you mean what would the precision be for 64 bits from the sun to AC? 2011-06-06T15:09:56 y 2011-06-06T15:09:57 thats another problem, representation incompatibility 2011-06-06T15:10:09 mcstar: huh? 2011-06-06T15:10:16 to janzert 2011-06-06T15:10:26 or worse sum(0.01 for r in range(100)) - sum(0.1 for r in range(10)) :) 2011-06-06T15:10:49 im sorry to say but it is working 2011-06-06T15:11:03 7.7715611723760958e-16 ? 2011-06-06T15:11:20 janzert: that 0.1 cant be represented as a binary string is a completely different problem 2011-06-06T15:11:31 (finite b s) 2011-06-06T15:11:44 mcstar: about 2.3 millimeters of precision 2011-06-06T15:11:47 oops 2011-06-06T15:11:49 McLeopold: ^^ 2011-06-06T15:11:52 ooooh 2011-06-06T15:11:59 mine is much longer 2011-06-06T15:12:07 oh sry 2011-06-06T15:12:09 ooooooooooh 2011-06-06T15:12:11 ;D 2011-06-06T15:12:48 McLeopold: that is assuming that wolfram alpha is a reasonably accurate calculator and encyclopedia for this 2011-06-06T15:13:20 only if they didn't use floating point.... 2011-06-06T15:13:42 doesn't look like it 2011-06-06T15:13:55 these numbers seem pretty good (just tried it for pluto, too) 2011-06-06T15:14:09 i had already done the math for pluto when developing my fixed point library 2011-06-06T15:14:58 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has joined #aichallenge 2011-06-06T15:19:45 with 80 bits (the number of bits that extended precision floating point gets), fixed point can specify any point between the sun and alpha centauri to within 0.034 micrometers 2011-06-06T15:20:16 jmcarthur:why do you say that "with uniform precision"? it is correct that the sun-pluton distance approxmiated in microns is 6000000000000000000, and a 64 bit integer can hold a value as large as 18446744073709551616, nevertheless, the relatice accuracy of floating point numbers depend on the magnitude i.e. the exponent of the numbers in question 2011-06-06T15:20:27 exactly 2011-06-06T15:20:35 McLeopold: a quicker test than converting the whole trueskill module to use Decimal might be to re-enable the use of scipy and check the results with that 2011-06-06T15:20:38 Chuck Norris can do it with 0.0037 2011-06-06T15:20:40 this is an advantage of fixed point precision over floating point precision 2011-06-06T15:21:03 janzert: never used scipy 2011-06-06T15:21:29 neither have I really 2011-06-06T15:21:50 but looking at the trueskill code it was originally written to use it 2011-06-06T15:22:03 mcstar: well, i say it's an advantage, but it's really more of a tradeoff 2011-06-06T15:22:08 just need to install scipy and uncomment the imports at the top 2011-06-06T15:22:12 *** Frontier has joined #aichallenge 2011-06-06T15:22:24 mcstar: floating point is nice when you *need* to be able to work with potentially unpredictable orders of magnitude 2011-06-06T15:22:39 mcstar: otherwise, fixed point might be nicer 2011-06-06T15:23:16 i dont think you understand me 2011-06-06T15:23:26 im saying that precision depends upon magnitude 2011-06-06T15:23:35 yes 2011-06-06T15:23:41 with floating point 2011-06-06T15:23:51 but you said uniform precision in a previous comment 2011-06-06T15:23:54 yes 2011-06-06T15:23:58 which is misleadind i think 2011-06-06T15:24:01 i was talking about fixed point precision 2011-06-06T15:24:15 ah ok 2011-06-06T15:24:30 because floating point doesn't have that uniformity 2011-06-06T15:24:44 if you know the absolute magnitudes you can use fixed point sure 2011-06-06T15:24:51 it wasnt clear for me what did u mean then 2011-06-06T15:25:02 that is a very common case (knowing your range) 2011-06-06T15:25:27 e.g. it's common to do silly things like representing RGB components as floats in the range 0-1 2011-06-06T15:26:34 well, sometimes you want nan 2011-06-06T15:26:36 sure, you get a decent precision in that range compared to larger ranges, but you could do so much better with fixed point in this case (which people sometimes do without realizing when they represent a component as, say, 0-255 or 0-65535) 2011-06-06T15:26:50 with fixed point, it will cycle back 2011-06-06T15:26:56 or not 2011-06-06T15:27:31 btw i never heard of fixed point representation when i dealt with c 2011-06-06T15:27:49 is there a common one?(implementTION) 2011-06-06T15:27:55 C is one of the nicest languages to use fixed point in because you can take advantage of so many of the properties at a low level 2011-06-06T15:28:02 i have no idea 2011-06-06T15:28:07 i've only ever done it manually 2011-06-06T15:28:34 it requires a fair bit of documentation to keep track of it since C's type system doesn't express everything i would like it to 2011-06-06T15:29:22 s/documentation/comments/ 2011-06-06T15:29:43 actually, its a good idea, using fixed point on the gpu, instead of the slower floating point units 2011-06-06T15:29:59 floating point is not necessarily that slow 2011-06-06T15:30:10 also, it's nice for stuff like HDR 2011-06-06T15:30:11 in many cases we know the ranges of our variables that we simulate 2011-06-06T15:30:30 the main benefit to fixed point on something like a GPU is lower memory requirements 2011-06-06T15:30:45 no, these have 64 bit fpu-s 2011-06-06T15:30:45 why use a float when 8 bits will do, for example 2011-06-06T15:31:07 it has to do with physical simulations 2011-06-06T15:31:09 *** Naktibalda has joined #aichallenge 2011-06-06T15:31:12 ah 2011-06-06T15:31:19 8bit is nowhere near enough 2011-06-06T15:31:19 i was thinking graphics 2011-06-06T15:31:34 8 bits is fine for an RGB component intended for the screen 2011-06-06T15:31:42 actually we had problems with 32bit fp precision previoulsy 2011-06-06T15:32:04 nvidia fpu-s dont comply with the usual ieee standards 2011-06-06T15:32:07 McLeopold: well using scipy, at least for the example in test_truskill.py, makes no difference 2011-06-06T15:32:12 but yeah, if you care about accuracy and know your range, fixed point rocks 2011-06-06T15:32:40 janzert: can you see how it differers from the online calc? 2011-06-06T15:33:33 yeah, at least for that example I see mu differs by as much as 0.2 2011-06-06T15:33:50 did you see the jskills link in the logs? 2011-06-06T15:34:06 https://github.com/nsp/JSkills 2011-06-06T15:34:08 no 2011-06-06T15:34:14 oh, that yes 2011-06-06T15:36:32 *** bhasker has quit IRC (Quit: bhasker) 2011-06-06T15:45:04 jmcarthur: You were talking about C and floating point. Take a look at D. There is a huge interface to the low level functionality: http://www.digitalmars.com/d/2.0/phobos/std_math.html 2011-06-06T15:47:20 McLeopold: btw, I think ghosts need to disappear after some time (or game) limit even if the new version doesn't surpass the previous one (not sure if there are already plans to do that or not) 2011-06-06T15:56:57 Is it difficult to turn off these lines on the sides of the map in visualizer? I mean these blackened pieces of map that represents the other side. 2011-06-06T15:58:05 I wanna turn it off in local visualizer. 2011-06-06T15:58:20 UncleVasya: Do they irritate you? 2011-06-06T15:58:52 I had something like this, but I removed it, because I considered it an unnessecary extra option. 2011-06-06T15:59:08 *** bhasker has joined #aichallenge 2011-06-06T15:59:36 * UncleVasya is going to search word irritate in the vocabulary. 2011-06-06T15:59:55 confuse 2011-06-06T15:59:58 yes 2011-06-06T16:00:24 They are bowl out me much. 2011-06-06T16:00:53 hmm some attacks happen across map boundaries, so it is somewhat important to see the copy of the map 2011-06-06T16:01:38 ok 2011-06-06T16:02:13 would you prefer if there was just white around the map? 2011-06-06T16:02:44 I did it when just downloaded source code :) 2011-06-06T16:02:55 White background and black water 2011-06-06T16:03:00 Actually you are the second person to ask for that, so I may actually add that feature again. 2011-06-06T16:03:12 Haha, ok then :) 2011-06-06T16:04:02 yeeehah, my engine+visualizer is working, ants can eat each other 2011-06-06T16:04:41 mcstar: What amazing features does your engine offer? Is it fast? 2011-06-06T16:04:55 i think it is 2011-06-06T16:05:05 *** bhasker_ has joined #aichallenge 2011-06-06T16:05:05 Can it load bots written in other languages? 2011-06-06T16:05:31 oh no, it cant interface with other bots, i didnt want it to 2011-06-06T16:05:39 Does it help debugging? 2011-06-06T16:05:45 i want to develop my genetic algorithm with it 2011-06-06T16:06:09 Yeah I though of that, but it is too much for me. 2011-06-06T16:06:19 Are we waiting for ground.invaders this time? :D 2011-06-06T16:06:23 what? writing an engine? 2011-06-06T16:06:35 no, doing a genetic algorithm 2011-06-06T16:06:39 ah 2011-06-06T16:06:47 thx, you could have saif it erlier 2011-06-06T16:06:55 but its too late for me now isnt it? 2011-06-06T16:06:57 :D 2011-06-06T16:07:02 said 2011-06-06T16:07:40 *** bhasker has quit IRC (Ping timeout: 260 seconds) 2011-06-06T16:08:12 I'd probably write some general strategies (expand, defend, attack, some other strategy...) and let something like a neural net decide for the right strategy depending on the situation on the map 2011-06-06T16:08:29 done it before? 2011-06-06T16:08:43 (the neural thingy) 2011-06-06T16:09:04 *** amstan has joined #aichallenge 2011-06-06T16:09:04 *** ChanServ sets mode: +o amstan 2011-06-06T16:09:07 no, but I learnt that if you give feedback on the outcome of a strategy to the neural net, it can adjust itself and chose another one next time 2011-06-06T16:09:52 well, that is called learning isnt it? 2011-06-06T16:10:28 yes, so it would be something you can call ai. most of the stuff in planet wars was far away from ai :) 2011-06-06T16:10:55 but I have to work on the visualizer first anyway! 2011-06-06T16:11:09 i dont know, the programmer at least would be able to pass the turing test lol 2011-06-06T16:12:03 i will really try the GA stuff, if it wont work well, or at all, then ill give heuristics a try 2011-06-06T16:12:22 but i have to say it was quite pleasant to write the engine in lisp 2011-06-06T16:12:44 *** olexs has joined #aichallenge 2011-06-06T16:13:17 *** roginn has joined #aichallenge 2011-06-06T16:15:06 *** Naktibalda has quit IRC (Quit: ChatZilla 0.9.87 [Firefox 4.0.1/20110518052719]) 2011-06-06T16:15:17 *** needsch has joined #aichallenge 2011-06-06T16:21:24 Frontier: well i was talking about C and fixed point. floating point is still floating point even if highly configurable 2011-06-06T16:22:26 i like that documentation. very precise 2011-06-06T16:22:58 and i love that D prefers purity even though it's an impure language 2011-06-06T16:23:00 I hate that documentation. Once it goes into classes, structs or templates you get totally lost :p 2011-06-06T16:23:14 well yeah. i just mean this floating point stuff specifically 2011-06-06T16:23:53 is there a pure language? 2011-06-06T16:24:11 how could you write anything meaningful without ever changing the outside world? 2011-06-06T16:24:45 files, console, network, shared memory 2011-06-06T16:33:04 *** Palmik has quit IRC (Remote host closed the connection) 2011-06-06T16:37:21 purity doesn't even technically mean you can't communicate with the outside world, at least not by many definitions 2011-06-06T16:38:53 well 2011-06-06T16:38:57 let me rephrase 2011-06-06T16:39:11 :) 2011-06-06T16:39:55 *purity* has to do with how the meanings of functions are derived from their parts, but doesn't not require that the *meaning* of those functions can't have anything to do with the outside world 2011-06-06T16:40:31 it's perhaps a subtle distinction, especially when looking at currently popular pure languages which aren't as different from imperative programming as they at first seem 2011-06-06T16:41:37 seems I could extend purity to cover pretty much anything going from a definition that includes outside world state though 2011-06-06T16:42:07 janzert: exactly! the problem is that you then have to have a definition for outside world state. it's harder than it appears 2011-06-06T16:42:12 but I'm certainly not a functional programmer by any means ;) 2011-06-06T16:42:15 s/definition/model/ 2011-06-06T16:43:01 its soo nice that when i have a conceptually signifacant bug, the program still manages to run for a time, and appear to work ok, hm 2011-06-06T16:43:21 one of my biggest problems with haskell is that most haskell developers have accepted IO as *the* solution to communicating with the outside world. to me, all it does is give us an imperative DSL. the problem is that it doesn't have a precise model at all. 2011-06-06T16:43:49 so it's only "pure" if you just treat it entirely as a black box. kind of cheating, IMO 2011-06-06T16:44:26 *** delt0r___ has joined #aichallenge 2011-06-06T16:44:38 it's still pure i guess. getLine's meaning is just a big question mark. at least you can combine question marks and get other question marks. 2011-06-06T16:44:55 hehe 2011-06-06T16:45:56 *** delt0r_ has quit IRC (Ping timeout: 250 seconds) 2011-06-06T16:50:45 *sigh*, well I guess I like the definition where a pure function is simply one that doesn't depend on or modify anything outside of its scope. 2011-06-06T16:51:52 although the generalization sounds good on second thought 2011-06-06T16:52:38 well, your definition sounds about the same as mine. the *meaning* of a pure function doesn't depend on or modify anything outside of scope 2011-06-06T16:54:10 because purity can be taken to mean different things, i use "denotative" to describe the kind of thing i'm talking about 2011-06-06T16:54:29 which has a precise meaning 2011-06-06T16:54:33 *** bhasker_ has quit IRC (Quit: bhasker_) 2011-06-06T16:55:27 in a denotative language, expressions can be made up of subexpressions, each subexpression is associated with a meaning (like a mathematical object, a number, a function, etc.), and the meaning of an expression depends only on the meanings of its subexpressions 2011-06-06T16:55:29 *** bhasker has joined #aichallenge 2011-06-06T16:56:14 jmcarthur: havent you discussed this a couple of weeks ago? i have deja vu 2011-06-06T16:56:33 i may have said the same thing here before, but i think it was at least a couple months ago 2011-06-06T17:02:04 *** UncleVasya has quit IRC (Ping timeout: 260 seconds) 2011-06-06T17:06:48 *** mcstar has quit IRC (Quit: WeeChat 0.3.5) 2011-06-06T17:19:45 *** Redgis has joined #aichallenge 2011-06-06T17:32:58 *** roginn has quit IRC (Quit: Page closed) 2011-06-06T17:33:13 *** McLeopold has quit IRC (Read error: Connection reset by peer) 2011-06-06T17:33:36 *** bhasker has quit IRC (Ping timeout: 250 seconds) 2011-06-06T17:41:42 *** Zaphus has joined #aichallenge 2011-06-06T17:42:35 *** bhasker has joined #aichallenge 2011-06-06T17:45:51 *** Mathnerd314 has joined #aichallenge 2011-06-06T17:57:26 *** amstan_ has joined #aichallenge 2011-06-06T17:57:26 *** ChanServ sets mode: +o amstan_ 2011-06-06T18:02:15 *** ltriant has joined #aichallenge 2011-06-06T18:02:30 *** bhasker has quit IRC (Quit: bhasker) 2011-06-06T18:04:56 *** bhasker has joined #aichallenge 2011-06-06T18:05:45 *** amstan_ has quit IRC (Ping timeout: 255 seconds) 2011-06-06T18:29:17 *** sigh has joined #aichallenge 2011-06-06T18:44:18 *** sigh has quit IRC (Remote host closed the connection) 2011-06-06T18:48:19 *** Redgis has quit IRC (Quit: ... mains libres) 2011-06-06T18:49:29 *** Zaphus has quit IRC (Quit: Page closed) 2011-06-06T18:51:14 *** xQuasar has joined #aichallenge 2011-06-06T18:51:31 If 'r' has been removed from the game, I think you guys should update this page: http://aichallengebeta.hypertriangle.com/specification.php 2011-06-06T18:54:18 *** Mathnerd314 has quit IRC (Disconnected by services) 2011-06-06T18:54:59 *** xQuasar has quit IRC () 2011-06-06T18:55:10 *** Mathnerd314_ has joined #aichallenge 2011-06-06T18:56:11 *** Mathnerd314_ is now known as Mathnerd314 2011-06-06T18:57:44 *** okayzed is now known as okay 2011-06-06T19:20:14 *** needsch has quit IRC (Quit: Leaving.) 2011-06-06T19:24:35 *** Kingpin13 has quit IRC (Quit: time to go) 2011-06-06T19:37:20 *** batukan has joined #aichallenge 2011-06-06T19:39:09 *** batukan has quit IRC (Quit: quiting) 2011-06-06T19:54:50 *** bhasker has quit IRC (Ping timeout: 248 seconds) 2011-06-06T19:58:48 *** bhasker has joined #aichallenge 2011-06-06T20:09:22 *** olexs has quit IRC (Quit: Leaving.) 2011-06-06T20:34:42 *** Frontier has quit IRC (Ping timeout: 255 seconds) 2011-06-06T21:03:13 *** bhasker has quit IRC (Quit: bhasker) 2011-06-06T21:05:09 *** FireFly has quit IRC (Quit: swatted to death) 2011-06-06T22:02:54 @later tell xQuasar I've updated the specs page to remove the 'r' line stuff on github, it'll show up on the beta site specs next time amstan "redoes" stuff 2011-06-06T22:02:54 antimatroid: Ready to serve. 2011-06-06T22:08:05 *** Eruonen has quit IRC () 2011-06-06T22:10:00 *** bhasker has joined #aichallenge 2011-06-06T22:34:23 *** ltriant has quit IRC (Read error: Connection reset by peer) 2011-06-06T22:35:24 *** ltriant has joined #aichallenge 2011-06-06T22:44:23 *** computerwiz_222 has joined #aichallenge 2011-06-06T22:44:28 *** bhasker has quit IRC (Quit: bhasker) 2011-06-06T22:45:32 *** qacek has joined #aichallenge 2011-06-06T22:45:33 *** ltriant has quit IRC (Read error: Connection reset by peer) 2011-06-06T22:45:41 *** ltriant has joined #aichallenge 2011-06-06T22:47:11 *** ltriant has quit IRC (Read error: Connection reset by peer) 2011-06-06T23:06:21 *** Mathnerd314 has quit IRC (Disconnected by services) 2011-06-06T23:07:33 *** Mathnerd314_ has joined #aichallenge 2011-06-06T23:08:20 *** Mathnerd314_ is now known as Mathnerd314 2011-06-06T23:16:06 *** Mathnerd314 has quit IRC (Ping timeout: 240 seconds) 2011-06-06T23:19:41 *** okay is now known as okayzed 2011-06-06T23:30:31 *** ltriant has joined #aichallenge 2011-06-06T23:52:29 *** computerwiz_222 has quit IRC (Ping timeout: 276 seconds) 2011-06-06T23:57:08 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has quit IRC (Quit: Lost terminal)