2011-06-30T00:09:53 *** McLeopold has quit IRC (Quit: Leaving.) 2011-06-30T00:29:23 *** ltriant has quit IRC (Ping timeout: 240 seconds) 2011-06-30T00:30:06 *** ltriant has joined #aichallenge 2011-06-30T00:46:36 contestbot: later tell McLeopold one of the workers stopped because it received a blank task http://pastebin.com/GpNPuQdq not sure if you were working on something at the time or if there's a bug lurking somewhere 2011-06-30T00:46:36 janzert1: Job's done. 2011-06-30T00:46:53 *** janzert1 is now known as janzert 2011-06-30T00:55:31 *** bhasker has joined #aichallenge 2011-06-30T00:57:37 *** UncleVasya has joined #aichallenge 2011-06-30T01:06:01 *** mcstar has joined #aichallenge 2011-06-30T01:10:15 mcstar: what if bot input had "a row col player directionToPreviousSquare" 2011-06-30T01:10:31 where direction to previous square is {N,E,S,W} 2011-06-30T01:10:37 wait, 5 4 2011-06-30T01:10:40 3 2 1 2011-06-30T01:10:44 ok, eyes open 2011-06-30T01:10:50 erm, {N,E,S,W,X|* 2011-06-30T01:10:56 where X is the ant didn't move 2011-06-30T01:11:02 C 2011-06-30T01:11:08 as inc centre 2011-06-30T01:11:13 ? 2011-06-30T01:11:26 so you agree? 2011-06-30T01:11:43 i dunno, i kind of liked the idea of having to try and match dead ants to old enemy ants 2011-06-30T01:11:46 but i don;t really mind 2011-06-30T01:12:07 old bots should still work fine 2011-06-30T01:12:26 as they are meant to ignore input they don't understand 2011-06-30T01:12:47 in the meantime i thought about it, and if theres a discrepancy, the ant which takes a place of another for some reason could just assume the role of that one 2011-06-30T01:13:28 so, it might not be too hard to live without it, but i think it would still make life easier 2011-06-30T01:13:35 *** carlop has quit IRC (Read error: Operation timed out) 2011-06-30T01:13:59 the engine can't do everything for you :P 2011-06-30T01:14:07 *** carlop has joined #aichallenge 2011-06-30T01:14:21 we will see.... 2011-06-30T01:14:28 :) 2011-06-30T01:14:47 you should implement an s line 2011-06-30T01:14:58 as in startegy which gives the next best move of an ant 2011-06-30T01:17:42 this coffee needed more sugar 2011-06-30T01:30:34 antimatroid1: you said your on buntu now 2011-06-30T01:30:38 do you have gnome 3? 2011-06-30T01:39:15 *** mcstar has quit IRC (Ping timeout: 246 seconds) 2011-06-30T01:41:55 *** mcstar has joined #aichallenge 2011-06-30T01:42:08 power was interrupted 2011-06-30T01:46:00 strictly speaking where an ant was the previous move is completely irrelevant as it makes no difference for the current or future state 2011-06-30T01:49:02 janzert: it would aid the tracking of them 2011-06-30T01:50:42 and I mean tracking them is completely irrelevant since the past state doesn't have any ongoing effect :) 2011-06-30T01:51:32 not true: startegies dont change turn by turn 2011-06-30T01:51:37 ra 2011-06-30T01:52:16 it depends on the bot whether it has memory of past states or not 2011-06-30T01:52:24 but whether an ant got to a position by stepping there from the north or south or simply having spawned there makes no difference whatsoever the same actions are available to it 2011-06-30T01:52:31 thats a different story that the engine doesnt need to know the past 2011-06-30T01:53:07 yes, thats true, thats why i said before that it wouldnt be hard to implement tracking 2011-06-30T01:53:10 I'm saying from a game theoretic stance keeping track of past states is useless 2011-06-30T01:53:33 a bot of course may want to do some incremental updating of its internal state 2011-06-30T01:53:41 well, from an implementation point of view its not useless 2011-06-30T01:54:06 *** stocha has joined #aichallenge 2011-06-30T01:54:13 but I think it should simply be smart enough to do that by looking at the present state 2011-06-30T01:54:28 present state of the game that is 2011-06-30T01:54:43 why give a bonus to the last botst that has not timed out, when it has utterly been defeated by all other bots ? http://aichallengebeta.hypertriangle.com/visualizer.php?game=45689&user=107 2011-06-30T01:54:46 ok, but it would be easier if we could just match up past ants with present ones 2011-06-30T01:55:01 need to move less data around, not destroy some structures... 2011-06-30T01:55:39 stocha: I would personally rather that any timed out bot is set to last place 2011-06-30T01:56:10 janzert: okay. But still why give a bonus to the loosing bot ? 2011-06-30T01:56:11 i.e. doesn't get any points at all 2011-06-30T01:56:23 it was the winning bot not the losing bot 2011-06-30T01:56:30 go tighten them if their loose 2011-06-30T01:56:34 like he has 0 point, and no chance of getting much more (like the game i linked). But he get the 1000+ bonus and sudenly ends up first ? 2011-06-30T01:56:39 by virtue of playing correctly for longer than any other bot 2011-06-30T01:56:57 first is the position he should be in 2011-06-30T01:57:02 janzert: it certainly isn't the case of the linked game 2011-06-30T01:57:23 btw, I don't agree much with the scoring method so I'm the wrong one to defend it ;) 2011-06-30T01:57:27 he should be last. He has no chance of having more point than the other bots in the game, with the strategy it use. 2011-06-30T01:57:38 janzert: ah okay. 2011-06-30T01:57:59 all other bots died before him so he should be first, that part I agree with 2011-06-30T01:58:01 janzert: what is the state of the concensus with the scoring method then ? 2011-06-30T01:58:15 that we're probably not going to change it? 2011-06-30T01:58:24 stocha: your problem is that stocha won? 2011-06-30T01:59:05 janzert: although i'm not to defend to be that harsh with timeout bot. I think the ant should lay still (big thing already). They you added the it gets no more score. That's enough for me. Why make it autoloose. 2011-06-30T01:59:19 mcstar: i'm just discussing the scoring method. 2011-06-30T01:59:25 actually I would fight much more strongly against the current scoring method if stocha hadn't come first in that game ;) 2011-06-30T01:59:47 mcstar: it happens that i look more at my bots games, so they will often be participating in the example i give :) 2011-06-30T02:00:15 janzert: there need to be a serious discussion on the scoring then. 2011-06-30T02:00:21 actually you have a point 2011-06-30T02:00:36 stocha: rather late for that 2011-06-30T02:00:41 janzert: i'm happy with the scoring in itself. Trying to have a bonus for bots that do combat. 2011-06-30T02:01:00 the scoring assumes that your bot could have eaten all of sir macelon's after it timed out 2011-06-30T02:01:00 janzert: although there should be more points for killing than grabing food then. 2011-06-30T02:01:02 the serious scoring discussion was months ago 2011-06-30T02:01:17 janzert: the difference is that now there are data. 2011-06-30T02:01:28 and so far as I'm concerned it's working good enough 2011-06-30T02:01:35 so don't touch it 2011-06-30T02:01:36 month ago, i would have serious doubt on the current scoring. 2011-06-30T02:01:50 we need to get other things finished off and get the contest going 2011-06-30T02:02:00 +1 2011-06-30T02:02:09 But, i don't like this timeout make the last no time out automatically win. This is not coherent with the scoring. 2011-06-30T02:02:57 janzert: incoherent scoring will get some people endlessly arguing. So if it is easy to settle logically it should be done. 2011-06-30T02:03:17 you can console yourself that it pretty certainly won't effect the top rankings at all 2011-06-30T02:03:21 What i ask for, is that the game still plays out when only one bot is remaining without timing out. 2011-06-30T02:03:30 and that the score follow the usual logic. 2011-06-30T02:03:35 we can go back to a timeout/crash removes you from the contest ;) 2011-06-30T02:03:48 *** carlop has quit IRC (Quit: ChatZilla 0.9.87 [Firefox 3.6.12/20101027124959]) 2011-06-30T02:03:52 :S 2011-06-30T02:04:02 janzert: that's insane. Because the timeout might be to blame on the contest, not the bot. 2011-06-30T02:05:00 janzert: who am i to convince, if not you ? 2011-06-30T02:05:09 personally I think timeout basically removing you from the current game is good enough though removing from the rest of the contest is too harsh :) 2011-06-30T02:05:11 janzert: i though you was the big influencial guy around. 2011-06-30T02:05:26 well I'll pretty much oppose score changes at this point 2011-06-30T02:05:53 and really oppose anything that lessens the last surviving bots chance of being the overall winner 2011-06-30T02:05:56 janzert: what i don't like is : bot B get 100 score. Bot c get 200 score. Bot d get 1 score (but still there). Bot e get 10000 score, but times out 2011-06-30T02:06:12 bot d wins over B and C. although it have no chance of catching up. 2011-06-30T02:06:30 that is not fair for B and C, that won by the rules. But the tiemout screw it up. 2011-06-30T02:06:39 being last survivor should trump everything else so far as I'm concerned 2011-06-30T02:07:01 the bot has shown itself to be better than all the rest 2011-06-30T02:07:26 janzert: yeah, but either you go for last survivor win, or for best fighter win. But not with best fighter win unless a strong bot timeout in whatever case that is last remaining bot win 2011-06-30T02:07:42 and I'm opposed to best fighter win 2011-06-30T02:07:43 I want it to be coherent, not schizophrenic scoring. 2011-06-30T02:07:58 janzert: yet you said you were happy with the scoring. 2011-06-30T02:08:07 no, I said I don't want it changed 2011-06-30T02:08:10 janzert: don't you see the need for coherence in there ? 2011-06-30T02:08:12 and it works good enough 2011-06-30T02:08:25 nope 2011-06-30T02:08:43 janzert: when there is no time out yes. When there are, especially if it's a strong bot, things are screwed. And coward bots get a bit bonus. 2011-06-30T02:08:44 coherence in scoring is a really minor concern for me 2011-06-30T02:09:02 (wich is the case of the stocha bot here, with sir_macelon timing out very often.) 2011-06-30T02:09:17 janzert: whom am i to speak with then ? 2011-06-30T02:10:30 to get it changed? like I said I'm actively opposed to changing it, so in order to change it you'd have to convince amstan and mcleopold to vote against me I suppose 2011-06-30T02:11:11 mcstar: i agree with you in that i think a lot of people will want to track their ants. But changing the communication now seems a bit unwise. 2011-06-30T02:11:45 mcstar: so i didn't tryed to assess if you proposal was a good way to help people track their ant or not. 2011-06-30T02:12:09 assess it then :) 2011-06-30T02:13:02 as janzert said, it shouldnt cause problems for existing bots 2011-06-30T02:13:20 mine ignores superfluous arguments too 2011-06-30T02:13:21 janzert: wrt tracking ant deaths 2011-06-30T02:13:40 if you propogate information about where an enemy ant might be based on where you last saw a live one, you can not propogate information from that ant 2011-06-30T02:13:42 @later tell amstan there is an issue with time out that makes the scoring that is supposed to support the best fighting not working. bot B get 100 score. Bot c get 200 score. Bot d get 1 score (but still there). Bot e get 10000 score, but times out. In my view, when only one bot remain by timeout effect (especially if it lost against other nottimeout eliminated bots) the game should play out. The the scoring stay coherent. 2011-06-30T02:13:42 stocha: As you wish. 2011-06-30T02:13:52 which if you're hunting down an enemy in a maze could be quite helpful no? 2011-06-30T02:14:24 antimatroid1: who was talking about tracking ant deaths? ;) 2011-06-30T02:14:42 ignore me then 2011-06-30T02:14:48 I haven't put much thought into that at all 2011-06-30T02:14:49 oh wait, ant moves? 2011-06-30T02:14:53 mcstar: i don't like the idea that the protocol change. A lot of beta participant has already buit the ant tracking system anyway. 2011-06-30T02:15:29 *** okayzed is now known as okay 2011-06-30T02:15:35 stocha: your bot doesn't need to change 2011-06-30T02:15:38 antimatroid1: yes, the idea was to give where an ant moved from 2011-06-30T02:15:44 janzert: why is he pinging me at this ungodly hour? lol 2011-06-30T02:15:47 wouldn;t that include dead ants? 2011-06-30T02:16:14 I don't think so 2011-06-30T02:16:25 a dead ant no longer exists in the game state right 2011-06-30T02:16:34 only the still living ants of yours would get the info of where teyre coming from 2011-06-30T02:16:36 no, but you receive information if you see an ant die 2011-06-30T02:16:48 stocha: idk, i can't concentrate about this so late 2011-06-30T02:17:03 stocha: and i'm not really into the ant logistics 2011-06-30T02:17:20 i would have expected the information to be coming under the idea that you can see the movement of stuff in view 2011-06-30T02:17:22 stocha: i would go with what antimatroid things 2011-06-30T02:17:23 so we'll give you that information 2011-06-30T02:17:25 thinks* 2011-06-30T02:17:41 amstan: cause he doesn't like my stance on the topic ;) 2011-06-30T02:17:53 amstan: he hasnt given his oppinion to stocha's problem :D 2011-06-30T02:18:03 janzert: he doesn't like a lot of things though 2011-06-30T02:18:07 * antimatroid1 is lost 2011-06-30T02:18:42 amstan: sorry, I let slip that to overule me he has to convince you and mcleopold :P 2011-06-30T02:18:53 oh np, i'm just kidding around 2011-06-30T02:19:13 @later tell McLeopold there is an issue with time out that makes the scoring that is supposed to support the best fighting not working. bot B get 100 score. Bot c get 200 score. Bot d get 1 score (but still there). Bot e get 10000 score, but times out. In my view, when only one bot remain by timeout effect (especially if it lost against other nottimeout eliminated bots) the game should play out. The the scoring stay coherent. 2011-06-30T02:19:13 stocha: Aye, aye, sir 2011-06-30T02:19:28 @later tell McLeopold so you are my last hope ... 2011-06-30T02:19:28 stocha: As you wish. 2011-06-30T02:19:36 stocha: i'm not exactly sure what you mean though 2011-06-30T02:19:37 hehe 2011-06-30T02:20:06 stocha: are you talking about the scoring of a crashed/timed out bot? i think they should act like they're still part of the game but not sending any moves 2011-06-30T02:20:44 he doesn't like that the last surviving ant is likely to win if all the other bots timeout 2011-06-30T02:20:51 amstan: http://aichallengebeta.hypertriangle.com/visualizer.php?game=45689&user=107 Stocha should loose here. There is a lot of game like this with coward but wich go in a tiny part and do nothing. Then remain last, and the winning bot timeout. 2011-06-30T02:20:55 at least that's my understanding 2011-06-30T02:21:06 janzert: not all other bot timeout !! 2011-06-30T02:21:22 what, that ant gets the food bonus yes? 2011-06-30T02:21:23 if all the other remaining ants timeout 2011-06-30T02:21:24 janzert: all other bot are eliminated yet. (with big score). And only the winning bot timeout. 2011-06-30T02:21:25 what's wrong with that? 2011-06-30T02:21:52 janzert: then the loosing bot sudendly becomes the winner. Although there is no chance of that but for the winning bot to timeout. 2011-06-30T02:22:04 stocha: 2 contests ago timing out got you disqualified from the contest 2011-06-30T02:22:14 antimatroid1: it make the scoring system not coherent when the best best bot time out a few turn before wining. 2011-06-30T02:22:28 the other bot should've been careful not to timeout then :} 2011-06-30T02:22:31 yes, but since the game can continue if someone times out, i think it's only fair that their ants just stay frozen 2011-06-30T02:22:49 i think stocha doesnt realize that somebody must win 2011-06-30T02:22:52 antimatroid: The problem is that either you should go with last bot win. Or with biggest scorte win. In that last case, don't make incoherent when best bot timeout. 2011-06-30T02:22:54 amstan: yep, I don't think that is at issue at all 2011-06-30T02:22:55 amstan: the only reason the ants are left there is to try and not make the rest of the game too unfair 2011-06-30T02:23:19 stocha: it clearly wasn't the best bot in that game if it times out 2011-06-30T02:23:29 janzert: the other bot didn't time out !!!!!!!!! 2011-06-30T02:23:44 janzert: they were ELIMINATED with BIG SCORES. 2011-06-30T02:24:03 so they should go 2d, 3rd 4th etc. And the loosing bot should be LAST in any cases. 2011-06-30T02:24:06 omg, you said it they were ELIMINATED 2011-06-30T02:24:16 the other bot is sir_macelon in this case, not the rest of the field 2011-06-30T02:24:23 here it receives the big bonus and wins. 2011-06-30T02:24:29 i'm still unsure on what the problem is 2011-06-30T02:24:39 janzert: janzert the other bot was winning over stocha in this game 2011-06-30T02:24:48 janzert: there was no way for stocha to catch up. 2011-06-30T02:24:52 anyway I'm need to get to bed 2011-06-30T02:25:04 Then this game is not the best example. But it happens a lot. 2011-06-30T02:25:46 and like I said my preference would actually be that timed out bots got no points awarded 2011-06-30T02:26:02 but I also don't want to change anything at this point 2011-06-30T02:26:17 with the scoring that is 2011-06-30T02:26:52 but the problem is not with time out bot. it's with the others bot. Especially that the loosing one get the bonus for winning. When it has not. 2011-06-30T02:28:59 i guess i'l have to live with it, unless McLeopold somewhat sees the point. 2011-06-30T02:29:05 *** stocha has quit IRC (Quit: Page closed) 2011-06-30T02:31:23 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has quit IRC (Quit: Lost terminal) 2011-06-30T02:32:13 did stocha expect that eliminated bots get a share of the bonus? 2011-06-30T02:32:27 i call that incoherent with his words 2011-06-30T02:33:38 antimatroid1: im going to implement a* as well, then maybe ill use my own too if it works not too bad 2011-06-30T02:34:03 mcstar: that reminds me, i need to work out if my heuristic function is monotonic or whatever 2011-06-30T02:34:15 i have a feeling it's not, which means i don't think my implementation does enough 2011-06-30T02:34:31 isnt it the simple (manhattan?) distance? 2011-06-30T02:35:04 if i have a set of targets, then it's min_T mdist(x, T) where T are target locations 2011-06-30T02:35:04 i read thats totally ok 2011-06-30T02:35:24 and this is for h(x ) 2011-06-30T02:36:38 *** Palmik has joined #aichallenge 2011-06-30T02:41:52 The h(x) part of the f(x) function must be an admissible heuristic; that is, it must not overestimate the distance to the goal. Thus, for an application like routing, h(x) might represent the straight-line distance to the goal, since that is physically the smallest possible distance between any two points or nodes. 2011-06-30T02:42:22 with your h, the distance cant be less than that, so its ok 2011-06-30T02:42:42 yes it's admissible 2011-06-30T02:42:52 the question i have is is it monotone 2011-06-30T02:42:54 i thin it is 2011-06-30T02:43:20 i'm curious of an example of an admissible nonmonotone heuristic function 2011-06-30T02:43:40 monotone in what? in the distance? since it is the distance it is monotone in itself 2011-06-30T02:44:00 just scale it up 2011-06-30T02:44:14 it says you need h(x) <= d(x,y) + h(y) 2011-06-30T02:44:16 c*d_min 2011-06-30T02:44:24 but i don't see how you can have an admissible function without that 2011-06-30T02:44:41 if your metric is all good with a metric space then you should be fine from the triangle inequality 2011-06-30T02:45:03 this is with multiple targets 2011-06-30T02:45:11 it might matter in what direction you travel 2011-06-30T02:45:19 but i don't understand what a non-monotone admissible heuristic would be? 2011-06-30T02:45:31 consider directed graph with different weight on edges travelling opposite directions ? 2011-06-30T02:45:48 ah yeah possibly 2011-06-30T02:45:52 negative edge weights too? 2011-06-30T02:45:55 or whatever 2011-06-30T02:46:03 maybe, not sure about that 2011-06-30T02:46:08 i'm not very knowledgeable on graph theory 2011-06-30T02:46:29 i think the problem doesnt change if you give a constant to all edge costs 2011-06-30T02:47:10 when i'm searching from food, i will add ants "locations" to the search queue, with g(x) = time they are available 2011-06-30T02:47:34 so if i collect food, the food square will become one of my sources for collecting more food, and g(x) := the turn the ant will be ready to start moving 2011-06-30T02:47:52 for* food 2011-06-30T02:48:44 hehe 2011-06-30T02:48:53 why not just use a meta a* search? 2011-06-30T02:49:00 one a* for pathfinding 2011-06-30T02:49:04 meta? 2011-06-30T02:49:11 another for searching among foods 2011-06-30T02:49:27 get what i mean? 2011-06-30T02:49:27 nah, my method is cool :) 2011-06-30T02:49:44 http://en.wikipedia.org/wiki/File:Astar_progress_animation.gif 2011-06-30T02:49:49 i'm a bit iffy about that animation 2011-06-30T02:49:59 it seems like they're doing a poor job of sorting the search queue 2011-06-30T02:50:05 you could minimise f and then h on a draw 2011-06-30T02:50:28 otherwise you search more than is necessary 2011-06-30T02:51:06 like, it searches way more than neccessary 2011-06-30T02:51:09 do you see what i mean? 2011-06-30T02:51:17 on the image the path can move diagonally 2011-06-30T02:51:23 no not really 2011-06-30T02:51:27 yes, doesn't matter 2011-06-30T02:51:58 it searches from the closest square to the source with equal f value, because that was the first one that was added to the search queue 2011-06-30T02:51:59 do you think on the image it touches too many nodes? 2011-06-30T02:52:21 but you should really search from the equal f valued location with the closest estimate to a target 2011-06-30T02:52:34 yes, more than necessary, but from the way it orders the search queue 2011-06-30T02:52:52 you dont ,like the way the open set i sgenerated? 2011-06-30T02:53:06 i don't like the way the priority queue is ordered 2011-06-30T02:53:44 did you visualize your search algorithm? 2011-06-30T02:54:15 (i think) it's ordered based off min f, where if f(x) == f(y), x you can easily sort the open set with the heuristic function 2011-06-30T02:55:27 i don't use open/closed set 2011-06-30T02:55:32 you don't need to if it's monotone 2011-06-30T02:55:48 just add search locations to the priority queue and mark it as visited 2011-06-30T02:55:51 then that's your search queue 2011-06-30T02:56:19 well, its the same 2011-06-30T02:56:33 yeah but this way is way simpler 2011-06-30T02:56:52 from an implementation point of view 2011-06-30T02:57:03 yeah, and thinking about it 2011-06-30T02:57:04 nevertheless, i will be more knowledgeable after i implement it 2011-06-30T02:57:23 my implementation is the same as i would do a bfs, except i order my queue 2011-06-30T02:57:29 so did you visualize what your alg. is doing? 2011-06-30T02:57:43 i debug the search path 2011-06-30T02:57:46 that's about it 2011-06-30T02:58:52 *** okay is now known as okayzed 2011-06-30T03:05:08 *** Accoun has joined #aichallenge 2011-06-30T03:19:55 *** aerique has joined #aichallenge 2011-06-30T03:22:01 a* with multiple targets is b* 2011-06-30T03:29:10 *** ltriant has quit IRC (Quit: Computer has gone to sleep) 2011-06-30T03:54:57 *** sigh has joined #aichallenge 2011-06-30T03:56:30 *** eashoka has joined #aichallenge 2011-06-30T03:59:34 *** sYnfo has joined #aichallenge 2011-06-30T04:01:08 *** boegel has joined #aichallenge 2011-06-30T04:02:33 *** Naktibalda has joined #aichallenge 2011-06-30T04:10:17 *** locutus2 has joined #aichallenge 2011-06-30T04:28:29 anybody knows when the gaming server will be back ? 2011-06-30T04:31:57 soon as the ec2 spot price drops again ;) 2011-06-30T04:32:07 hopefully within the hour 2011-06-30T04:36:09 tx, jancert :-) 2011-06-30T04:37:24 janzcert : can I ask another one.. when wil the competion proper is likely to start ? and for how long this time..... 2011-06-30T04:40:52 mcstar: it does seem to be 2011-06-30T04:41:01 is the actual algorithm the way i've done it? 2011-06-30T04:43:14 i guess no, but im not sure 2011-06-30T04:45:59 wikipedia page doesn't really describe it very well imo 2011-06-30T04:49:57 a1k0n_: do you know any examples of an admissible heuristic function which is not monotone? 2011-06-30T04:50:41 *** stocha has joined #aichallenge 2011-06-30T04:51:10 *** amstan has quit IRC (Ping timeout: 260 seconds) 2011-06-30T04:52:46 mcstar: i'm not confused by the a* wiki page 2011-06-30T04:53:18 mcstar: what i want it that a bot isn't given any bonus score, unless only his ants remains. 2011-06-30T04:53:27 "If a closed set is used, then h must also be monotonic (or consistent) for A* to be optimal", i would have thought that was if we don't use a closed set? 2011-06-30T04:53:58 stocha: so we should leave the game running and let the bot try and collct all the timet out ants? 2011-06-30T04:54:02 seems rather silly 2011-06-30T04:54:15 now* confused :P aha 2011-06-30T04:54:30 antimatroid1: at least until it catch up with the others (non timed out eliminated ants) 2011-06-30T04:54:50 antimatroid1: it's not more silly that what the actual system is doing. 2011-06-30T04:55:05 antimatroid1: the other option is to go for reverse elimination order for the rank. 2011-06-30T04:55:15 (which i belive janzert supports) 2011-06-30T04:56:59 mcstar: the part on the a* page about implementation details covers my concern from before i think 2011-06-30T04:57:04 stocha: timed-out or eliminated bots should get NO bonuses, thats my oppinion 2011-06-30T04:57:22 mcstar: yeah. Who disagree with that ? 2011-06-30T04:57:38 and it seems fair that the only left player gets it 2011-06-30T04:58:01 maybe this is where we differ 2011-06-30T04:58:29 mcstar: that's the point i disagree. If the last remaing player has 1 ant from the start of the game, and timedout has 5000. And the three eliminated had up to 1000 each, a few turn back. 2011-06-30T04:59:05 mcstar: unless you go with reverse elimination order 2011-06-30T04:59:27 Otherwise depending on when to bot timeout, the ranking is all different. That bothers me a lot. 2011-06-30T04:59:53 look, when the contest starts bot will not time out because of exhausted resources, or some llike that 2011-06-30T05:00:12 only because they are badly written, or they dont use timers 2011-06-30T05:00:32 mcstar: what timing technology will be used ? 2011-06-30T05:00:39 mcstar: on the server parts ? 2011-06-30T05:00:42 whatever you use in your program 2011-06-30T05:01:02 i mean timing out or crashing is unforgivable 2011-06-30T05:01:11 mcstar: what prevent that a deamon server thread just don't give any time to the bot, so it times out ? 2011-06-30T05:01:56 im not the man to ask to question from 2011-06-30T05:02:08 to->those 2011-06-30T05:02:08 mcstar: if bot time out is very rare (which hopefully will be) then there is little problem to be fair. 2011-06-30T05:02:16 *** antimatroid has joined #aichallenge 2011-06-30T05:02:37 mcstar: don't assume that the time out will always be to blame on the bot, unless the server technology can garanty that. 2011-06-30T05:02:56 but thats a different problem 2011-06-30T05:03:07 tell them to guarantee fair timeslices 2011-06-30T05:03:16 mcstar: if the server make the bots times out randomly, then the incoherency will be problematic. 2011-06-30T05:03:27 mcstar: as far as i know, you can't. 2011-06-30T05:03:42 *** eashoka has quit IRC (Quit: Page closed) 2011-06-30T05:03:43 *** antimatroid1 has quit IRC (Ping timeout: 240 seconds) 2011-06-30T05:03:53 mcstar: although the probability that things go real crazy can be very low if things are done well. 2011-06-30T05:03:54 lets hope it wont happen, but i dont think that because of this you should change the scoring system 2011-06-30T05:04:21 I amazed as how well i'm able not to make my point clear :) 2011-06-30T05:04:42 let's stop this, and state that i'm simply not suited for explaining my own opinion. 2011-06-30T05:05:21 stocha: i think i understand you, but i disagree on this, ok? 2011-06-30T05:05:26 My wife always tell me : don't talke. 2011-06-30T05:05:48 but she says that in bed... 2011-06-30T05:05:53 mcstar: yeah, but the point is, we don't seem to disagree. So i think i explained the thing wrong. 2011-06-30T05:06:03 mcstar: no she says that whenever i say something. 2011-06-30T05:06:11 unless it's funny. Maybe. 2011-06-30T05:06:18 sorry to hear that :D 2011-06-30T05:06:39 mcstar: at a point in my life, i din't speak much (if at all). It was not that bad. 2011-06-30T05:06:55 maybe i was more clear, in a lot ways :) 2011-06-30T05:08:22 Most probably, in any case, the impact of what bothers me will be limited. Game mecanics wise. What worries me is that some people will complains how unfair the thing is. And they will be right (for one or two games over 100 ) 2011-06-30T05:08:34 Although it would certainly be a minor issue to solve the case. 2011-06-30T05:09:02 *** Kingpin13 has joined #aichallenge 2011-06-30T05:11:11 *** locutus2 has quit IRC (Ping timeout: 240 seconds) 2011-06-30T05:12:58 *** locutus2 has joined #aichallenge 2011-06-30T05:28:50 *** |UncleVasya| <|UncleVasya|!kvirc@178-133-58-157.dialup.umc.net.ua> has joined #aichallenge 2011-06-30T05:37:26 *** |UncleVasya| <|UncleVasya|!kvirc@178-133-58-157.dialup.umc.net.ua> has quit IRC (Ping timeout: 260 seconds) 2011-06-30T05:44:40 *** UncleVasya has joined #aichallenge 2011-06-30T05:49:02 *** stocha has quit IRC (Quit: Page closed) 2011-06-30T05:50:50 *** InsaneMalkavian has left #aichallenge 2011-06-30T05:50:51 *** UncleVasya has quit IRC (Ping timeout: 260 seconds) 2011-06-30T05:51:09 *** foRei has joined #aichallenge 2011-06-30T05:52:14 *** stocha has joined #aichallenge 2011-06-30T05:53:45 @later tell McLeopold, i think there are now 3 map generators. RandomWalk, Octogonal and Maze. Why not close the map maping, and simply test with those 3 ? (Though it's possible to upgrade them, having more will make beta testing the map generator much more difficult i think) 2011-06-30T05:53:45 stocha: Job's done. 2011-06-30T05:53:53 *** stocha has quit IRC (Client Quit) 2011-06-30T06:03:56 my A* is very fast: can produce 2 MB of buggy printouts in a second 2011-06-30T06:13:48 ok, after correcting a simple indexing error, it worked 2011-06-30T06:13:55 first write 2011-06-30T06:14:07 *** olexs has joined #aichallenge 2011-06-30T06:15:26 *** locutus2 has quit IRC (Ping timeout: 252 seconds) 2011-06-30T06:35:19 *** locutus2 has joined #aichallenge 2011-06-30T06:48:57 *** sYnfo has quit IRC (Quit: Page closed) 2011-06-30T06:49:03 mcstar: :) 2011-06-30T06:49:08 told you a* is worth it 2011-06-30T06:50:30 *** olexs has quit IRC (Quit: Leaving.) 2011-06-30T07:07:02 *** gobang has joined #aichallenge 2011-06-30T07:37:07 http://i.imgur.com/3XxRW.png 2011-06-30T07:37:21 creating this image in mathematica took more time than writing the algorithm 2011-06-30T07:37:37 it seems i unlearned some of its quirks 2011-06-30T07:38:55 antimatroid: but i have to work on it, because it not very fast 2011-06-30T07:38:59 s 2011-06-30T07:50:14 *** sYnfo has joined #aichallenge 2011-06-30T07:53:46 mcstar: what is the picture? 2011-06-30T07:54:09 did you see it? 2011-06-30T07:55:05 antimatroid: 2011-06-30T07:55:14 yeah it's a path 2011-06-30T07:55:16 from what to what/ 2011-06-30T07:55:25 from upper left to lower right 2011-06-30T07:55:44 red is obstacle 2011-06-30T07:55:50 yep okay 2011-06-30T07:56:10 you should also highlight two other things 2011-06-30T07:56:20 locations added to the search queue and locations taken off the search queue as well 2011-06-30T07:56:39 i will try to make a movie out of it 2011-06-30T07:56:49 that would be sweet :) 2011-06-30T07:56:50 when i put it in my visualization framework 2011-06-30T07:56:56 i would like to see a movie of one with multiple targets 2011-06-30T07:57:03 but i soon have to go somewhere 2011-06-30T07:57:14 thats another story 2011-06-30T07:57:24 it's really quite simple to do multiple targets 2011-06-30T07:57:40 yeah, but first i must speed up this one 2011-06-30T07:57:47 the implementation isnt optimal 2011-06-30T07:57:49 what's so sow abou tit? 2011-06-30T07:57:54 you want to see it? 2011-06-30T07:57:57 i'm not sure how else i could implement it 2011-06-30T07:58:09 i just have a priority_queue as my search queue 2011-06-30T07:59:02 well, priority_queue, greater > and operator>(const FoodSearch &a, const FoodSearch &b) 2011-06-30T07:59:51 i prefer not to share actual strategy code 2011-06-30T07:59:57 not until the end 2011-06-30T07:59:59 *** boegel has quit IRC (Ping timeout: 240 seconds) 2011-06-30T08:00:02 *** UncleVasya has joined #aichallenge 2011-06-30T08:00:11 lol: http://aichallengebeta.hypertriangle.com/visualizer.php?game=46380&user=26 2011-06-30T08:00:22 *** locutus2 has quit IRC (Ping timeout: 250 seconds) 2011-06-30T08:00:42 UncleVasya: you should probably try to avoid enemies if you have one ant left :P 2011-06-30T08:01:44 http://paste.lisp.org/display/123047 2011-06-30T08:01:46 this is it 2011-06-30T08:01:51 Since refactor of my bot this system was broken. 2011-06-30T08:01:56 mcstar: ^^ :P 2011-06-30T08:02:07 Usually it works this way: 2011-06-30T08:02:17 < 25 - avoiding enemies 2011-06-30T08:02:24 the first part is the uploading, the astar function is important 2011-06-30T08:02:55 25 < 60 - not very strong attack 2011-06-30T08:02:58 mcstar: imo it's better if people know their bot is being compared to each individual persons effort 2011-06-30T08:03:08 > 60 - strong attack 2011-06-30T08:03:09 rather than how many people uploaded an instance of variation of a released bot 2011-06-30T08:03:16 antimatroid: i just read about sharing 2011-06-30T08:03:17 or copied other peoples released code 2011-06-30T08:03:23 it doesn't really matter :) 2011-06-30T08:03:25 well, its really a very general code 2011-06-30T08:03:34 but in general i prefer to keep to that 2011-06-30T08:04:00 that seemed to be the general consensus of people when code code released for pw anyway 2011-06-30T08:04:06 i am fairly indifferent personally 2011-06-30T08:04:22 i dont plan to share code until the end, i told stocha that too 2011-06-30T08:04:27 you know, you was there 2011-06-30T08:04:45 were 2011-06-30T08:04:47 sry 2011-06-30T08:04:51 :) 2011-06-30T08:04:57 picking up the bogan language 2011-06-30T08:05:20 bogan? 2011-06-30T08:05:59 http://en.wikipedia.org/wiki/Bogan 2011-06-30T08:06:08 although i'd almost have to classify myself as bogan with that definition 2011-06-30T08:06:37 hey, i forgot to colorize it 2011-06-30T08:06:38 damn 2011-06-30T08:06:40 Don't you know what happened to the server? I had only about 11 games for last 13 hours. 2011-06-30T08:06:52 to me bogans are the people in australia who like to cause trouble for other people for no reason 2011-06-30T08:06:52 lower left corner: pls select common lisp and format 2011-06-30T08:07:05 if by any chance you want to read that code.... 2011-06-30T08:07:07 which i don't self classify with 2011-06-30T08:07:08 id be glad 2011-06-30T08:07:43 well, i create trouble 2011-06-30T08:07:51 but not physical/demeaning/racial trouble 2011-06-30T08:08:00 i see, the tassie synonyme for redneck 2011-06-30T08:08:06 australian* 2011-06-30T08:08:08 but yeah kind of 2011-06-30T08:08:34 the scum of the nation 2011-06-30T08:09:14 back to the code pls 2011-06-30T08:09:16 :D 2011-06-30T08:09:18 :p 2011-06-30T08:10:37 *** locutus2 has joined #aichallenge 2011-06-30T08:12:29 *** onensora has joined #aichallenge 2011-06-30T08:21:30 this image is somewhat better http://imgur.com/WZB59 2011-06-30T08:22:08 What's that? 2011-06-30T08:23:00 *** boegel has joined #aichallenge 2011-06-30T08:23:13 pathfinding? 2011-06-30T08:23:30 from darker blue to lighter blue 2011-06-30T08:23:42 I see 2011-06-30T08:26:24 sYnfo: have you done pathfinding yet? 2011-06-30T08:26:27 are you using a*? 2011-06-30T08:27:17 mcstar: is the open/closed set thing meant to speed it up somehow? 2011-06-30T08:29:21 well, i scan the open set to minimize the h function 2011-06-30T08:29:24 *** jmreardon has joined #aichallenge 2011-06-30T08:29:57 i don't understand why you can't do the open/closed set way if the heuristic function isn't monotone, but you can use the ordinary prique methof 2011-06-30T08:30:10 doesn't that sound counterintuitive? i would have expected it to be the other way around 2011-06-30T08:30:34 for example, the path is 273 long, and it touched 3877 tiles 2011-06-30T08:30:42 antimatroid: Nope, my bot is far from working right now, I don't have much time alas. But a* is what I'll be going for 2011-06-30T08:31:43 antimatroid: i will try to make it simpler tonite 2011-06-30T08:31:50 *** boegel has quit IRC (Quit: *poof!*) 2011-06-30T08:31:57 i want to understand the monotone stuff though 2011-06-30T08:32:08 if it's faster to use open/closed sets i will do it, but i don't see why that would be faster 2011-06-30T08:32:27 antimatroid: could you give me an approximation, that in similar circumstances how many nodes does your code visit before finding the path? 2011-06-30T08:32:55 mcstar: not easily, i haven't got it open and would need to set it up to do just finding a path etc. 2011-06-30T08:34:04 *** pvarga has quit IRC (Quit: pvarga) 2011-06-30T08:35:47 if i save this map in an ascii file, could you use it? 2011-06-30T08:36:12 i'd have to write specific code 2011-06-30T08:36:16 really cbf atm sorry :P 2011-06-30T08:36:28 thats a 1 liner isnt it? 2011-06-30T08:36:32 ok 2011-06-30T08:36:34 cbf? 2011-06-30T08:36:34 no 2011-06-30T08:36:38 can't be f'd 2011-06-30T08:36:40 cba 2011-06-30T08:37:23 actually maybe, pastebin the map 2011-06-30T08:37:43 what format dou you want? 2011-06-30T08:37:51 like an actual game 2011-06-30T08:37:52 want the whole map, or al ist of the obstacles? 2011-06-30T08:38:07 copy the sample input data from the specs page 2011-06-30T08:38:09 in ants specification? 2011-06-30T08:38:09 and make it like that 2011-06-30T08:38:23 oh, i need all water squares listed 2011-06-30T08:38:32 i haven't got a function to read a state from a char grid 2011-06-30T08:38:50 just from normal state input 2011-06-30T08:38:54 and i have my grid set up as a graph etc. 2011-06-30T08:39:21 i think its the easiest if i give you the coordinate pairs on separate lines 2011-06-30T08:39:50 *** UncleVasya has quit IRC (Ping timeout: 252 seconds) 2011-06-30T08:39:51 nah, if i can pass input to my bot like it's a game, then i can do it 2011-06-30T08:39:57 otherwise it's too much effort :P 2011-06-30T08:41:31 pfff 2011-06-30T08:41:42 you want it like w row col 2011-06-30T08:41:49 ok, it the same for me 2011-06-30T08:43:14 http://pastebin.com/DbBYCzkS 2011-06-30T08:44:16 bah, that isn't a game state :P 2011-06-30T08:44:46 bah, 2011-06-30T08:44:53 what do you mean game state? 2011-06-30T08:44:58 you want all the info? 2011-06-30T08:45:12 you can simply paste it there yourself 2011-06-30T08:45:32 yeah, i want my bot to just think it's playing a game when i feed the input to it 2011-06-30T08:45:35 hey i got to go, and a fing thunderstorm is coming so im powering down 2011-06-30T08:45:49 then pls paste it in yourself :) 2011-06-30T08:46:05 http://pastebin.com/Lq8pwhFg 2011-06-30T08:46:16 powering down for a thunderstorm? 2011-06-30T08:46:18 are you female? 2011-06-30T08:46:38 :) 2011-06-30T08:48:12 thehttp://pastebin.com/G9qif0uB 2011-06-30T08:48:17 http://pastebin.com/G9qif0uB 2011-06-30T08:48:28 a thunderstorm is very serious 2011-06-30T08:48:42 power can be severed 2011-06-30T08:48:54 and after one such event i couldnt boot up my pc 2011-06-30T08:49:04 for many many minutes 2011-06-30T08:49:26 you laugh until something of yours will be broken 2011-06-30T08:49:40 nowadays we have some very serios lightnings 2011-06-30T08:49:54 we don't get lightning much anyway 2011-06-30T08:49:59 i hope that paste helped im really off now 2011-06-30T08:50:01 bb 2011-06-30T08:50:02 i love lightning 2011-06-30T08:50:11 nye a couple of years ago was the best lightning of my life 2011-06-30T08:50:15 me too, but equipments dont 2011-06-30T08:50:19 the entire sky was lighting up 2011-06-30T08:50:28 very nice 2011-06-30T08:50:29 bb 2011-06-30T08:50:31 *** mcstar has quit IRC (Quit: WeeChat 0.3.5) 2011-06-30T08:56:33 *** olexs has joined #aichallenge 2011-06-30T09:46:45 *** NielsE has joined #aichallenge 2011-06-30T09:49:09 hi, with the default test bot from the beta I'm getting ---> o 3 18 w # move blocked 2011-06-30T09:49:15 but I'm pretty sure it's a valid move... 2011-06-30T09:49:21 *** locutus2 has quit IRC (Ping timeout: 264 seconds) 2011-06-30T09:49:33 is there anyone else that has seen the same problem 2011-06-30T09:49:33 ? 2011-06-30T09:49:38 *** locutus2 has joined #aichallenge 2011-06-30T09:54:08 NielsE: I had a few of those with my bot. It always was moving either on water or onto food squares. 2011-06-30T09:57:51 yeah its onto a food square 2011-06-30T09:58:13 but how is that a problem and should it not be an ant when you are next to it? 2011-06-30T10:01:40 Hmh, I 2011-06-30T10:01:42 'Damn 2011-06-30T10:01:49 NielsE: I had the same issue yesterday. It is possible to end up next to a food square without it spawning an ant, it is then an illegal move to move onto it. 2011-06-30T10:01:59 NielsE: https://github.com/aichallenge/aichallenge/issues/204 2011-06-30T10:03:08 aerique: but why didn't the food become an ant? 2011-06-30T10:03:19 so you should basically treat such a square like water 2011-06-30T10:03:39 should I treat every food as water? 2011-06-30T10:03:58 NielsE: it might be next to an ant that just spawned 2011-06-30T10:04:02 NielsE: it's due to the order of ant moves and ant spawning in the engine 2011-06-30T10:04:05 or it may have just spawned next to you? 2011-06-30T10:04:19 Hmh, I've made a veeery slight modification to the Python3 started bot's ants.py to make it a liiitle bit more efficient, is there any way to modify the code on github? (Obv a newbie here :) ) 2011-06-30T10:04:42 sYnfo: the starter bots are really just a place for people to start 2011-06-30T10:04:58 i usually rewrite my own one 2011-06-30T10:05:20 although i wrote the c++ starter bot a lot like i structure my own bots normally, just a bit simpler in places 2011-06-30T10:05:25 Even the ants.py? Ok, I thought people usually let that be :) 2011-06-30T10:05:36 yeah i don't touch ants.py 2011-06-30T10:05:42 but that doesn't affect your bot 2011-06-30T10:05:51 Well 2011-06-30T10:05:54 or maybe it does for the python bot? 2011-06-30T10:05:58 No no no 2011-06-30T10:06:00 i don't like python very much :P 2011-06-30T10:06:45 My point is, if people don't touch the ants.py, any gain in efficiency in it will translate to more games played on the server, right? 2011-06-30T10:06:56 I don't get you people, Python is awesome:P 2011-06-30T10:07:07 i find it confusing to program in python 2011-06-30T10:07:25 but the only programming i've ever really done in it is with the map generators for this, so i don't know it very well 2011-06-30T10:07:32 c++ is so close to perfect :( 2011-06-30T10:07:32 My friend feels the same, I on the other hand, not at all, python is a blessing for me 2011-06-30T10:07:39 :) 2011-06-30T10:07:43 Deppends I guess 2011-06-30T10:08:15 the beta site seems to have stopped working, it does not seem to do any games anymore 2011-06-30T10:12:21 *** onensora has quit IRC () 2011-06-30T10:15:47 aerique: thanks btw, I've fixed my code so it does not complain anymore 2011-06-30T10:16:30 *** onensora has joined #aichallenge 2011-06-30T10:19:54 *** locutus2 has quit IRC (Ping timeout: 250 seconds) 2011-06-30T10:26:33 no workers are active, no servers to play games on, only the web server is running 2011-06-30T10:29:02 *** locutus2 has joined #aichallenge 2011-06-30T10:34:44 *** gobang has quit IRC (Read error: Connection reset by peer) 2011-06-30T10:34:45 *** sigh has quit IRC (Remote host closed the connection) 2011-06-30T10:42:06 *** aerique has quit IRC (Quit: ...) 2011-06-30T10:59:55 *** locutus2 has quit IRC (Ping timeout: 240 seconds) 2011-06-30T11:03:10 antimatroid: non-monotonic heuristics are technically admissible i think but don't really work very well in my experience 2011-06-30T11:03:34 a1k0n_: I can't even think of an example of one though? 2011-06-30T11:03:41 *** TimTheEnchanter has joined #aichallenge 2011-06-30T11:03:43 like you can have a function which has some penalty term added, and the penalty suddenly goes away when you reach some state 2011-06-30T11:03:52 but that doesn't "guide" the search at all 2011-06-30T11:03:53 surely any kind of metric for a metric space is monotonic from the triangle inequality? 2011-06-30T11:03:59 *** jheming3 has joined #aichallenge 2011-06-30T11:04:27 bbiab, got to go to work 2011-06-30T11:04:40 but yeah, that's probably a good argument against trying anything like that 2011-06-30T11:05:08 a1k0n_: the other thing that bugs me is the open/closed set implementation rather than just a priority queue (ordered queue), wikipedia says the non-monotonic one doesn't work for the open/closed set version, does that mean the open closed set version would be quicker with monotonic? 2011-06-30T11:05:16 i would have expected it to be the other way around... :s 2011-06-30T11:05:22 ah okay 2011-06-30T11:05:31 well answer that at some point for me :) it's been bugging me all day :P 2011-06-30T11:05:37 enjoy work! 2011-06-30T11:07:40 *** jheming3 has left #aichallenge 2011-06-30T11:08:53 *** locutus2 has joined #aichallenge 2011-06-30T11:10:05 *** TimTheEnchanter has quit IRC (Quit: Page closed) 2011-06-30T11:14:35 *** gobang has joined #aichallenge 2011-06-30T11:22:15 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has joined #aichallenge 2011-06-30T11:23:55 *** gobang has quit IRC (Ping timeout: 240 seconds) 2011-06-30T11:29:41 *** Kingpin13 has quit IRC (Ping timeout: 246 seconds) 2011-06-30T11:31:30 the closed set is just to handle the case where you already visited a node.. which.. yeah, probably wouldn't work with a nonmonotonic function 2011-06-30T11:31:40 but it speeds it up a lot on a grid-based thing 2011-06-30T11:32:54 but if you have, say, an "objective" of starting at point a, grabbing food at point b, and then going to point c, then, well, that should really be two searches, but if you wanted to do one, you'd have to know whether you already visited b at each state 2011-06-30T11:33:05 in which case the heuristic changes 2011-06-30T11:35:41 *** Naktibalda has quit IRC (Quit: ChatZilla 0.9.87 [Firefox 5.0/20110615151330]) 2011-06-30T11:36:53 *** gobang has joined #aichallenge 2011-06-30T11:38:27 *** stocha has joined #aichallenge 2011-06-30T11:39:40 why is my c little benchmark (-O3) showing that my same java little benchmark is faster ?? Is that usual ? (I'm on mac OSX and use -O3 on the Netbeans project) 2011-06-30T11:43:54 *** stocha has quit IRC (Ping timeout: 252 seconds) 2011-06-30T11:47:21 *** stocha has joined #aichallenge 2011-06-30T11:48:42 *** sYnfo has quit IRC (Quit: Page closed) 2011-06-30T11:50:34 *** Naktibalda has joined #aichallenge 2011-06-30T11:50:54 *** mleise has joined #aichallenge 2011-06-30T11:52:30 antimatroid wants people to use c++ so he has a chance to win :p 2011-06-30T11:52:47 how does that affect my chance of winning? 2011-06-30T11:54:23 antimatroid: because no one can make a bot using c++ ? :p 2011-06-30T11:54:52 a good one i mean 2011-06-30T11:55:05 @rankings 2011-06-30T11:55:17 stocha: Error: There was a problem accessing the interface to ai-contest.com 2011-06-30T11:55:45 contestbot: yeah. juste i what i though 2011-06-30T11:55:46 stocha: User error, it's not my fault. 2011-06-30T11:55:56 contestbot: who's fault is it then ? 2011-06-30T11:55:57 stocha: I have no idea what you mean. 2011-06-30T11:56:12 contestbot: nobody seems to, really. 2011-06-30T11:56:13 stocha: You have no gotten any error messages recently, so here's a random one just to let you know that we care. 2011-06-30T11:56:30 contestbot: thanks old chap. 2011-06-30T11:56:31 stocha: An error occured while trying to show the previous error. 2011-06-30T11:57:58 So how is the contest coming along? I've been busy playing Uru. 2011-06-30T11:58:15 *** nux67 has joined #aichallenge 2011-06-30T11:58:27 mleise: quite well in my opinion. There is still some work though. 2011-06-30T11:58:58 *** McLeopold has joined #aichallenge 2011-06-30T12:02:58 *** stocha has quit IRC (Ping timeout: 252 seconds) 2011-06-30T12:13:17 *** fab0109 has joined #aichallenge 2011-06-30T12:13:41 McLeopold: I'll add the food display you asked for. It's not much work since I already create dummy ants for the player spawns. 2011-06-30T12:16:40 what did I ask for? 2011-06-30T12:18:08 mleise: http://ants.zeroviz.us/NneQM/ check out how slow this is 2011-06-30T12:19:44 *** Accoun has quit IRC () 2011-06-30T12:20:32 McLeopold: omg what is this? it almost kills Opera ...now i see something 2011-06-30T12:20:59 oh my fucking god... 2011-06-30T12:21:26 how many ants? 2011-06-30T12:22:04 *** fab0109 has quit IRC (Quit: Page closed) 2011-06-30T12:22:07 haha 2011-06-30T12:22:26 I was trying to display a map showing who can access which parts first 2011-06-30T12:22:27 *** fab0109 has joined #aichallenge 2011-06-30T12:23:13 looks like the center is dead land 2011-06-30T12:23:13 I remember what I asked now, can maps display food 2011-06-30T12:23:18 *** nino has joined #aichallenge 2011-06-30T12:23:47 you don't intend to put up a flood fill map, do you? 2011-06-30T12:23:56 *** nino is now known as Outopia 2011-06-30T12:23:57 no, it's for analysis only 2011-06-30T12:24:30 I just want the visualizer to have a special map mode where it displays whatever the map actually looks like, ants, food, etc 2011-06-30T12:24:46 without going through the setup for actually playing a game 2011-06-30T12:24:54 maps with pre-placed food would be cool, though 2011-06-30T12:25:01 I'll just say pacman :) 2011-06-30T12:25:25 McLeopold: so... you want a more static version of it, right? 2011-06-30T12:25:40 when checking symmetry, I want to place some random food and look at the results 2011-06-30T12:25:44 right 2011-06-30T12:26:22 I have no idea how much work that is. It would only really be useful for the ant paste site right now 2011-06-30T12:27:56 well, the cursor, and attack circles for example are drawn over ant positions. so for that to work I need all the ants as 'animation objects' with their 'current' coordinates 2011-06-30T12:28:12 @later tell stocha scoring is not going to change, it is a compromise between surviving order, best food gatherer and best fighter 2011-06-30T12:28:12 McLeopold: Yep. 2011-06-30T12:28:17 there is also the label option for color blind people 2011-06-30T12:28:31 wich works on the same basis 2011-06-30T12:29:00 hmm... 2011-06-30T12:29:14 so what's the slowdown? creation of objects? 2011-06-30T12:30:02 pass it through chrome's profiler :) 2011-06-30T12:30:06 how about we get tricky, capitol letters won't be animated, lower case will? 2011-06-30T12:30:58 There are no letters in my data. I just have 'owner' as an index starting at 0 2011-06-30T12:31:38 unfortunately I redraw the whole map when the button tool tip changes ^^ 2011-06-30T12:32:29 hello, do you now when the chalenge will be closed aka how many time do we have before the end 2011-06-30T12:33:01 you ask the wrong question son 2011-06-30T12:33:12 the answer is 2 months though 2011-06-30T12:33:20 most probably at least 2011-06-30T12:34:58 thanks you look like it will be my summer chalenge 2011-06-30T12:35:28 fab0109: (it hasn't officially started yet) 2011-06-30T12:38:27 McLeopold: First I get a list of all ants visible in the current turn (results are cached), then I calculate the effective coordinates with wrapping and map scrolling in mind, checking which ants are outside of the visible area and discarding those (has no effect when zoomed out of course) 2011-06-30T12:38:48 can you skip visibillity when in map mode? 2011-06-30T12:39:23 McLeopold: The remaining ants are drawn and their size (spawning ants pop up a bit) and location checked again for possible wrapping situations where they have to be drawn multiple times 2011-06-30T12:40:02 McLeopold: you can zoom-in in map mode, so it is not obsolete 2011-06-30T12:41:53 I'm currently on Windows and don't have Chrome installed. With a non-minified version of the JavaScript, it can be profiled pretty well. 2011-06-30T12:44:03 k 2011-06-30T12:45:17 ah you use the non-minified version already. ok :) 2011-06-30T12:46:23 McLeopold: It will show either 'interpolate' or 'draw' as taking most of the time. 2011-06-30T12:50:45 *** gobang has quit IRC (Ping timeout: 255 seconds) 2011-06-30T12:52:39 *** gobang has joined #aichallenge 2011-06-30T12:57:20 *** Naktibalda has quit IRC (Quit: ChatZilla 0.9.87 [Firefox 5.0/20110615151330]) 2011-06-30T13:05:01 *** amstan has joined #aichallenge 2011-06-30T13:05:01 *** ChanServ sets mode: +o amstan 2011-06-30T13:07:10 *** McLeopold has quit IRC (Read error: Connection reset by peer) 2011-06-30T13:10:43 *** McLeopold has joined #aichallenge 2011-06-30T13:11:52 *** Apophis_ has quit IRC (Quit: Verlassend) 2011-06-30T13:16:54 *** Apophis has quit IRC (Read error: Connection reset by peer) 2011-06-30T13:22:44 *** Apophis has joined #aichallenge 2011-06-30T13:25:51 *** gobang has quit IRC (Ping timeout: 255 seconds) 2011-06-30T13:26:24 *** gobang has joined #aichallenge 2011-06-30T13:31:32 *** choas has joined #aichallenge 2011-06-30T13:33:50 *** NielsE has quit IRC (Quit: Page closed) 2011-06-30T13:55:50 *** gobang has quit IRC (Read error: Connection reset by peer) 2011-06-30T13:55:56 *** smiley1983 has joined #aichallenge 2011-06-30T13:57:49 *** gobang has joined #aichallenge 2011-06-30T13:59:02 wow, ksaskosj finally got rewarded for not giving up ;) 2011-06-30T14:00:21 *** smiley1983 has quit IRC (Ping timeout: 240 seconds) 2011-06-30T14:00:23 @later tell antimatroid wasn't this already implemented? https://github.com/aichallenge/aichallenge/pull/203 2011-06-30T14:00:23 amstan: OK 2011-06-30T14:00:37 @later tell mcleopold wasn't this already implemented? https://github.com/aichallenge/aichallenge/pull/203 2011-06-30T14:00:37 amstan: Yep. 2011-06-30T14:00:43 *** smiley1983 has joined #aichallenge 2011-06-30T14:01:37 amstan: let me check 2011-06-30T14:02:00 i tought we alreaday handled every kind of possible symmetry 2011-06-30T14:03:15 amstan: let's just leave the issue, map generation may continue throughout the contest, it it is up to the one who write the generator to choose symmetries 2011-06-30T14:03:41 also, do we really have to commit maps in the repo? it kinda makes everything messy 2011-06-30T14:03:46 We can detect all symmetries that we can fairly spawn food on 2011-06-30T14:04:09 maps should be up to the server admin to put in effect, it doesn't really have much to do with the code 2011-06-30T14:04:17 amstan: no, we won't for launch, but I wanted the beta test maps in there 2011-06-30T14:04:34 I just wanted a few with different types of symmetry as a test 2011-06-30T14:05:06 We may want to include some generated maps in the tools.zip download? 2011-06-30T14:05:29 or a utility to download official maps 2011-06-30T14:05:58 i propose just a script that makes them 2011-06-30T14:06:10 I'm working on that 2011-06-30T14:06:34 and we shouldn't really encourage getting maps from the server, unless you have to debug for the one game your bot crashed in 2011-06-30T14:07:16 why not? 2011-06-30T14:08:07 the maps we provide should only be samples, they don't really have to have anything in common with the ones on the server 2011-06-30T14:08:31 True, but some may still want to download them 2011-06-30T14:08:32 we had this happen a few times, lots of outrage over map types 2011-06-30T14:08:46 it should be understood that there's nothing about them that's special 2011-06-30T14:09:06 if they want to download them they can analyze the json and get the map name and use the map/ url 2011-06-30T14:09:06 But blocking access doesn't make sense 2011-06-30T14:09:21 we're not blocking access, we're just not offering a tool 2011-06-30T14:09:36 But why? Are you worried about bandwidth? 2011-06-30T14:09:36 oh, they can also look in the replay, the map is listed there right? 2011-06-30T14:10:15 Are you worried that some will only test on official maps and then complain that they weren't a fair sample? 2011-06-30T14:10:27 no, but it's extra work for not much gain 2011-06-30T14:10:48 McLeopold: partially, yes 2011-06-30T14:11:27 I believe opening up the contest as much as possible and providing the best tools for what people want is the best policy. 2011-06-30T14:11:41 And *I* want to be able to download officially used maps. 2011-06-30T14:12:04 if you really want to make a tool, go for it 2011-06-30T14:12:08 I don't want to guide the contest based on if people might complain 2011-06-30T14:12:17 but i don't think it's important 2011-06-30T14:12:23 ok, after launch 2011-06-30T14:12:50 also, keep in mind the ammount of tools that will be made during the contest 2011-06-30T14:13:01 someone might make it themselves 2011-06-30T14:13:04 btw, the tool I used to generate new maps on the server would be about the same that users would use for their own map generation 2011-06-30T14:13:13 yes, i know 2011-06-30T14:16:31 *** gobang has quit IRC (Read error: Connection reset by peer) 2011-06-30T14:16:52 *** gobang has joined #aichallenge 2011-06-30T14:17:41 now say, i'd like to get a copy of octagonal_73.map, what would i have to do ? it's not so obvious on beta 2011-06-30T14:19:19 gobang: since they're in the repo, you could find it there 2011-06-30T14:21:03 just saying, you can run it in the visualizer, but not download it 2011-06-30T14:22:10 *** mcstar has joined #aichallenge 2011-06-30T14:28:57 k, git pull did it. 2011-06-30T14:34:44 gobang: there is a url 2011-06-30T14:35:05 http://aichallengebeta.hypertriangle.com/map/ 2011-06-30T14:37:27 ;) thanks, i was just trying to point out, that it's not obvious. 2011-06-30T14:37:40 *** Naktibalda has joined #aichallenge 2011-06-30T14:41:50 *** smiley1983 has quit IRC (Quit: leaving) 2011-06-30T14:43:42 there's a lot of information, that's pretty common to lurkers of this channel, but imagine being a n00b, trying to get somewhwre whithout that 2011-06-30T14:55:04 *** mleise has quit IRC (Quit: Leaving.) 2011-06-30T15:05:12 *** nux67 has quit IRC (Ping timeout: 252 seconds) 2011-06-30T15:17:50 *** olexs has quit IRC (Quit: Leaving.) 2011-06-30T15:21:32 i was gonna do an ants bot last week but instead i made this: http://a1k0n.net/2011/06/26/obfuscated-c-yahoo-logo.html 2011-06-30T15:39:47 hehe nice 2011-06-30T15:40:03 i bet you have some sort of script that obfuscates code 2011-06-30T15:40:31 i dont think you do all of this on a one-by-one basis 2011-06-30T15:40:33 or do you? 2011-06-30T15:44:03 *** stocha has joined #aichallenge 2011-06-30T15:46:52 @later tell McLeopold why do everyone thinks i want scoring to change ? I just want that the last remaining bot doesn't get the bonus if it isn't the last remaining bot (one or more timedout bot): the game should playout normaly. However i more worryed about the map policy. Maps should be genarated with fixed rules that don't evolve after beta. They should be random for each map, with a fair seed. And the map generator have to 2011-06-30T15:46:52 stocha: Ready to serve. 2011-06-30T15:47:24 @later tell McLeopold If not the case, then there should be a fixed set of map, that people have access to. 2011-06-30T15:47:24 stocha: Ready to serve. 2011-06-30T15:48:42 have to .. what? 2011-06-30T15:48:57 *** gobang has quit IRC (Ping timeout: 255 seconds) 2011-06-30T15:49:38 a1k0n_: you there? 2011-06-30T15:51:46 *** stocha has quit IRC (Ping timeout: 252 seconds) 2011-06-30T15:55:57 *** stocha has joined #aichallenge 2011-06-30T15:56:56 *** Apophis has quit IRC (Read error: Connection reset by peer) 2011-06-30T16:01:44 *** okayzed is now known as okay 2011-06-30T16:04:54 janzert: I'm going to try cutting games off earlier. 66% of ants+food for 150 turns. This means holdbots will go for a min of 150. The cutoff counter starts sooner, but you have longer to turn the tide. 2011-06-30T16:05:51 ok, it sounds pretty early to me but might as well try it 2011-06-30T16:06:03 66 is now 2011-06-30T16:06:05 low 2011-06-30T16:06:20 but I want to see what happens 2011-06-30T16:06:23 also it seems the current maps have a lower skill separation 2011-06-30T16:06:30 yeah, they suck 2011-06-30T16:06:40 I just wanted to test symmetries 2011-06-30T16:06:48 *** stocha has quit IRC (Ping timeout: 252 seconds) 2011-06-30T16:06:50 did antimatroid get some new symmetries built? 2011-06-30T16:06:53 I don't feel the need for fixing them until launch 2011-06-30T16:07:05 yes 2011-06-30T16:07:11 we have a few generators in the works 2011-06-30T16:07:25 might as well throw the current ones out at least since we've seen them 2011-06-30T16:07:34 ok 2011-06-30T16:07:45 *** Accoun has joined #aichallenge 2011-06-30T16:07:49 It also shows up map selection may need to be tuned 2011-06-30T16:07:54 btw, I'll be gone for at least 24 hours in about an hour here 2011-06-30T16:08:06 can I have access to workers? 2011-06-30T16:08:07 and don't know what my connection status will be for the next week 2011-06-30T16:08:09 just in case? 2011-06-30T16:08:32 it's fine with me, but ask amstan, we probably need jeff's permission 2011-06-30T16:08:50 *** hacklash has joined #aichallenge 2011-06-30T16:08:54 we can switch to a different ssh key when we start the real contest so that part shouldn't be a problem 2011-06-30T16:08:56 are the workers using jeff amazon account? 2011-06-30T16:09:00 yes 2011-06-30T16:09:04 ok 2011-06-30T16:09:29 do you have a howto on spinning up a new worker or taking one down? 2011-06-30T16:09:51 also so amstan knows the reason he couldn't get access at first with them was jeff had used a different ssh key we didn't have 2011-06-30T16:10:01 oh 2011-06-30T16:10:16 amstan can look in the bash history on the contest user on jeff's server 2011-06-30T16:10:20 so amstan can spin up a new worker? 2011-06-30T16:10:23 all the commands are there 2011-06-30T16:11:23 basic overview is, you have to sudo to the contest user, then run the ~/ec2-check/ec2-env.sh to set up the ec2 access for the ec2 scripts 2011-06-30T16:12:08 the ec2-describe-instances will list the current instances, the instance name that looks like i-hexnumber is the important part 2011-06-30T16:12:23 then use ec2-terminate-instance to get ec2 to kill and start a new worker 2011-06-30T16:12:43 or ec2-describe-spot-instance-requests to list the current requests 2011-06-30T16:13:02 and ec2-cancel-spot-instance-request to cancel them 2011-06-30T16:13:18 I don't remember the full command to create a new request but it'll be in the history 2011-06-30T16:13:39 you manager ec2 instances from command line? 2011-06-30T16:13:51 for jeff's account I do :) 2011-06-30T16:14:06 don't have his ec2 login details to access the web page 2011-06-30T16:14:34 also you can look at http://cloudexchange.org/charts/us-east-1.linux.m2.xlarge.html to see the current spot price 2011-06-30T16:14:47 if it's above 0.20 that's why you don't have any workers :/ 2011-06-30T16:15:22 Are the m2.xlarge instances that much more effective than the m1.s? We can actually run 5 times as many games? 2011-06-30T16:15:28 supposedly the hotel I'll be at has internet access so after tomorrow evening I should be able to check in 2011-06-30T16:15:53 (I'm not sure; I just would've thought that it would've been cheaper on the smaller instances.) 2011-06-30T16:15:56 jbroman: look at the memory size 2011-06-30T16:16:24 we can't run any games with more than 2 players on m1.small and that would be tight 2011-06-30T16:16:51 also we'd probably get lots of complaints about the slow cpu speed 2011-06-30T16:18:02 http://aws.amazon.com/ec2/instance-types/ has the specs btw 2011-06-30T16:18:04 those prices have odd spikes 2011-06-30T16:18:06 janzert: Oh, I didn't realize that players were getting on the order of a gigabyte or two of memory. 2011-06-30T16:18:10 * jbroman (I've seen the specs before.) 2011-06-30T16:18:36 yeah, java starts throwing fits when you try and reduce it much below 1GB :( 2011-06-30T16:18:44 Aww. 2011-06-30T16:18:56 McLeopold: yeah, the last 48 hours has been weird with the spot pricing 2011-06-30T16:19:38 seems to happen about once a month although this was longer spikyness(?) than usual 2011-06-30T16:27:53 *** Outopia has quit IRC (Quit: ChatZilla 0.9.87 [Firefox 5.0/20110615151330]) 2011-06-30T16:32:59 *** stocha has joined #aichallenge 2011-06-30T16:34:06 janzert: it's not necessarily that the map has lower skill separation. It might be that the bots features where well tuned for the older maps (no starting proximity especially, and low reward for exploration, small maps) 2011-06-30T16:35:09 jmreardon: i think the three generator, the old random walk, the current one, and the futur maze generator offers a good range of maps. 2011-06-30T16:40:53 janzert: is this for the lowes prices only? that availability depends on the number of premium users? 2011-06-30T16:40:58 lowest 2011-06-30T16:45:58 antimatroid: i have a movie of the pathfinding process 2011-06-30T16:46:07 i suggest you download it with wget 2011-06-30T16:46:15 i put it up on github 2011-06-30T16:46:25 and it doesnt really like binary files :) 2011-06-30T16:46:34 wget https://github.com/liquid-phynix/cl-ants-engine/blob/master/astar_1.mpeg?raw=true 2011-06-30T16:54:28 *** Naktibalda has quit IRC (Quit: ChatZilla 0.9.87 [Firefox 5.0/20110622232440]) 2011-06-30T17:05:35 *** janzert has left #aichallenge 2011-06-30T17:10:18 *** mleise has joined #aichallenge 2011-06-30T17:16:15 *** olexs has joined #aichallenge 2011-06-30T17:19:52 how is the food spawned on the different map types? 2011-06-30T17:20:38 are there certain spots that spawn food or have a higher probability of spawning food? 2011-06-30T17:26:16 *** Cyndre has joined #aichallenge 2011-06-30T17:36:37 hacklash: you can look at the code 2011-06-30T17:36:41 mcstar: i don't have a script that obfuscates code, though i do have scripts which compress ascii art into various representations 2011-06-30T17:37:04 (which wasn't used in this case) 2011-06-30T17:37:35 hacklash: I don't think it is intended to have higher food spawn rates on any square 2011-06-30T17:38:08 but it seems like on the octagonla one, that there are certain squares that spawn food 2011-06-30T17:38:15 when a certain square is blocked in symmetric mode it is delayed until the ant that occupies the square moves 2011-06-30T17:39:12 *** stocha has quit IRC (Ping timeout: 252 seconds) 2011-06-30T17:39:13 hacklash: You probably have to ask McLeopold about it, but I strongly belive that this is coincidence 2011-06-30T17:39:21 ok, thanks 2011-06-30T17:39:25 ? 2011-06-30T17:39:54 McLeopold: would there be a reason for cetain squares on a map to spawn more food than others? 2011-06-30T17:40:18 there shouldn 2011-06-30T17:40:21 't be 2011-06-30T17:40:44 all land squares are divided into symmetric sets 2011-06-30T17:41:03 each square should spawn food at least once before any squares spawns twice 2011-06-30T17:41:24 McLeopold: oh really? i didn't know that! 2011-06-30T17:41:50 so a square is less likely to have food on it if it just had food spawn there? 2011-06-30T17:42:33 yes 2011-06-30T17:44:21 but that may be several 1000 squares so it would be hard to predict in the important start phase where food will spawn 2011-06-30T17:44:52 if you can explore the entire map, then apply the symmetric_food_sets function to it, then track where food has been spawned and detect the turn where the sets are shuffelled, then you *might* get an advantage 2011-06-30T17:45:11 after turn 1000 2011-06-30T17:46:03 the thought has always been that by the time you can use food spawn and map symmetry information to your advantage, it might be too late 2011-06-30T17:47:28 hacklash: if there is a game you think food might not be fair, we can run the map through the symmetric_sets function to see if there are issues 2011-06-30T17:47:47 *** Mathnerd314 has joined #aichallenge 2011-06-30T17:56:09 ok, i just thought the theoretical implications of that decision were interesting. If i see any practical repercussions i'll let you know 2011-06-30T17:59:39 *** olexs1 has joined #aichallenge 2011-06-30T17:59:41 *** olexs has quit IRC (Read error: Connection reset by peer) 2011-06-30T18:02:34 If you can predict where food will spawn, make sure not to have an ant on that spot, because it won't spawn of not free, just wait for the ant to move 2011-06-30T18:13:40 oh, was a max board size ever decided on? 2011-06-30T18:14:50 256x256 2011-06-30T18:14:57 practically it may be less 2011-06-30T18:15:33 *** sigh has joined #aichallenge 2011-06-30T18:15:35 ok, that lets me know if i'll have enough memory for some strategies 2011-06-30T18:16:06 *** Kingpin13 has joined #aichallenge 2011-06-30T18:18:36 i had the crazy idea of caching the pathfinding but estimated that total size of that would grow to about 4 GB, a tad too large to be practical 2011-06-30T18:19:04 1sec per turn also 2011-06-30T18:19:21 searching in a 4gb hashtable takes time... 2011-06-30T18:19:30 you keep that in memory between turns so you don't have to recalculate paths you've already found 2011-06-30T18:20:00 hacklash: maybe keep the last 1000 or so 2011-06-30T18:20:56 wait... getting the value out of a large hashtable is slow? 2011-06-30T18:21:21 that doesn't sound right 2011-06-30T18:21:56 *** Palmik has quit IRC (Remote host closed the connection) 2011-06-30T18:22:26 mcstar: care to explain? 2011-06-30T18:22:35 well, you must use long hashes with that kind of amount of paths 2011-06-30T18:22:54 i dont know numbers, but i guess its very large 2011-06-30T18:23:15 do you have experience with that large hashtables? 2011-06-30T18:23:39 no 2011-06-30T18:23:41 you have to take into account re-hashing 2011-06-30T18:23:42 that's why i asked 2011-06-30T18:23:48 which takes time 2011-06-30T18:24:07 a hash-key only works for a certain amount of hashed values 2011-06-30T18:24:08 if you know how big it's going to be yiou can intialize with the coprrect size and prevent re-hashing 2011-06-30T18:24:16 which yhou could do if you knoew the board size 2011-06-30T18:24:19 when the table gets bigger you have to re-hash everything 2011-06-30T18:24:35 and it could easily take more time than 1 sec i guess 2011-06-30T18:24:51 but a longer key takes longer to find 2011-06-30T18:25:07 there's no free lunch so to speak 2011-06-30T18:25:16 see, i don't get that part 2011-06-30T18:25:34 it should just be a mod operation and a jump instruction 2011-06-30T18:26:19 if the hashtable is adjacent in memory 2011-06-30T18:27:43 is that wrong? 2011-06-30T18:27:45 you want to hash pointers? 2011-06-30T18:27:50 thats good i guess 2011-06-30T18:28:21 obviously you always hash pointers.... 2011-06-30T18:28:25 doestn matter 2011-06-30T18:28:54 what i think, is that its easier to generaTE paths 2011-06-30T18:29:01 than storing them like this 2011-06-30T18:29:04 isnt it? 2011-06-30T18:29:23 generating is not much worse 2011-06-30T18:29:56 theres only a small chance that i want go from A to B more than once 2011-06-30T18:30:00 and much cheaper on memory that you need to keep initialized 2011-06-30T18:30:11 so i dont get the point why not just generate a new path every time 2011-06-30T18:31:01 i guess you already keep a copy of the explored map right? 2011-06-30T18:31:36 *** sigh has quit IRC (Remote host closed the connection) 2011-06-30T18:31:55 yeah, i'm storing boolean arrays, one for water, one for land, null means unseen 2011-06-30T18:32:59 *** hacklash has quit IRC (Quit: Page closed) 2011-06-30T18:34:37 *** olexs1 has quit IRC (Read error: Connection reset by peer) 2011-06-30T18:42:40 hehehehehehehe 2011-06-30T18:42:51 im laughing at myself big time 2011-06-30T18:42:59 because of 1 addition 2011-06-30T18:43:26 my A* scanned 20 times more nodes than it was supposed to 2011-06-30T18:43:28 :) 2011-06-30T18:46:34 true false or null? what language might that be? 2011-06-30T18:46:49 *** jmreardon has quit IRC (Quit: jmreardon) 2011-06-30T18:49:37 lol 2011-06-30T18:49:39 0 2011-06-30T18:49:43 null as in zero 2011-06-30T18:50:24 hacklash said boolean array. so boolean must be some kind of object in that language 2011-06-30T18:50:34 ot it uses "variants" 2011-06-30T18:51:02 ah, ok, i guess i didt read the beginning 2011-06-30T18:51:15 2 separate arrays, ok 2011-06-30T18:51:25 He's talking about memory requirements, but that may be difficult if the language stores an unknown amount of extra info 2011-06-30T18:51:40 or the language uses ints as bools 2011-06-30T18:52:07 https://github.com/liquid-phynix/cl-ants-engine/blob/master/astar_2.mpeg?raw=true 2011-06-30T18:52:20 mleise: now its working properly 2011-06-30T18:52:26 did you see my earlier video? 2011-06-30T18:52:31 nope 2011-06-30T18:52:43 poor github 2011-06-30T18:52:44 compare the 2, quite a performance difference 2011-06-30T18:53:08 i left in an addition in the heuristic function, unfortunately 2011-06-30T18:53:18 why poor github? 2011-06-30T18:53:22 it seems to load or something. there is a rotating progress icon in the browser tab 2011-06-30T18:53:28 ah no 2011-06-30T18:53:33 wget-it 2011-06-30T18:53:43 dont use the browser 2011-06-30T18:53:50 its a binary file 2011-06-30T18:54:10 lol, I whish I rebooted to Linux now 2011-06-30T18:54:16 oh... 2011-06-30T18:54:31 maybe you can save it with right-click 2011-06-30T18:54:50 yeah, you can i believe 2011-06-30T18:55:42 oh god, no media player... wait a minute 2011-06-30T18:55:57 *** mleise has quit IRC (Read error: Connection reset by peer) 2011-06-30T18:57:52 *** mleise has joined #aichallenge 2011-06-30T18:58:13 ok what's the link again? 2011-06-30T18:59:07 https://github.com/liquid-phynix/cl-ants-engine/blob/master/astar_2.mpeg?raw=true 2011-06-30T18:59:25 https://github.com/liquid-phynix/cl-ants-engine/blob/master/astar_1.mpeg?raw=true is the previous one, the faulty 2011-06-30T19:00:26 haha nice. how did you make a video from your algorithm? 2011-06-30T19:00:53 the video? thats easy, xvidcap 2011-06-30T19:01:55 has a little gui, you can select a rectangle you want to record 2011-06-30T19:02:20 but i forgot the command to mencoder and it seems its not in my path, so i left it in mpeg 2011-06-30T19:02:53 ok, nice demo. the first version looks seriously flawed :p 2011-06-30T19:03:33 yeah :S 2011-06-30T19:03:59 of course the visualization has nothing to do with xvidcap, its only for recording the window 2011-06-30T19:05:14 what did you use for the gfx? cpp & cairo? 2011-06-30T19:05:53 common lisp + sdl bindings: lispbuilder-sdl 2011-06-30T19:06:24 i wrote a little abstraction on top of it, the visualizer runs in another thread 2011-06-30T19:06:51 and the algorithm tells it to update the screen after each turn 2011-06-30T19:07:00 *** choas has quit IRC (Quit: leaving) 2011-06-30T19:08:28 it is a good test case with the solid line in the center 2011-06-30T19:09:07 and it makes me wonder if the big red area could be skipped heuristically 2011-06-30T19:09:22 yeah, ill put in it my map-loader function so that i can load up ants maps 2011-06-30T19:10:02 welll maybe 2011-06-30T19:10:22 the thing is: i dont think i need better heuristics 2011-06-30T19:10:29 the maps wont be soo random 2011-06-30T19:10:48 and the planned paths will be much much shorter than the presented one 2011-06-30T19:11:28 and, yes 2011-06-30T19:11:46 before this, i tried to come up with my own "heuristic" pathfinding 2011-06-30T19:12:02 i computed the outlines of individual connected areas of water 2011-06-30T19:12:18 and i simply walked the map, until i hit such an outline 2011-06-30T19:12:40 then, i did a little bit of breadth first search, in the outline, which is not much 2011-06-30T19:12:55 and followed it until an exit condition presented itself 2011-06-30T19:13:11 namely the ant reached the tile in the outline closest to the target 2011-06-30T19:13:27 and i think this can be combined with a* heuristics 2011-06-30T19:14:11 maybe i should give points in the open set(white) higher priority when they are part of such an outline 2011-06-30T19:14:34 but in the ants game it wouldn make much difference 2011-06-30T19:15:15 t 2011-06-30T19:16:25 right 2011-06-30T19:17:00 I had the outline generation for the visualizer, when I was working on rendering coast-lines 2011-06-30T19:18:27 But with the current map sizes it takes to long to render the whole map and it still looked pretty boring with only one terrain texture 2011-06-30T19:19:29 but computing the outline is fast 2011-06-30T19:19:51 so was it supposed to be eyecandy? 2011-06-30T19:19:58 hehe, that's true even in slower JavaScript implementations. 2011-06-30T19:20:22 But I rendered the outline as a shadow (blurred) 2011-06-30T19:20:36 to use it as the base for the coast-line effect 2011-06-30T19:20:57 for me its better not to see such frills 2011-06-30T19:21:20 if ants are little squares i dont need coastlines :) 2011-06-30T19:22:16 Sure, for debugging it is horrible: soft coast-lines, ants and food as graphics. 2011-06-30T19:23:47 If someone was asking for more immersion that would be the ideal place for the eye-candy-mode. But as I couldn't make it good looking enough within certain computational time constraints I never finished that mode. 2011-06-30T19:25:31 good night 2011-06-30T19:25:49 nite 2011-06-30T19:26:08 its early, only 1:25 2011-06-30T19:26:14 quit 2011-06-30T19:26:22 *** fab0109 has left #aichallenge 2011-06-30T19:26:31 :) 2011-06-30T19:30:48 *** mleise has quit IRC (Ping timeout: 255 seconds) 2011-06-30T19:54:56 *** mcstar has quit IRC (Quit: WeeChat 0.3.5) 2011-06-30T20:06:09 *** Apophis has joined #aichallenge 2011-06-30T20:09:15 *** foRei has quit IRC (Read error: Connection reset by peer) 2011-06-30T20:35:10 *** Apophis has quit IRC (Read error: Connection reset by peer) 2011-06-30T20:35:38 *** Apophis has joined #aichallenge 2011-06-30T20:37:47 *** filter_ has joined #aichallenge 2011-06-30T20:39:40 *** filter has quit IRC (Ping timeout: 260 seconds) 2011-06-30T20:45:41 *** tnbd has joined #aichallenge 2011-06-30T20:50:16 *** Kingpin13 has quit IRC (Ping timeout: 240 seconds) 2011-06-30T20:55:21 aichallenge: McLeopold epsilon * r47e2319 / (4 files in 2 dirs): updates to map generators - http://bit.ly/jNl9O8 2011-06-30T21:10:09 *** jbrechtel has joined #aichallenge 2011-06-30T21:10:35 *** tnbd has quit IRC (Quit: Leaving) 2011-06-30T21:11:17 Hi, anyone know why the Scala environment for the ai contest is 2.7.7? Anything that a willing volunteer could do to help get that on a more modern version? 2011-06-30T21:11:29 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has quit IRC (Read error: Connection reset by peer) 2011-06-30T21:19:16 jbrechtel: provide janzert with the install instructions for ubuntu (apt-get or a *.deb package) 2011-06-30T21:19:59 jbrechtel: As for the why, "it's the one in the Ubuntu repos" 2011-06-30T21:21:58 *** McLeopold has left #aichallenge 2011-06-30T21:23:00 *** Evious has quit IRC (Read error: Connection reset by peer) 2011-06-30T21:30:06 jbroman: Cool, thanks for the info. Which version of Ubuntu? 2011-06-30T21:30:25 jbrechtel: I believe it's natty, but not 100% sure on that. 2011-06-30T21:37:11 *** filter_ is now known as filter 2011-06-30T21:47:03 *** locutus2 has quit IRC (Ping timeout: 240 seconds) 2011-06-30T21:52:20 *** okay is now known as okayzed 2011-06-30T21:57:56 jbrechtel: How can I contact janzert? 2011-06-30T21:58:00 oops 2011-06-30T21:58:11 jbroman: I meant 2011-06-30T21:58:45 I've always spoken to him over IRC. amstan may know another way if one exists. 2011-06-30T22:00:11 Ok. I've found a couple PPAs with the latest Scala (2.9.0.1). I imagine those would be OK if a deb would be. 2011-06-30T22:00:13 jbrechtel: hey 2011-06-30T22:00:22 amstan: Hi 2011-06-30T22:00:28 jbrechtel: if you want to look at worker/setup.py 2011-06-30T22:00:39 there might be other languages which get the packages from ppas 2011-06-30T22:02:15 jbrechtel: if you don't want to do all that, you could just submit an issue with the ppa that you would like to be used 2011-06-30T22:02:30 even would go a long way 2011-06-30T22:03:07 Where would worker/setup.py be found? I know of no other software besides the starter packages I can download. 2011-06-30T22:04:31 jbrechtel: oh, sorry: https://github.com/aichallenge/aichallenge 2011-06-30T22:04:44 everything that we develop can be found on github 2011-06-30T22:05:31 Cool, thanks. 2011-06-30T22:05:45 I suppose that'd be the place to submit the issue to as well then or is there another issue tracker somewhere 2011-06-30T22:05:55 nope, we just use the github issue tracker 2011-06-30T22:06:24 Cool 2011-06-30T22:09:53 oh...look at that...it's already using scala 2.8.1 2011-06-30T22:10:03 The webpage is just slightly incorrect 2011-06-30T22:10:26 jbrechtel: yeah, i don't really like that page, it's all manually updated 2011-06-30T22:11:22 Ah. That's unfortunate. 2011-06-30T22:12:33 I'll go ahead and update this to Scala 2.9.0.1 and submit a pull request. 2011-06-30T22:12:39 jbrechtel: thanks 2011-06-30T22:17:02 *** onensora has quit IRC () 2011-06-30T22:23:33 *** jbrechtel has quit IRC (Ping timeout: 264 seconds) 2011-06-30T23:51:10 mcstar: i'm shaped :(