2011-12-10T00:01:32 there 2011-12-10T00:01:44 how do create combat scenarious for testing ? 2011-12-10T00:01:52 http://paste.aichallenge.org/Hvijl/ 2011-12-10T00:01:56 okay, basically, he just keeps 3 guys alive on his front 2011-12-10T00:02:01 best: make a .map (they're super simple) with a and b ants 2011-12-10T00:02:06 then run with --scenario, --food none 2011-12-10T00:02:17 but the gap has to be filled 2011-12-10T00:02:46 remember in that scenario he would have died anyway 2011-12-10T00:02:50 he was one of the front-line ants 2011-12-10T00:02:56 right 2011-12-10T00:03:13 so the reasoning goes "if I don't move, I could be killed" "if I move forward I could also be killed but take some of the bastards with me" 2011-12-10T00:03:32 but you don't take em, you save your own 2011-12-10T00:03:42 HOWEVER it should be P(they'll move forward) * "could be killed" vs P(they'll stand still) * "will be killed" 2011-12-10T00:04:01 and P(they'll move forward) is very small 2011-12-10T00:04:05 so it should feel safe staying in place 2011-12-10T00:09:58 okay, time to implement a1k0n's combat method... 2011-12-10T00:10:02 *** gazpachoking_ has joined #aichallenge 2011-12-10T00:10:09 how did the other one work out? 2011-12-10T00:10:13 aichallenge: janzert epsilon * r9120e1c / (website/css/layout.css website/footer.php): Add totaalnet sponsorship thank you - http://git.io/ayU9ew 2011-12-10T00:11:15 *** gazpachoking has quit IRC (Ping timeout: 260 seconds) 2011-12-10T00:11:43 aichallenge: janzert epsilon * r087cce5 / website/images/totaalnet.png : And add totaalnet logo - http://git.io/R3vazw 2011-12-10T00:11:46 janzert: oh yeah, how's that going? 2011-12-10T00:12:10 janzert: p tag is not padded 2011-12-10T00:12:14 in the side menu 2011-12-10T00:12:36 shouldn't need it 2011-12-10T00:12:42 but something is off 2011-12-10T00:12:52 looks like the css isn't in effect at all 2011-12-10T00:13:01 even though I rebuilt it 2011-12-10T00:13:17 you don't have the sponsor part part of the menu 2011-12-10T00:13:27 not suppose to be 2011-12-10T00:13:33 underscore/dash 2011-12-10T00:13:40 yep, that's it 2011-12-10T00:13:58 *** TheLinker has quit IRC (Quit: Bye) 2011-12-10T00:14:10 janzert: so did we already get the servers? or what happened? you didn't send any emails about it 2011-12-10T00:14:37 yep, two have been running for 6 hours or so 2011-12-10T00:14:42 working on the rest now 2011-12-10T00:14:53 and planning on a forum announcement once I get there :) 2011-12-10T00:15:24 noo! you rebooted! 2011-12-10T00:15:26 lol 2011-12-10T00:15:30 I didn't 2011-12-10T00:15:33 manager uptime went from 50days to 18min 2011-12-10T00:15:35 I saw that it dod though 2011-12-10T00:15:38 did 2011-12-10T00:15:45 kicked my ssh session out 2011-12-10T00:16:25 fixed the hostname 2011-12-10T00:16:33 why did it reboot though? 2011-12-10T00:16:34 aichallenge: janzert epsilon * r08984cf / website/css/layout.css : underscores not dashes - http://git.io/kwp6KA 2011-12-10T00:16:39 11 servers and 5.6 gpm, wow 2011-12-10T00:16:54 I know amazon had a maintenance reboot coming up but I thought it wasn't until sunday 2011-12-10T00:16:59 so not sure 2011-12-10T00:17:03 oh, that too 2011-12-10T00:17:16 said it received a ctrl-alt-del so I assume amazon was the cause though 2011-12-10T00:17:43 there we go 2011-12-10T00:19:04 janzert: are you intending to add to credits.php? 2011-12-10T00:19:42 heh, it still says "Google AI Challenge" 2011-12-10T00:19:52 yeah, if you're editing that, can you change it? 2011-12-10T00:20:24 it had crossed my mind that we should create a sponsors section in there, but I didn't have any specific plans 2011-12-10T00:20:26 i'm thinking of either a Sponsors section or put both google and totaalnet in special thanks 2011-12-10T00:20:32 *** amstan_ is now known as amstan 2011-12-10T00:20:39 yeah, sponsors section would be good 2011-12-10T00:21:15 I think Jeff will need to check with Google before we add them there, and it'd probably be nice to add the logo for that 2011-12-10T00:22:05 they're already in the title as "Sponsored by", i don't think it matters if we add another line in the credits mentioning their name 2011-12-10T00:22:16 the only issue is logo usage 2011-12-10T00:24:28 janzert: is this not vmd? 2011-12-10T00:24:52 oh, nvm 2011-12-10T00:24:56 pci bridge vendor vmware 2011-12-10T00:25:26 janzert: what about cpu freq? 2011-12-10T00:27:35 3.0ghz as compared to 2.667 2011-12-10T00:28:04 yes 2011-12-10T00:28:13 isn't that a problem? 2011-12-10T00:28:21 I don't think so 2011-12-10T00:28:42 hkraal_ was mentioning that vmware could limit the freq, so we could match the ec2 2011-12-10T00:28:57 they'll be a little faster, but there is variance from game to game anyway 2011-12-10T00:29:05 ok 2011-12-10T00:33:24 happening to be on a noisy box on ec2 can affect your performance quite a bit. Commonly people will spin up more servers than they need, benchmark them, and keep only the fastest ones. 2011-12-10T00:33:42 So there is already a decent amount of variance 2011-12-10T00:35:24 21:28 <@amstan> yes 2011-12-10T00:35:26 oop 2011-12-10T00:35:32 http://aichallenge.org/maps.php 2011-12-10T00:35:37 BenJackson: hmm? 2011-12-10T00:35:39 yay games 2011-12-10T00:35:43 it would be neat if that page had some other stats 2011-12-10T00:35:44 thanks totaalnet 2011-12-10T00:35:46 amstan: just a mispaste 2011-12-10T00:36:00 a1k0n: there's more 2011-12-10T00:36:02 coming 2011-12-10T00:36:08 awesome 2011-12-10T00:36:27 like the mean and variation of the relative scores of people in each position 2011-12-10T00:36:34 that'd be a hell of a query though :) 2011-12-10T00:36:55 BenJackson: if you can make it, sure 2011-12-10T00:37:00 BenJackson: i can test it for you 2011-12-10T00:37:08 to make sure it doesn't take ages 2011-12-10T00:37:11 and it doesn't DoS the server :P 2011-12-10T00:37:35 yes, nothing that a mysql restart and then restart of all workers won't fix 2011-12-10T00:37:51 heh 2011-12-10T00:40:14 resubmitting will be much more exciting now 2011-12-10T00:40:23 congrats & thanks to whoever brought the new sponsors on board 2011-12-10T00:40:31 *** gazpachoking_ is now known as gazpachoking 2011-12-10T00:42:39 haha wow: http://aichallenge.org/visualizer.php?game=178562&user=432 2011-12-10T00:43:06 and that is why you don't need a very dense gas of ants for collecting food 2011-12-10T00:43:20 that's a lot of debug code, lol 2011-12-10T00:44:19 oh man, you're just plowing through them 2011-12-10T00:49:35 I wonder if I should attempt to implement an evenly spread "gas" like that 2011-12-10T00:50:10 what a nice looking map 2011-12-10T00:50:34 some of the starting hills are a little close there 2011-12-10T00:50:46 it's frankly amazing i didn't lose both right away 2011-12-10T00:50:46 not according to the spec :) 2011-12-10T00:51:09 3 per player 2011-12-10T00:51:26 oh right. i did lose 2/3 right away 2011-12-10T00:52:04 though really you're better off without those hills 2011-12-10T00:52:22 better defense for the last? 2011-12-10T00:52:25 *** bmh has joined #aichallenge 2011-12-10T00:52:51 well, once you finally get your economy going into the midgame you can stream ants out of the one hill and thus defend it by default 2011-12-10T00:52:51 a1k0n: are you still using the combat you described? 2011-12-10T00:52:55 the others are just too contested 2011-12-10T00:52:57 i am 2011-12-10T00:53:08 I'm going to try and implement it now 2011-12-10T00:53:21 although, I'm still slightly confused about it 2011-12-10T00:53:24 McLeopold: well I'm going to try to implement a *better* strategy 2011-12-10T00:53:43 everyone seems slightly confused about it, probably including me 2011-12-10T00:53:49 a1k0n: I still think examining one ant at a time is bound to get you stuck at local maxima. 2011-12-10T00:53:55 but whatever mess i made happens to seem to work sorta 2011-12-10T00:54:06 well, naturally 2011-12-10T00:54:32 can I summarize and you tell me where I go wrong? 2011-12-10T00:54:37 sure 2011-12-10T00:54:40 1) pick one ant (yours or enemies) 2011-12-10T00:55:01 so far so good. 2011-12-10T00:55:12 2) try all 5 moves and pick the set of best moves and inc that ants D array by 1 2011-12-10T00:55:19 yep 2011-12-10T00:55:21 for each best move 2011-12-10T00:55:30 3) ??? 2011-12-10T00:55:37 4) profit 2011-12-10T00:55:40 that's it 2011-12-10T00:55:48 3) If you have time left, go to 1 2011-12-10T00:55:59 3) move the ant? 2011-12-10T00:56:02 no, 3) pick a random move according to your new D distribution 2011-12-10T00:56:14 for every ant or just the one? 2011-12-10T00:56:33 so if you have [a,b,c,d,e] you pick move 0 with probability a/(a+b+c+d+e), move 1 with b/(a+b+c+d+e), etc 2011-12-10T00:56:36 for just the one ant 2011-12-10T00:56:39 ok 2011-12-10T00:56:44 3) move the ant! 2011-12-10T00:56:50 yes. 2011-12-10T00:56:55 4) goto 1 2011-12-10T00:57:04 4) maybe look for stabilized ant, otherwise go to 1 2011-12-10T00:57:10 yeah. 2011-12-10T00:57:19 and the magic is in the score? 2011-12-10T00:57:20 if(not over time) go to 1 2011-12-10T00:57:30 :p 2011-12-10T00:57:35 yes. each time you move the ant, you need to resolve combat and check the score. 2011-12-10T00:58:02 so, I assume you tried a score of how many of your ants die? 2011-12-10T00:58:12 if so, how did that do? 2011-12-10T00:58:13 and how many they kill, yes. 2011-12-10T00:58:28 a1k0n have you considered running minimax on "all defensive" vs. "all offensive" then trying to "perturb" the chosen state? 2011-12-10T00:58:37 rather than "perturbing" it from the starting state? 2011-12-10T00:58:39 well, it helps to weigh your ants more heavily than your enemies 2011-12-10T00:58:40 antimatroid: minimax is expensive 2011-12-10T00:58:51 not when you and your oponents have 2 moves each 2011-12-10T00:58:52 antimatroid: i have 2011-12-10T00:59:04 so, you lose 1 is worse than they lose 1? 2011-12-10T00:59:12 yeah. 2011-12-10T00:59:18 a1k0n: don't forget to tell the good people to more readily throw your ants down the oubliette when they're near the hill 2011-12-10T00:59:20 otherwise 1:1 trades are fine, and they really aren't fine 2011-12-10T00:59:38 yes, having some notion of defending your hill in there is probably good 2011-12-10T00:59:40 I have my on secret for deciding on a 1:1 kill :) 2011-12-10T01:00:16 which everyone probably already does :( 2011-12-10T01:00:59 you have less friends than enemies in the battle partition? 2011-12-10T01:01:10 nope, ready for the game changer? 2011-12-10T01:01:12 or you have more ants so you want to take them down? 2011-12-10T01:01:22 the suspense is killing me :P 2011-12-10T01:01:38 it's a miracle i survived this: http://aichallenge.org/visualizer.php?game=178562&user=432&row=42&col=14&turn=165 2011-12-10T01:01:55 the cost of an ant dying is it's distance from its nearest hill 2011-12-10T01:02:15 * McLeopold does punctuation fail 2011-12-10T01:02:23 McLeopold: linear costing seems... too simple? 2011-12-10T01:02:26 wouldn't you want to keep ants alive near your hill? 2011-12-10T01:02:40 I was thinking of using a 1/sqrt(n) costing 2011-12-10T01:02:47 no 2011-12-10T01:02:53 I don't see why 2011-12-10T01:02:58 hill defense 2011-12-10T01:03:01 well, it helps to make last-ditch suicides to keep your hill 2011-12-10T01:03:09 yes, but minimax deals with that :P 2011-12-10T01:03:13 this is for a 1:1 trade only 2011-12-10T01:03:22 good luck with that 2011-12-10T01:03:22 what about n:n trade? 2011-12-10T01:03:25 obviously formations would trump it 2011-12-10T01:03:29 a1k0n: that's epic. 2011-12-10T01:04:26 the time to replace an ant is the walking distance from the hill, if you have a good economy 2011-12-10T01:04:31 minimax only really works if you can exhaustively search your opponents' refutation of the moves you're considering 2011-12-10T01:04:46 so, n:n trades near your hill push the enemy back, all things being equal 2011-12-10T01:05:01 so maybe your idea will work with perturbing the obvious moves 2011-12-10T01:05:04 *** Antimony has quit IRC (Ping timeout: 268 seconds) 2011-12-10T01:05:18 instead of just peturbing blindly like me 2011-12-10T01:05:30 that's my plan, but I'm not perturbing yet 2011-12-10T01:05:41 just running away from conflict 2011-12-10T01:05:50 a1k0n: i pick offensive moves as follows, order ants based on closest enemy, then move into the square that minimises the positive possible number of enemys it can engage in battle with next turn 2011-12-10T01:05:51 hunterbot kicks my but :( 2011-12-10T01:05:59 i do actually have an optional seeding step where i step all my ants towards/away from enemies and my enemy does the same, in order to start the dirichlet distributions off with something 2011-12-10T01:06:14 defensive is order ants based on furthest enemy and move to the square with minimum possible number of battle enemies 2011-12-10T01:06:17 *** media-281 has joined #aichallenge 2011-12-10T01:06:33 whether it helps against other bots is questionable, but it definitely beats not having it in a 1v1 2011-12-10T01:06:38 a1k0n: did you notice a change in the time to stabalize? 2011-12-10T01:07:02 i'm considering adding in "neutral moves" where it minimises static battle enemies 2011-12-10T01:07:12 so as to assume they're not going to try and step into your territory 2011-12-10T01:07:18 McLeopold: not really 2011-12-10T01:07:26 i just check for 95% o the "weight" in one move 2011-12-10T01:07:51 I guess that number depends on the scoring then 2011-12-10T01:08:02 you can also further tiebreak on things like no of unmoved ants that could move into the square or something 2011-12-10T01:08:10 well, maybe, but not really 2011-12-10T01:08:32 omg, my kid is going crazy i beter get him in bed 2011-12-10T01:08:40 do you every update the D array more than 1? 2011-12-10T01:08:57 if two moves have equal score, then both get incremented 2011-12-10T01:09:11 do you split ants into theaters? I'm working on that now 2011-12-10T01:09:16 nope 2011-12-10T01:09:24 but it would probably be a good idea 2011-12-10T01:09:47 but a single move for a single ant never gets more than 1 on an update? 2011-12-10T01:09:50 i think i've now officially spent more idea discussing my implementation than implementing it 2011-12-10T01:09:56 :) 2011-12-10T01:09:57 right 2011-12-10T01:10:00 ok 2011-12-10T01:10:03 off I go... 2011-12-10T01:10:27 that's only because code spills forth from your mind like ambrosia on the firma 2011-12-10T01:12:54 unlike words, which get transposed freely 2011-12-10T01:12:58 s/idea/time/ 2011-12-10T01:13:47 my kid left some effing toy in my car, and can't go to bed without it, and the goddamn sprinklers are on so i can't even go outside without getting wet and why are sprinklers on in december anyway 2011-12-10T01:14:33 a1k0n: in a drought region, no less 2011-12-10T01:19:56 a1k0n: I deny my kid all the plesures of life, so he doesn't know the comforts of going to bed with a toy. no problems so far 2011-12-10T01:20:30 suckers, it's like 30 degrees here :P 2011-12-10T01:20:33 McLeopold: good thinking. My parents spoiled me and now I'm an ungrateful git. 2011-12-10T01:20:40 antimatroid: c or f? 2011-12-10T01:20:46 actually, 24 apparently, but feels warmer i swear 2011-12-10T01:20:47 c 2011-12-10T01:20:53 antimatroid: over here, 30 is freezing 2011-12-10T01:21:20 antimatroid: 'eh, it was around 19c today 2011-12-10T01:21:54 I think it almost made it up to 30f here today ;) 2011-12-10T01:22:08 janzert: CA > MI :) 2011-12-10T01:22:12 yeah but you guys use whack measurements :P 2011-12-10T01:22:15 although it's back into the teens now :/ 2011-12-10T01:22:24 if your not using kelvin, you can talk 2011-12-10T01:22:29 bmh: heh, that's what my brother decided too :P 2011-12-10T01:22:35 and i am :P 2011-12-10T01:22:47 30 kelvin? are you a brain in a jar? 2011-12-10T01:22:48 anyway, off to a bbq 2011-12-10T01:23:00 *** media-281 has quit IRC (Read error: Connection reset by peer) 2011-12-10T01:24:56 on another note, the python for..else makes me happy :) 2011-12-10T01:24:57 McLeopold: what is errors per minute on the server stats page measuring now? cause it surely isn't timeouts and crashes anymore 2011-12-10T01:25:09 or maybe it's simply broken 2011-12-10T01:25:11 I don't know :) 2011-12-10T01:25:17 ok :) 2011-12-10T01:25:26 I think it queried the matchup table 2011-12-10T01:25:36 for games that didn't post 2011-12-10T01:25:48 ahh 2011-12-10T01:25:58 so, it should always be 0 2011-12-10T01:26:05 *** dragonfyre13 has joined #aichallenge 2011-12-10T01:26:08 yeah, for that 2011-12-10T01:26:10 hey there everyone 2011-12-10T01:26:29 Quick question if someone knows. Any idea of numexpr is able to be used on the servers? 2011-12-10T01:26:30 I used to use it to make sure a worker wasn't timing out every bot 2011-12-10T01:28:37 dragonfyre13: I'd be rather surprised if it was 2011-12-10T01:28:58 darn. Wonder if me submitting a request for it at this point would do any good.... 2011-12-10T01:29:06 you can always try submitting a quick bot that imports and tries to use it to check though 2011-12-10T01:29:16 janzert: yeah, that's what I was going to try next 2011-12-10T01:29:20 janzert: thanks 2011-12-10T01:29:48 yeah, with only a week to go I'm probably not going to make much of any language changes 2011-12-10T01:31:25 but janzert, I program brilliantly fast in FrTime. Please consider supporting it! 2011-12-10T01:31:35 :) 2011-12-10T01:32:37 The purpose of Lisp is not to enable programming. It's to enable the creation of a language where your program can be expressed in a single statement. 2011-12-10T01:34:19 janzert: thought this was pretty much just running an ubuntu server (I know next to nothing about it, something I thought I read on the forums) 2011-12-10T01:34:27 sudo apt-get install python-numexpr 2011-12-10T01:36:10 *** Ttech is now known as PhthaloFox 2011-12-10T01:38:46 *** mindeavor has joined #aichallenge 2011-12-10T01:39:21 janzert: uploaded, we'll see if it runs. 2011-12-10T01:40:15 ImportError: No module named numexpr 2011-12-10T01:40:17 yay 2011-12-10T01:41:49 I can now group ants http://paste.aichallenge.org/zyslQ/ 2011-12-10T01:42:45 McLeopold: did I just watch you lose to HunterBot? 2011-12-10T01:42:54 ignore that part :) 2011-12-10T01:43:39 question how does one add visualization information 2011-12-10T01:43:48 i remember reading on the forum but can't remember 2011-12-10T01:44:02 bhasker: you should start by cloning the j-h-a aichallenge repo from github 2011-12-10T01:44:15 the instructions are all there 2011-12-10T01:44:16 hmm k let me see 2011-12-10T01:44:22 McLeopold: want to compete against me on my server in a few minutes? :) 2011-12-10T01:44:29 sure, but I'm going to lose 2011-12-10T01:45:53 just need to get properly LOS maintenence working 2011-12-10T01:54:50 bmh: ready? 2011-12-10T01:54:59 McLeopold: I can only code so fast... 2011-12-10T01:55:05 I'll just run the older version 2011-12-10T01:55:08 yeah 2011-12-10T01:55:14 that will be more fun that waiting 2011-12-10T01:55:21 what's your server name? 2011-12-10T01:55:37 bhickey.net 2081 -- hang on though 2011-12-10T01:55:39 BenJackson: you there? 2011-12-10T01:55:58 guess not, ok, launching 2011-12-10T01:56:13 damn, got paired with Ben's bot 2011-12-10T01:56:34 better! 2011-12-10T01:56:52 I'm playing a game now... 2011-12-10T01:57:03 yep, that's me 2011-12-10T01:57:05 normally I timeout after about 100 moves 2011-12-10T01:57:23 ohh, slowing down... 2011-12-10T01:57:42 if it gets faster, it's cause your killing my ants :) 2011-12-10T01:58:00 this bot zergs. 2011-12-10T01:58:08 which one? 2011-12-10T01:58:17 skynet, as of this morning 2011-12-10T01:59:43 I have 41 ants total in my view, including enemy, that can't be good... 2011-12-10T02:00:22 *** dvladim has joined #aichallenge 2011-12-10T02:01:40 maybe I'll time out 2011-12-10T02:01:46 I'm slowing down 2011-12-10T02:01:48 I don't actually check the clock 2011-12-10T02:01:51 it's my exploration code 2011-12-10T02:02:14 it's still the tutorial version :( 2011-12-10T02:03:09 http://bhickey.net:2080/replay.4891 2011-12-10T02:03:22 wow 2011-12-10T02:03:39 it looks so much better when watching the debug code :( 2011-12-10T02:03:48 that's the ghastly dumb version of my bot 2011-12-10T02:03:56 which happens to be my bot... 2011-12-10T02:04:58 I have know idea what happened there 2011-12-10T02:06:09 *** Palmik has joined #aichallenge 2011-12-10T02:09:05 antimatroid around? 2011-12-10T02:09:45 bhasker: he's at a BBQ. 2011-12-10T02:09:52 ah k 2011-12-10T02:12:42 bmh, thanks for the game, I've got some bugs to find 2011-12-10T02:12:54 McLeopold: goodluck, I've got my own debugging 2011-12-10T02:13:02 maybe later tonight 2011-12-10T02:16:06 bmh: btw, that is an invalid map 2011-12-10T02:16:35 McLeopold: oh? What's wrong with it? 2011-12-10T02:16:47 *** dvladim has quit IRC (Ping timeout: 244 seconds) 2011-12-10T02:17:29 bmh: probably starting hill distances, it was removed from official 2011-12-10T02:22:00 ah 2011-12-10T02:31:22 *** zyberkiddy has joined #aichallenge 2011-12-10T02:31:25 *** dvladim has joined #aichallenge 2011-12-10T02:34:42 *** Scryer has quit IRC (Quit: Page closed) 2011-12-10T02:36:38 *** amstan has quit IRC (Quit: Konversation terminated!) 2011-12-10T02:36:42 *** amstan_ has joined #aichallenge 2011-12-10T02:36:46 *** ChanServ sets mode: +o amstan_ 2011-12-10T02:39:35 *** codetiger has joined #aichallenge 2011-12-10T02:41:23 *** codetiger has quit IRC (Remote host closed the connection) 2011-12-10T02:44:11 *** jeffymj has joined #aichallenge 2011-12-10T02:45:55 bmh: so, the bug was an incorrect indentaion :( 2011-12-10T02:46:07 McLeopold: I 2011-12-10T02:46:13 There are reasons that I don't use Python. 2011-12-10T02:46:54 it was actually logical, I but a line that removed an unseen location form the list to check for in the wrong spot 2011-12-10T02:46:59 not sure about starting hill distance on that map, but for sure total area and per player area are both too big 2011-12-10T02:47:08 it was something 2011-12-10T02:47:30 total area is just over the limit at 25935 and per player is at 12967.5 2011-12-10T02:48:00 bmh: okay, ready for another match? 2011-12-10T02:48:15 *** dvladim has quit IRC (Read error: Operation timed out) 2011-12-10T02:53:34 *** zyberkiddy has left #aichallenge 2011-12-10T02:54:29 McLeopold: not yet! :) 2011-12-10T02:56:53 *** epicmonkey has joined #aichallenge 2011-12-10T02:58:31 okay, I've got some more to do too :) 2011-12-10T03:12:17 McLeopold: ready when you are 2011-12-10T03:12:34 *** dvladim has joined #aichallenge 2011-12-10T03:13:36 *** sigh has quit IRC (Remote host closed the connection) 2011-12-10T03:14:34 *** Jak_o_Shadows has joined #aichallenge 2011-12-10T03:19:07 *** dom7b5 has quit IRC (Ping timeout: 240 seconds) 2011-12-10T03:19:41 whoa. The student has become the master. Skynet beats BenJackson v4 2011-12-10T03:20:12 *** dom7b5 has joined #aichallenge 2011-12-10T03:21:04 McLeopold: http://bhickey.net:2080/replay.4952 2011-12-10T03:21:55 looking... 2011-12-10T03:22:35 I got the a1k0n "hill defence" strategy working 2011-12-10T03:22:54 okay, I'm on 2011-12-10T03:23:08 and what is the hill defense strategy? 2011-12-10T03:23:54 maximize the walking distance from each hill to the nearest square that isn't visible 2011-12-10T03:25:12 oh, so don't keep ants close if not needed 2011-12-10T03:26:13 why'd you time out against bj? 2011-12-10T03:27:16 don't know 2011-12-10T03:27:28 well, my code attempted to move an ant twice 2011-12-10T03:27:33 don't know why 2011-12-10T03:27:49 probably a future ant that got the wrong time value 2011-12-10T03:29:11 future ants? What manner of sorcery is this!? 2011-12-10T03:32:17 it's my way of avoiding game trees 2011-12-10T03:33:17 look at that game I posted on paste where I lost to Hunter 2011-12-10T03:33:21 http://paste.aichallenge.org/zyslQ/ 2011-12-10T03:33:39 turn on my visualizations, you can see 2 white lines coming from the hill to other food 2011-12-10T03:33:46 oh, whoa. state 2011-12-10T03:33:49 those are ants that will spawn 2011-12-10T03:34:05 I don't do any planning. 2011-12-10T03:34:27 I looks more impressive with more food 2011-12-10T03:34:50 I can plan for an ant to gather several food items, and you can see its path 2011-12-10T03:34:59 it frees up other ants to do other things 2011-12-10T03:36:06 I need to incorporate "distance from the hill" into my scores somehow 2011-12-10T03:37:01 just wait until I get my quantum ants working :) 2011-12-10T03:49:39 *** bmh has quit IRC (Ping timeout: 258 seconds) 2011-12-10T03:50:39 *** raemde has joined #aichallenge 2011-12-10T03:53:10 *** raemde_ has quit IRC (Ping timeout: 248 seconds) 2011-12-10T04:11:10 *** g0llum has joined #aichallenge 2011-12-10T04:12:00 *** dvladim has quit IRC (Ping timeout: 244 seconds) 2011-12-10T04:13:06 Ah nice, I see janzert put the new servers up :) 2011-12-10T04:14:44 is BenJackson around? 2011-12-10T04:27:11 *** pairofdice has joined #aichallenge 2011-12-10T04:33:27 *** dom7b5 has quit IRC (Ping timeout: 240 seconds) 2011-12-10T04:41:34 *** jeffymj has quit IRC (Read error: Connection reset by peer) 2011-12-10T04:42:06 *** jeffymj has joined #aichallenge 2011-12-10T04:47:30 *** dom7b5 has joined #aichallenge 2011-12-10T04:52:47 *** dom7b5 has quit IRC (Ping timeout: 240 seconds) 2011-12-10T04:56:25 *** foRei has joined #aichallenge 2011-12-10T04:59:13 *** foRei has quit IRC (Read error: Connection reset by peer) 2011-12-10T05:00:11 *** foRei has joined #aichallenge 2011-12-10T05:20:46 *** LouisMartin has joined #aichallenge 2011-12-10T05:22:35 On the ~12,000 registered contestant, how many would you guess are serious about being competitive? 2011-12-10T05:22:42 *** jeffymj has quit IRC (Quit: Ex-Chat) 2011-12-10T05:30:15 I'm at # 2500 and I'm not to serious so I would gues around 1500? 2011-12-10T05:31:50 *** LouisMartin has quit IRC (Quit: Page closed) 2011-12-10T05:33:05 LouisMartin, i think this is the watershed: http://aichallenge.org/profile.php?user=12359 2011-12-10T05:59:43 *** dom7b5 has joined #aichallenge 2011-12-10T06:11:44 *** Anilm3 has joined #aichallenge 2011-12-10T06:16:43 *** Anilm3 has quit IRC (Quit: leaving) 2011-12-10T06:23:15 *** Antimony has joined #aichallenge 2011-12-10T06:23:22 *** ikaros has joined #aichallenge 2011-12-10T06:35:13 *** Kurnevsky has joined #aichallenge 2011-12-10T06:39:23 *** mj41 has joined #aichallenge 2011-12-10T06:39:28 *** Kurnevsky has quit IRC (Ping timeout: 240 seconds) 2011-12-10T06:41:28 Git is awesome. 2011-12-10T06:44:28 *** Donnetje has joined #aichallenge 2011-12-10T06:45:18 Hi 2011-12-10T06:48:01 hi 2011-12-10T07:03:16 *** iglo has joined #aichallenge 2011-12-10T07:03:52 *** Antimony has quit IRC (Ping timeout: 240 seconds) 2011-12-10T07:06:44 *** Redgis has joined #aichallenge 2011-12-10T07:08:47 *** delt0r_ has quit IRC (Ping timeout: 240 seconds) 2011-12-10T07:09:31 *** olexs has joined #aichallenge 2011-12-10T07:09:56 *** Kurnevsky has joined #aichallenge 2011-12-10T07:11:01 *** Donnetje has quit IRC (Quit: Page closed) 2011-12-10T07:11:59 *** Redgis has quit IRC (Ping timeout: 268 seconds) 2011-12-10T07:18:50 *** pairofdice has quit IRC (Quit: leaving) 2011-12-10T07:21:20 *** delt0r_ has joined #aichallenge 2011-12-10T07:23:07 *** pairofdice has joined #aichallenge 2011-12-10T07:24:37 *** flowenol has joined #aichallenge 2011-12-10T07:25:49 I think you should have a bot with skill at least 60-70 if you're serious 2011-12-10T07:26:12 at least if you've been in the contest for a while 2011-12-10T07:29:02 *** Jak_o_Shadows has quit IRC (Read error: Connection reset by peer) 2011-12-10T07:32:55 *** kire has joined #aichallenge 2011-12-10T07:36:41 *** dvladim has joined #aichallenge 2011-12-10T07:39:07 *** mleise has joined #aichallenge 2011-12-10T07:42:02 *** kire has quit IRC (Read error: Connection reset by peer) 2011-12-10T07:43:04 yeah, sounds reasonable 2011-12-10T07:43:21 tutorial bot + minimal routing around obstacles should be around 60 2011-12-10T07:48:37 At 60 you have way more than tutorial bot 2011-12-10T07:49:38 Well, 60 is rank 1143 2011-12-10T07:49:57 It's as good a cutoff as any I guess 2011-12-10T07:50:59 *** foRei has quit IRC (Ping timeout: 252 seconds) 2011-12-10T07:51:46 This Java starter is confusing me too much 2011-12-10T07:56:41 it's pretty straightforward 2011-12-10T07:57:54 *** raemde_ has joined #aichallenge 2011-12-10T07:58:59 *** bluegaspode has joined #aichallenge 2011-12-10T08:01:07 *** raemde has quit IRC (Ping timeout: 252 seconds) 2011-12-10T08:01:52 judging by that starter packs ive seen the java one had the highest quality 2011-12-10T08:03:43 *** Anilm3 has joined #aichallenge 2011-12-10T08:05:48 *** replore_ has joined #aichallenge 2011-12-10T08:08:15 *** raemde_ is now known as raemde 2011-12-10T08:17:46 *** bluegaspode has quit IRC (Quit: Page closed) 2011-12-10T08:19:53 *** jeffymj has joined #aichallenge 2011-12-10T08:21:07 *** dom7b5 has quit IRC (Ping timeout: 240 seconds) 2011-12-10T08:22:46 *** dom7b5 has joined #aichallenge 2011-12-10T08:23:26 *** Redgis has joined #aichallenge 2011-12-10T08:27:09 *** iglo has quit IRC (Remote host closed the connection) 2011-12-10T08:40:34 *** jbob_ has joined #aichallenge 2011-12-10T08:47:04 *** dvladim has quit IRC (Ping timeout: 240 seconds) 2011-12-10T08:51:26 *** Dlayne has joined #aichallenge 2011-12-10T08:55:03 wow a million servers :D 2011-12-10T09:01:20 *** mleise1 has joined #aichallenge 2011-12-10T09:01:21 *** mleise has quit IRC (Read error: Connection reset by peer) 2011-12-10T09:01:53 Oh, nice 2011-12-10T09:12:14 *** Redgis has quit IRC (Ping timeout: 268 seconds) 2011-12-10T09:16:22 Anilm3: there will be more ;) 2011-12-10T09:17:05 I work for the server sponsor and the systems can have at least 12 more servers :) 2011-12-10T09:17:21 prolly even 18 more 2011-12-10T09:24:27 *** delt0r_ has quit IRC (Ping timeout: 240 seconds) 2011-12-10T09:29:39 *** delt0r_ has joined #aichallenge 2011-12-10T09:30:10 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has joined #aichallenge 2011-12-10T09:34:01 *** flowenol has quit IRC (Ping timeout: 252 seconds) 2011-12-10T09:44:18 *** epicmonkey has quit IRC (Ping timeout: 268 seconds) 2011-12-10T09:47:47 *** dom7b5 has quit IRC (Ping timeout: 240 seconds) 2011-12-10T09:49:16 *** dom7b5 has joined #aichallenge 2011-12-10T09:49:21 *** jbob_ has quit IRC (Quit: jbob_) 2011-12-10T09:49:50 *** Redgis has joined #aichallenge 2011-12-10T09:50:48 *** g0llum has quit IRC (Read error: Connection reset by peer) 2011-12-10T09:56:02 *** kilae has joined #aichallenge 2011-12-10T09:57:10 *** kilae has quit IRC (Client Quit) 2011-12-10T09:57:42 *** kilae has joined #aichallenge 2011-12-10T10:01:17 *** besh has quit IRC (Quit: Page closed) 2011-12-10T10:02:11 *** Redgis has quit IRC (Ping timeout: 268 seconds) 2011-12-10T10:05:06 *** jstrong has joined #aichallenge 2011-12-10T10:07:11 hm so i massively improved my combat 2011-12-10T10:07:17 problem? I now easily get shoehorned into my base 2011-12-10T10:07:25 if I don't get shoehorned I do really well. 2011-12-10T10:07:27 but if I do 2011-12-10T10:08:48 *** besh has joined #aichallenge 2011-12-10T10:09:40 *** Antimony has joined #aichallenge 2011-12-10T10:09:57 hkraal_: that's great!! 2011-12-10T10:14:41 *** avdg has joined #aichallenge 2011-12-10T10:14:46 oldman is doing well again 2011-12-10T10:24:03 needs to rename his bot to skeletron 2011-12-10T10:24:19 *** Garf has quit IRC (Quit: Make a new plan, Stan!) 2011-12-10T10:24:33 *** Garf has joined #aichallenge 2011-12-10T10:25:02 *** sigh has joined #aichallenge 2011-12-10T10:25:13 *** mleise has joined #aichallenge 2011-12-10T10:25:16 it would be nice that each user could have multiple bots 2011-12-10T10:25:34 *** mleise1 has quit IRC (Quit: Leaving.) 2011-12-10T10:30:27 *** mj41 has quit IRC (Ping timeout: 252 seconds) 2011-12-10T10:32:43 *** dvladim has joined #aichallenge 2011-12-10T10:35:46 do you treat all enemies as "one player" in payoff matrix, or each independently? 2011-12-10T10:36:14 if you implement payoff matrix* 2011-12-10T10:38:33 Fluxid: all as one enemy 2011-12-10T10:38:55 i'm not really constructing a tree or payoff matrix, but you could really think of it as either 2011-12-10T10:38:59 memetix doesn't I think 2011-12-10T10:39:19 in the payoff matrix case i'm just picking the strategy which maximises my worst outcome 2011-12-10T10:39:33 altough I don't know if what he does is a payoff matrix 2011-12-10T10:40:16 *** Antimony has quit IRC (Ping timeout: 240 seconds) 2011-12-10T10:40:42 i've just finished rewriting all my battle code stuff so that i don't copy state 2011-12-10T10:40:46 because that was an obvious fail 2011-12-10T10:40:49 it's testing time 2011-12-10T10:44:11 Fluxid: how many moves are you considering for your ants? 2011-12-10T10:44:25 i basically pick out 2 sets of moves for my ants, and 2 sets of moves for their ants 2011-12-10T10:44:39 and a set of moves is a move for each of my or each of their ants 2011-12-10T10:44:45 so my game is really 2x2 2011-12-10T10:45:13 although it's written so i can just throw more sets of moves in if i want/can afford the time 2011-12-10T10:45:32 antimatroid: yeah, I do that 2011-12-10T10:45:40 thestinger: i know :P 2011-12-10T10:45:46 it seems all about the positioning though 2011-12-10T10:45:50 I need to get that working well 2011-12-10T10:46:00 as long as your ants don't suicide, you'll do fine with basic combat 2011-12-10T10:46:02 yep, i've worked out how i'm going to do that 2011-12-10T10:46:09 no idea how well it'll do, but it's waht i'm going with 2011-12-10T10:46:31 i just want to finish my bot and submit it 2011-12-10T10:48:19 *** pguillory has joined #aichallenge 2011-12-10T10:54:37 Anilm3: that would require lots more servers to keep the frequency of the games on acceptable levels. Besides I would think you wanted to combine all your programming efforts on a single bot? ;) 2011-12-10T10:57:53 well I just meant it for the purpose of evaluating and debugging 2011-12-10T10:58:17 the tcp servers are kind of brutal for some of os 2011-12-10T10:58:24 us 2011-12-10T10:58:59 ahh right, yeah I can see where your going :) 2011-12-10T10:59:09 but the ranking would be pretty useless then 2011-12-10T10:59:36 *** xathis has joined #aichallenge 2011-12-10T10:59:40 It would indeed be nice to have 10 game slots / day or something so you can release your new bot ahead in the queue 2011-12-10T11:00:07 *** amstan_ has quit IRC (Ping timeout: 240 seconds) 2011-12-10T11:00:09 yeah 2011-12-10T11:00:24 I've had quite a number of games today though 2011-12-10T11:00:46 I expect that a bunch of other servers will be launched as well. I recon the time between games could only be 1 hour if all are running 2011-12-10T11:01:03 so, that won't be to bad :) 2011-12-10T11:01:05 9 games 2011-12-10T11:01:21 and another one is near 2011-12-10T11:02:01 its already much better yeah :) 2011-12-10T11:02:18 *** avdg has quit IRC (Quit: Leaving.) 2011-12-10T11:02:31 *** Jacob_Strauss has joined #aichallenge 2011-12-10T11:02:42 best thing is that if prices in the Amazon cloud are going up these servers will continue to run :) 2011-12-10T11:03:12 that's good 2011-12-10T11:03:52 Anilm3: 4x5 2011-12-10T11:04:50 Fluxid: 4x5 what? 2011-12-10T11:04:54 matrix 2011-12-10T11:05:00 or whatever 2011-12-10T11:05:05 at least ATM 2011-12-10T11:05:14 I'm lost 2011-12-10T11:05:33 Fluxid dont you mean antimatroid ? 2011-12-10T11:05:50 uh 2011-12-10T11:05:54 right! 2011-12-10T11:05:59 good morning ;) 2011-12-10T11:06:05 it all makes sense now 2011-12-10T11:06:16 sorry, an[tab] ;) 2011-12-10T11:06:21 ghehe 2011-12-10T11:06:23 antimatroid: i use 4x5 2011-12-10T11:06:35 that's a strange number? 2011-12-10T11:07:07 i consider 4 moves of enemy and halt, and four own moves 2011-12-10T11:07:17 so for enemy it gives 5 options 2011-12-10T11:07:46 Fluxid: ah, I need to switch to something like that :) 2011-12-10T11:07:55 when I tried before my bot became too cautious 2011-12-10T11:07:55 Fluxid: normally you are the row player :P 2011-12-10T11:08:01 but what i have now isn't real payoff matrix or whatever, i'm trying to implement it now 2011-12-10T11:08:03 I have to keep it stupid and ignorant or it won't attack... 2011-12-10T11:08:23 antimatroid: i never had this area of game theory on uni 2011-12-10T11:08:24 or I could change the way kills/deaths are weighted 2011-12-10T11:08:40 the game theory being used here is fairly basic :P 2011-12-10T11:08:46 but that's just convention 2011-12-10T11:08:50 so i'm using terms learned on this channel and wikipedia 2011-12-10T11:08:59 imo the convention is WRONG when it comes to n>3 player games 2011-12-10T11:09:03 n>2* 2011-12-10T11:10:05 *** pguillory has quit IRC (Quit: Page closed) 2011-12-10T11:10:26 but it's funny when you do something, and then you learn how people call it. 2011-12-10T11:11:05 yup 2011-12-10T11:12:11 *** flowenol has joined #aichallenge 2011-12-10T11:20:01 *** g0llum has joined #aichallenge 2011-12-10T11:20:19 *** dvladim has quit IRC (Ping timeout: 244 seconds) 2011-12-10T11:20:49 hrmm, would it be against the symmetry requirements to have a (for ex) 4-player map with translational symmetry and the top half and bottom half are the same but the bottom half is offset 1/2 player space to the left, so hive to hive would be a path like /\/ 2011-12-10T11:21:33 The mapgenerator uses that type AFAIK 2011-12-10T11:25:31 *** rwest_ has joined #aichallenge 2011-12-10T11:25:39 *** avdg has joined #aichallenge 2011-12-10T11:25:42 ugh for some reason, all of my bots started timing out 2011-12-10T11:26:01 hmmm has to be a java issue 2011-12-10T11:26:13 ls 2011-12-10T11:26:13 haha 2011-12-10T11:26:27 like even my old bot which was taking 60ms to compute 2011-12-10T11:27:56 *** kire has joined #aichallenge 2011-12-10T11:32:33 i see some bots shake their ants during standoffs. can i swap positions of two ants? (which is equivalent to not moving both ants) 2011-12-10T11:32:41 yes 2011-12-10T11:33:39 so i have to modify isOccupied() to include the case wher the ant will move in the future 2011-12-10T11:34:12 my bot never produces that kind of move 2011-12-10T11:34:17 Extrarius: have you taken a look at how translationally symmetric maps are currently done? 2011-12-10T11:34:44 things are basically picked such that translate^n(location) = location for all locations 2011-12-10T11:36:45 *** mj41 has joined #aichallenge 2011-12-10T11:37:35 what types of symmetry are being used? 2011-12-10T11:38:56 besh, switching the locations of two ants is never strictly needed 2011-12-10T11:39:21 how do you people mix combat with gathering food?? 2011-12-10T11:41:07 I'm very stuck with that 2011-12-10T11:41:12 :-( 2011-12-10T11:41:20 *** sigh has quit IRC (Remote host closed the connection) 2011-12-10T11:41:54 pairofdice: yes. but it looks cool to the observer :) 2011-12-10T11:42:57 Anilm3: i do not mix them. I do combat first when it is clear i can kill some ants. Then foraging with remaing ants.. 2011-12-10T11:43:25 You don't get points for artistic aspects :) 2011-12-10T11:43:50 besh: that's the strategy I was thinking about but I think it could be better if it could be mixed 2011-12-10T11:45:05 Well i used to do it based on priorites but it screws up combat if one of the ants flies away to gather food or even explore unseen tile. 2011-12-10T11:45:46 haha I see 2011-12-10T11:46:20 pairofdice: maybe some attention :p 2011-12-10T12:00:02 *** Anilm3 has quit IRC (Quit: goodbye) 2011-12-10T12:12:04 *** bretep has joined #aichallenge 2011-12-10T12:21:53 How can you find older pastes on paste.aichallenge.org? 2011-12-10T12:30:21 *** yoden has joined #aichallenge 2011-12-10T12:31:05 *** nha has joined #aichallenge 2011-12-10T12:33:51 bretep: what are you looking for? 2011-12-10T12:35:50 *** flowenol has quit IRC (Remote host closed the connection) 2011-12-10T12:36:09 *** flowenol has joined #aichallenge 2011-12-10T12:37:44 *** Anilm3 has joined #aichallenge 2011-12-10T12:37:52 besh: what's your name in the site? 2011-12-10T12:42:47 McLeopold: I'm looking for one of those big maps fluxid posted on it 2011-12-10T12:44:06 *** delt0r_ has quit IRC (Remote host closed the connection) 2011-12-10T12:44:53 *** Antimony has joined #aichallenge 2011-12-10T12:51:48 *** amriedle has joined #aichallenge 2011-12-10T12:54:56 bretep: I'm working on paging now 2011-12-10T12:56:05 *** Conorach has joined #aichallenge 2011-12-10T12:58:28 *** amstan_ has joined #aichallenge 2011-12-10T12:58:28 *** ChanServ sets mode: +o amstan_ 2011-12-10T13:00:20 *** bmh has joined #aichallenge 2011-12-10T13:02:07 *** Israfel has quit IRC (Ping timeout: 240 seconds) 2011-12-10T13:07:12 *** PhthaloFox is now known as Ttech 2011-12-10T13:12:23 *** olexs has quit IRC (Ping timeout: 252 seconds) 2011-12-10T13:14:02 *** Israfel has joined #aichallenge 2011-12-10T13:14:07 amstan_: great job with the servers!! 2011-12-10T13:14:22 Anilm3: congratulate janzert too 2011-12-10T13:14:32 janzert: great job with the servers!! 2011-12-10T13:15:02 I'm having a game per hour now 2011-12-10T13:16:07 holy cow, 19 servers 2011-12-10T13:16:23 haha 2011-12-10T13:17:20 amstan_: I believe it works out to be a succes :) 2011-12-10T13:17:30 yep, cool stuff 2011-12-10T13:17:50 hkraal_: thanks for your initiative in getting this going 2011-12-10T13:18:08 oh hkraal_ great job you too! xD 2011-12-10T13:18:14 overal load is around 30%... so janzert can add quite a few more :D 2011-12-10T13:18:39 i am doing my testing on the official server now ;) 2011-12-10T13:18:50 besh: what's your name on official? 2011-12-10T13:18:57 dshawul 2011-12-10T13:19:01 a1k0n: I got your combat implemented 2011-12-10T13:19:02 amstan_: sure np... I really hate sweet hardware doing nothing 2011-12-10T13:19:04 oh that's you! 2011-12-10T13:19:15 does it work at all? 2011-12-10T13:19:21 a little 2011-12-10T13:19:25 heh 2011-12-10T13:19:34 what do you do if an ant move is occupied? 2011-12-10T13:19:34 yeah i flood the server with 22 versions so far 2011-12-10T13:19:48 I can only score 4 of the five positions 2011-12-10T13:19:54 i don't consider it for that sample 2011-12-10T13:19:56 or I have to move the other ant 2011-12-10T13:20:11 but then, that move won't be picked as the best 2011-12-10T13:20:17 right 2011-12-10T13:20:17 and can't be incremented 2011-12-10T13:20:49 so you give it a second-best move if it can't go its preferred way 2011-12-10T13:20:59 McLeopold: what details on the servers do you want? :) 2011-12-10T13:21:00 so, my bot does do well in 2v1 2011-12-10T13:21:25 hkraal_: more from our perspective, what the plans are for finals 2011-12-10T13:21:42 As I understand the finals will be run on Amazon 2011-12-10T13:22:39 That to make sure to tampering can be done. For the time to the finals these extra servers are used to lower the time between games 2011-12-10T13:22:51 *no tampering 2011-12-10T13:22:54 I'd implement monte carlo combat, but I really don't want to implement "find the nearby ants" 2011-12-10T13:22:56 * bmh is lazy 2011-12-10T13:23:38 a1k0n: http://paste.aichallenge.org/8BvMQ/ 2011-12-10T13:23:47 turn on the visualizations 2011-12-10T13:24:06 the magenta lines connect ant groups that I simulate individually 2011-12-10T13:24:17 the blue lines are the suggested combat moves 2011-12-10T13:24:19 Do you think being more aggresive attacking hills is better with many players ? 2011-12-10T13:24:23 a blue circle means hold 2011-12-10T13:25:19 turn 27 has a best move as hold, when it should be south 2011-12-10T13:25:36 also, I ignore 1v1 for now 2011-12-10T13:26:14 the yellow/red highlights are using the simpler combat method to mark KILL and DIE squares 2011-12-10T13:26:29 you also have the influence map stuff? 2011-12-10T13:26:33 ah 2011-12-10T13:27:00 what are the white circles around your hill? 2011-12-10T13:27:26 hive count 2011-12-10T13:27:34 to make sure I'm calculating it right 2011-12-10T13:27:46 ah, nice 2011-12-10T13:28:06 turn 34, I get it right, but I think the ants could have moved together 2011-12-10T13:28:40 the stars are spots I can't see, but have possible food/hills 2011-12-10T13:29:06 and, as you can see, I commit everything to razing hills, and don't have defense :) (which is really bad, btw) 2011-12-10T13:29:10 neat 2011-12-10T13:29:20 so, what's my problem? :P 2011-12-10T13:30:09 oh, and my score is just ant count differences 2011-12-10T13:30:15 dunno. clearly it really mispredicted turn 27 and not because of blocking 2011-12-10T13:30:39 I can paste the D updates 2011-12-10T13:31:25 Poor pguillory http://aichallenge.org/visualizer.php?game=181730&user=589 2011-12-10T13:33:40 a1k0n: http://pastebin.com/igreAdHh 2011-12-10T13:34:30 hmm, hold, e and s all got the same score 2011-12-10T13:34:31 *** mj41 has quit IRC (Ping timeout: 255 seconds) 2011-12-10T13:34:53 Antimony: poor him what? He's not aggressive enough 2011-12-10T13:35:28 *** Antimony has quit IRC (Ping timeout: 240 seconds) 2011-12-10T13:35:45 looks like both use same algorithm though 2011-12-10T13:36:42 you don't have nearly enough samples 2011-12-10T13:37:22 I'll up it and see what happens 2011-12-10T13:37:30 *** hkraal_ is now known as hkraal^afk 2011-12-10T13:37:51 *** UncleVasya has joined #aichallenge 2011-12-10T13:37:54 you probably want some weak attraction between ants in order to tiebreak 2011-12-10T13:38:19 at 100, it's still the same, but I think I get it now 2011-12-10T13:38:37 I'm scoring a survive and a 1v1 both as 0 2011-12-10T13:39:56 a1k0n: what is weak attraction? 2011-12-10T13:40:06 score higher if my ants are in groups? 2011-12-10T13:40:34 McLeopold: all things being equal, prefer moves that pull your ants together 2011-12-10T13:42:19 or towards enemies 2011-12-10T13:43:12 hmm, it doesn't mess up when I change the scoring, but it takes a little longer, and my ant graph doesn't go down as much 2011-12-10T13:43:29 thx for the help 2011-12-10T13:43:40 too bad I'm already disqualified :( 2011-12-10T13:44:43 so, I can beat hunter pretty good now, but if I play myself, it's nothing but stalemates 2011-12-10T13:45:54 McLeopold: want to roll over skynet? 2011-12-10T13:46:07 McLeopold: you can not participate in the contest,right? that sucks :( 2011-12-10T13:54:58 *** amriedle has quit IRC (Quit: Lost terminal) 2011-12-10T13:55:42 00:21 < bmh> McLeopold: http://bhickey.net:2080/replay.4952 2011-12-10T13:55:54 hahaa the hill battle at 0,128 at the start 2011-12-10T13:56:04 we suicided like 15 ants into each other there 2011-12-10T13:56:19 uh.. is that the right game? 2011-12-10T13:56:37 oh, got it 2011-12-10T13:57:18 Skynet is aggressive. 2011-12-10T13:59:16 McLeopold: you don't participate on the game? 2011-12-10T14:00:34 Anilm3: no 2011-12-10T14:00:40 I can't, I'm a contest organizer 2011-12-10T14:00:58 oh, I'm sorry 2011-12-10T14:01:04 bmh: okay, firing up the bot 2011-12-10T14:01:14 BenJackson: I think I might add combat and call it done 2011-12-10T14:01:47 I just happened to watch that corner 2011-12-10T14:01:55 as my hill was saved over and over by suicides 2011-12-10T14:02:00 BenJackson: is version 4 really old? 2011-12-10T14:02:24 4 Oct 30th 4:26am Deactivated: not in queue to play game 64.64 20 345 C++ 2011-12-10T14:02:38 64.64 didn't mean then what it does now 2011-12-10T14:02:43 probably 2011-12-10T14:02:44 so, that's the version I'm playing on tcp? 2011-12-10T14:02:52 on bmh's server 2011-12-10T14:03:07 bmh: what's your server url? 2011-12-10T14:03:17 Anilm3: bhickey.net:2080 / 2081 (tcp) 2011-12-10T14:03:19 yes many bots have got stronger past couple of days. Top 100 is tough now 2011-12-10T14:03:46 wow, you need 78.5 to get in the top 100 2011-12-10T14:04:28 wow for me 2011-12-10T14:04:30 FlagCapper just resubmitted 2011-12-10T14:04:40 i am doing what i can to get in there even though i was upto 35 some time back 2011-12-10T14:05:12 a1k0n: do you think you'll make top 10? 2011-12-10T14:05:12 so if those ranks were directly comparable, my V4 would be 783rd 2011-12-10T14:05:17 I'm sure it was higher than that at the time 2011-12-10T14:05:18 i'm in #10 right now, heh 2011-12-10T14:05:30 oh, I thought you were at 20... 2011-12-10T14:05:37 i had just resubmitted too 2011-12-10T14:05:38 @rankings 2011-12-10T14:05:39 Anilm3: Top 10 players: pguillory(92.7), xathis(91.8), ChrisH(90.2), lazarant(88.7), GreenTea(88.6), delineate(88.3), Memetix(88.0), fourmidable(87.9), RVeerdonk(87.4), a1k0n(86.0) 2011-12-10T14:05:41 still working my way up 2011-12-10T14:05:49 this game kinda blows my mind though: http://aichallenge.org/visualizer.php?game=179213&user=31 2011-12-10T14:06:05 he didn't even try to capture my hill, he just picked off all my ants one by one 2011-12-10T14:06:15 yeah, I saw that one 2011-12-10T14:06:21 on the same map I'm getting such similar games that the same battles play out the same way at each hill 2011-12-10T14:06:23 it looks chaotic 2011-12-10T14:07:10 a1k0n: "flag capper" yeah right! 2011-12-10T14:07:54 well, I can beat BenJackson v4... 2011-12-10T14:07:57 *** Antimony has joined #aichallenge 2011-12-10T14:07:59 trivia: flagcapper is a 16 year old high school student 2011-12-10T14:08:19 besh: good on him! 2011-12-10T14:08:21 I was probably also smarter as a 16yo hs student :) 2011-12-10T14:08:24 wow 2011-12-10T14:08:43 I probably had more time than now 2011-12-10T14:08:56 when i read that, i closed my laptop and went to bed :) 2011-12-10T14:09:55 I stand by my offer: If I come in nth place, I will buy a copy of MacKay for whoever is in (n+1)th 2011-12-10T14:10:38 you don't have to when you end rock bottom haha :) 2011-12-10T14:10:57 besh: That's what I was hoping for 2011-12-10T14:11:06 *** TheLinker has joined #aichallenge 2011-12-10T14:11:32 who's bulletv13? 2011-12-10T14:12:08 *** mleise has quit IRC (Ping timeout: 240 seconds) 2011-12-10T14:12:14 how do you implement weak attraction a1k0n? 2011-12-10T14:12:22 not you, but how could it be done 2011-12-10T14:12:41 slip something in the other ants' drinks 2011-12-10T14:12:54 :( 2011-12-10T14:13:13 1 point for adjacent ants? 2011-12-10T14:14:18 my decisions are not really based on points 2011-12-10T14:14:46 I may have to do something like that 2011-12-10T14:14:48 what is flagcapper doing for combat? 2011-12-10T14:15:02 well, it's a computer, so you have to do something with numbers 2011-12-10T14:15:09 maybe if he's 16yo he's actually driving them all with a joystick 2011-12-10T14:15:12 I suppose 2011-12-10T14:15:16 using the video game reflexes of youth 2011-12-10T14:15:19 lol 2011-12-10T14:15:28 McLeopold: If he's using Lisp, he's doing symbolic manipulation. :-P 2011-12-10T14:15:41 no one wants to talk serious today :) 2011-12-10T14:15:53 okay, produce a list of adjacent ant pairs, the longer list wins 2011-12-10T14:15:57 besh: And that's why I'm voting for Rick Perry. 2011-12-10T14:16:03 http://paste.aichallenge.org/MPYWt/?turn=1015 2011-12-10T14:16:43 *** epicmonkey has joined #aichallenge 2011-12-10T14:17:04 haha Murahshak . Better draw something than combat. 2011-12-10T14:17:53 my combat method can only take out vowells... 2011-12-10T14:18:45 now to make my bot detect words 2011-12-10T14:18:50 and insert/remove letters to change the meaning 2011-12-10T14:19:13 happy new year you say? how about happy now bear? 2011-12-10T14:20:55 McLeopold: how do you make your ants go to a distant hive all together? 2011-12-10T14:21:04 bfs 2011-12-10T14:21:08 braille would only take an average of 4 ants per letter 2011-12-10T14:21:18 at such distance? 2011-12-10T14:21:22 sure 2011-12-10T14:21:35 I search from the hive, pulling in all ants, so one search 2011-12-10T14:21:44 that's great 2011-12-10T14:22:00 actually, I search from all hives, pulling in the closest 2011-12-10T14:22:37 doesn't that time out big maps? 2011-12-10T14:22:38 but nobody speaks braille 2011-12-10T14:22:46 no, it's really fast 2011-12-10T14:22:52 it's just one paint of the map 2011-12-10T14:23:04 I'll have to try that 2011-12-10T14:23:13 do you have any pathfinding? 2011-12-10T14:23:25 I have A* to each objective 2011-12-10T14:23:35 from each ant 2011-12-10T14:23:48 I think I'm going to reverse and do bfs from every objective 2011-12-10T14:23:49 that is good for point to point, but I'm always sending multiples, so bfs gets them all 2011-12-10T14:24:00 I search from food to ants 2011-12-10T14:24:10 that's what I'm going to try, yes 2011-12-10T14:24:11 so, multi-source to multi-target 2011-12-10T14:24:17 it works great for what I've seen 2011-12-10T14:24:34 for exploring, I search each ant, which is the slow stupid part 2011-12-10T14:24:48 I have the same problem for exploring 2011-12-10T14:24:49 I need some vision edge detection 2011-12-10T14:24:56 that too 2011-12-10T14:25:16 for not playing on the contest your bot is quite nice 2011-12-10T14:25:46 *** foRei has joined #aichallenge 2011-12-10T14:27:10 McLeopold: why do you search each ant? For exploration I just do a BFS to all the unseen squares and all the unknown squares 2011-12-10T14:27:57 That ends up with bad behavior 2011-12-10T14:27:57 bmh: what are unknown squares? 2011-12-10T14:28:09 Anilm3: ones that you haven't seen before 2011-12-10T14:28:24 isn't that the same as unseen? 2011-12-10T14:28:40 Anilm3: Unseen - can't see now, Unknown - never seen 2011-12-10T14:29:09 I actually find the edge of the never-seen space 2011-12-10T14:29:12 and bfs from that 2011-12-10T14:29:24 oh, ok! 2011-12-10T14:29:28 I don't know if the effort of edge-detecting it saves time or wastes time 2011-12-10T14:29:52 BenJackson: and when everything is seen? 2011-12-10T14:30:07 possible enemy border 2011-12-10T14:30:09 Well, the perimeter is pretty important in many ways 2011-12-10T14:30:41 Anilm3: I'm constantly trying to keep as much possible in sight 2011-12-10T14:30:41 *** Antimony has quit IRC (Ping timeout: 240 seconds) 2011-12-10T14:30:54 also over time my concept of unseen has varied from truly never seen to just "not seen recently" 2011-12-10T14:31:00 BenJackson: you don't track where enemies might be? 2011-12-10T14:31:13 Checking one horizontal neighbour and one vertical neighbour should be enough for edge detection 2011-12-10T14:31:20 where enemies might be? 2011-12-10T14:31:43 BenJackson: my exploration method tries to minimize not seen recently squares, but on mazes doesn't work well 2011-12-10T14:32:14 BenJackson: start with everything as a possible enemy, anything that you see an isn't an enemy isn't a possible enemy, where enemies you see can move is a possible enemy and an enemy hill out of view spawns a possible enemy, then propogate the info each turn 2011-12-10T14:32:25 then you can sort of set up a "perimeter" around your hills 2011-12-10T14:32:30 where you don't actually need to keep visibility 2011-12-10T14:32:44 neat idea 2011-12-10T14:32:50 (particle filter) 2011-12-10T14:33:12 antimatroid: what's your bots name? 2011-12-10T14:33:16 on official I mean 2011-12-10T14:33:18 antimatroid 2011-12-10T14:33:24 it's not there or on tcp though :P 2011-12-10T14:33:52 it's not going to be very competitive, i haven't had enough time to play with it, but i'll upload it before the end 2011-12-10T14:34:36 haha if it's not as competitive as mcleopold I don't want to face you 2011-12-10T14:34:52 ? 2011-12-10T14:34:58 his bot is quite good 2011-12-10T14:35:16 better than mine probably 2011-12-10T14:35:16 but if i wasn't as competitive i'd be worse :P 2011-12-10T14:35:25 Oh no, my bot has fallen from peak rank 130 to 190 2011-12-10T14:35:58 antimatroid: I'm a poor english speaker :P 2011-12-10T14:36:10 anyway, bed 2011-12-10T14:36:40 gn 2011-12-10T14:36:53 bmh: have you seen the new official maps? 2011-12-10T14:37:13 Anilm3: yes... if you want to point me at a gzip archive of them, I'll swap 'em in 2011-12-10T14:37:28 you just have to download the updated tools package 2011-12-10T14:37:32 http://www.youtube.com/watch?v=y71lli8MS8s 2011-12-10T14:37:36 someone should do this for ants :) 2011-12-10T14:37:40 the maps are supposed to be there 2011-12-10T14:38:09 BenJackson: that shit is insane :P 2011-12-10T14:38:14 you reddit whore :P 2011-12-10T14:38:41 antimatroid: If you think that's bad.. I met my girlfriend at a Reddit meetup. 2011-12-10T14:38:50 haha 2011-12-10T14:38:56 *** xathis has quit IRC () 2011-12-10T14:43:53 i should really go to one of those meetups then, lol 2011-12-10T14:46:17 bmh: where to I start my bfs? from the hills? 2011-12-10T14:47:56 hahaha 2011-12-10T14:49:48 McLeopold: I have several that each start from many places 2011-12-10T14:49:52 and I use that info all over 2011-12-10T14:50:08 eg a bfs starting from all of your ants gives you the distance to the closest ant from every square 2011-12-10T14:50:13 and you can do that for any set of interesting things 2011-12-10T14:50:16 Zannick: I'm currently playing crawl :) 2011-12-10T14:50:26 then you can ask questions like "is this food closer to me or an enemy" in constant time 2011-12-10T14:51:44 *** Antimony has joined #aichallenge 2011-12-10T14:52:36 interesting mr BenJackson 2011-12-10T14:52:48 but that enemy may not be going to that food 2011-12-10T14:53:29 that question may not be *interesting* 2011-12-10T14:53:32 it's just an example 2011-12-10T14:53:43 ok 2011-12-10T14:53:58 *** bretep has quit IRC (Ping timeout: 258 seconds) 2011-12-10T14:54:48 *** Accoun has quit IRC () 2011-12-10T15:00:54 *** dvladim has joined #aichallenge 2011-12-10T15:04:47 BenJackson: skynet relegates: http://bhickey.net:2080/replay.5407 2011-12-10T15:06:48 time for the big leagues 2011-12-10T15:08:14 BenJackson: still need combat... 2011-12-10T15:08:32 (basically I don't want to write the code to find all ants within combat radius + 2) 2011-12-10T15:08:50 that's exactly the same code to find vision 2011-12-10T15:08:54 with a different kernel 2011-12-10T15:11:18 bj_v4 is kind of snakey 2011-12-10T15:11:36 all ants are going toward the closest goal 2011-12-10T15:11:43 so once ants are clumped they stay clumped 2011-12-10T15:11:56 interesting 2011-12-10T15:12:01 not very useful though 2011-12-10T15:12:04 Not interesting, bad 2011-12-10T15:12:17 2 ants can destroy your entire group 2011-12-10T15:12:25 yep 2011-12-10T15:12:29 *** jstrong has quit IRC (Quit: Leaving.) 2011-12-10T15:12:37 V7 was still pretty snakey and did fairly well though 2011-12-10T15:12:42 at the time peaking in the upper teens 2011-12-10T15:12:50 it had minimal combat, though 2011-12-10T15:12:56 Powercubes all around 2011-12-10T15:12:58 so the snakes actually worked out as fighting groups 2011-12-10T15:13:22 when I de-snaked and fixed combat it got a lot weaker at first 2011-12-10T15:13:32 *** Accoun has joined #aichallenge 2011-12-10T15:13:40 BenJackson: I desnake my ants by having my ants block BFS 2011-12-10T15:14:04 at that time I didn't have any costing implemented in my bfs 2011-12-10T15:14:22 bmh, how does that turn out in long corridors 2011-12-10T15:14:32 pairofdice: damned if I know 2011-12-10T15:16:04 *** bretep has joined #aichallenge 2011-12-10T15:16:36 *** epicmonkey has quit IRC (Ping timeout: 240 seconds) 2011-12-10T15:17:33 *** dvladim has quit IRC (Ping timeout: 244 seconds) 2011-12-10T15:17:59 bmh, did you try not blocking but increasing the distance for each ant 2011-12-10T15:18:24 ikaros: I thought of it, but it really doesn't fit with my BFS harness 2011-12-10T15:19:53 I think I figured a way to plug that in my bfs but haven't tested it 2011-12-10T15:20:04 *** bravewarrior has joined #aichallenge 2011-12-10T15:20:53 I had all kinds of problems when I tried last time. The solution is pretty simple though 2011-12-10T15:23:48 haha all my work email from today is about failing over to the emergency backup email 2011-12-10T15:24:11 BenJackson: checking work e-mail on a weekend? that's a paddlin' 2011-12-10T15:24:31 checking aicontest ranking during work? etc 2011-12-10T15:27:03 oh right, ikaros: there's my BFS code https://gist.github.com/1456201 2011-12-10T15:27:08 *** kire has quit IRC (Remote host closed the connection) 2011-12-10T15:30:23 dumbPoint, smartPoint 2011-12-10T15:30:39 is that haskell? 2011-12-10T15:30:39 what language is that 2011-12-10T15:30:52 Haskell. 2011-12-10T15:30:54 my bot no longer wants to kill hills 2011-12-10T15:30:58 it just kills all other ants 2011-12-10T15:31:10 this is not what i understand by just reading the code... 2011-12-10T15:31:10 sounds good too rwest :) 2011-12-10T15:31:15 rwest_, 2011-12-10T15:31:17 this one 2011-12-10T15:31:50 *** Bazzinga has joined #aichallenge 2011-12-10T15:32:03 Hello 2011-12-10T15:32:20 Bazzinga! 2011-12-10T15:32:55 oh shit I know why lol 2011-12-10T15:33:01 I fail so hard for doing that 2011-12-10T15:33:09 unfail then 2011-12-10T15:33:28 bmh, must admit i cant really read that :) the only thing ive done with haskell is factorial and such lol 2011-12-10T15:33:30 just did, I made the hill overwrite any value assigned to it after diffusion 2011-12-10T15:33:41 but that value I overwrote it with 2011-12-10T15:33:49 was lower than what was diffused around it 2011-12-10T15:33:52 *** Jacob_Strauss has quit IRC (Quit: Page closed) 2011-12-10T15:33:53 so they just circled it 2011-12-10T15:35:37 ikaros: it's really simple. I promise. It takes a vector of type a and returns a vector of type a 2011-12-10T15:36:14 then it takes two functions on points, one to determine if a point should get searched, and another to determine if the neighbors of that point should be searched 2011-12-10T15:36:58 The obscure part is that I'm constructing a queue inside the function :) 2011-12-10T15:37:21 its the syntax of haskell that makes me uncomfortable :) 2011-12-10T15:37:33 i know people love it.. but not me lol 2011-12-10T15:39:12 wanna see aweful syntax, you should try ZPL 2011-12-10T15:39:22 had to learn it for a parallel processing class 2011-12-10T15:40:27 awful even lol 2011-12-10T15:44:54 *** Bazzinga has quit IRC (Quit: Page closed) 2011-12-10T15:46:19 hehe 2011-12-10T15:46:32 i wouldnt call haskell syntax awful 2011-12-10T15:46:35 just dont like it 2011-12-10T15:46:59 too pure for my taste :)) 2011-12-10T15:47:52 I think I should get points the longer I survive with 1 ant 2011-12-10T15:48:28 I swear to god it's been at least 2-3 minutes I been waiting for this tcp game to end 2011-12-10T15:48:59 finally 2011-12-10T15:49:48 :) 2011-12-10T15:49:50 386 turns 2011-12-10T15:50:29 http://ants.fluxid.pl/replay.24133?turn=494 2011-12-10T15:51:24 haha I'm sure me and solifugid23tr timed out at the same turn 2011-12-10T15:53:24 aichallenge: janzert epsilon * r05a3e84 / (sql/2_generate_matchup.sql sql/opponent.sql): Remove hard game limit from pairing - http://git.io/hR1EYg 2011-12-10T15:53:55 http://aichallenge.org/visualizer.php?game=181039&user=3473 2011-12-10T15:53:58 ut oh, I timed out 2011-12-10T15:54:46 @rankings 2011-12-10T15:54:51 Anilm3: Top 10 players: pguillory(92.5), xathis(91.9), ChrisH(90.0), lazarant(88.7), GreenTea(88.6), delineate(88.3), fourmidable(88.3), Memetix(87.7), RVeerdonk(87.6), MomoBot(86.0) 2011-12-10T15:54:53 I been timing out a lot lately 2011-12-10T15:54:57 even my old code 2011-12-10T15:55:05 my old code took 60ms to run 2011-12-10T15:55:14 now it times out often 2011-12-10T15:55:17 makes no sense to me 2011-12-10T15:55:39 although I did a git pull yesterday ^^ 2011-12-10T15:55:41 I don't think it's likely that one I just pasted is a real timeout 2011-12-10T15:55:54 my most expensive operation is combat and I was in bigger fights earlier in that game 2011-12-10T15:56:15 same for orange 2011-12-10T15:56:18 *** Antimony has quit IRC (Ping timeout: 244 seconds) 2011-12-10T15:58:57 *** Antimony has joined #aichallenge 2011-12-10T16:00:23 *** UncleVasya has quit IRC () 2011-12-10T16:02:12 *** bmh has quit IRC (Ping timeout: 258 seconds) 2011-12-10T16:08:17 *** Kurnevsky has quit IRC (Quit: Instantbird 1.0) 2011-12-10T16:08:23 first test of new combat: success 2011-12-10T16:08:38 at least it does not wait, now test with real bot 2011-12-10T16:09:52 *** lilEzek has joined #aichallenge 2011-12-10T16:09:57 hi 2011-12-10T16:12:13 it failed hard 2011-12-10T16:13:33 haha 2011-12-10T16:14:03 you know I keep wondering if Go bots are actually using threads 2011-12-10T16:15:00 well, i have 5 informations about each enemy movement per ant movement per ant, so i have to properly use and weight it now... 2011-12-10T16:21:32 BenJackson: threads aren't allowed 2011-12-10T16:21:52 right 2011-12-10T16:22:39 *** Palmik has quit IRC (Read error: Connection reset by peer) 2011-12-10T16:22:49 you probbably meant microthreads 2011-12-10T16:23:00 I'm sure they use "goroutines" 2011-12-10T16:23:06 which may or may not use threads based on settings and etc 2011-12-10T16:23:26 coroutines, tasklets, microthreads, greenhtreads, it's all about the same i think ;) 2011-12-10T16:26:46 what's a microthread? 2011-12-10T16:29:16 oh don't worry 2011-12-10T16:29:22 *** dom7b5 has quit IRC (Quit: Ex-Chat) 2011-12-10T16:31:13 http://ants.fluxid.pl/replay.24224 <-- xathis combat >>>>>>>>>>>>>>>>>> mine 2011-12-10T16:33:16 it's kind of scary 2011-12-10T16:34:51 *** arscan has joined #aichallenge 2011-12-10T16:35:24 did they add some workers to the pool? games seem to be going by much faster 2011-12-10T16:37:33 yup 2011-12-10T16:37:47 how many do they have running, out of curiosity? 2011-12-10T16:37:51 you will see at the side bar that the contest has another sponsor 2011-12-10T16:38:07 you can see it youself at the bottom link "Server statistics" 2011-12-10T16:38:26 ah, thanks anim3 2011-12-10T16:38:33 *anilm3 2011-12-10T16:38:34 np 2011-12-10T16:43:37 http://forums.aichallenge.org/viewtopic.php?f=24&t=2083 2011-12-10T16:43:51 anyone here who understands it well? 2011-12-10T16:44:45 i was trying to implement it and see how well it performs with my bot 2011-12-10T16:50:22 memetix surely does 2011-12-10T16:51:36 *** UncleVasya has joined #aichallenge 2011-12-10T16:51:47 *** Antimony has quit IRC (Ping timeout: 252 seconds) 2011-12-10T16:52:00 *** pairofdice has quit IRC (Quit: leaving) 2011-12-10T16:52:01 *** Redgis has joined #aichallenge 2011-12-10T16:52:54 indeed 2011-12-10T16:53:43 flowenol: I use a very similar approach to combat, it's not perfect but I am still tweaking 2011-12-10T16:54:10 my problem is understanding the implementation 2011-12-10T16:54:18 english not being my first language 2011-12-10T16:54:29 makes it a little hard to me 2011-12-10T16:54:34 I actually use hashmaps for it 2011-12-10T16:54:43 from what i see 2011-12-10T16:54:45 since keeping an array the size of the map seems like a waste 2011-12-10T16:54:51 you only need to know the combat tiles 2011-12-10T16:54:57 i noticed that 2011-12-10T16:55:01 theres enough memory 2011-12-10T16:55:01 McLeopold: Hi. Do you remember an idea with kill ratio? To show in the visualizer two numbers: 1. how many of my ants were killed during the game and 2. how many enemy ants were killed by my ants. Is it going to happen? 2011-12-10T16:55:04 the influence part seems simple 2011-12-10T16:55:33 I counts the influence of enemy vs my own ants 2011-12-10T16:55:39 just counts of what tiles can be affected 2011-12-10T16:55:41 mine or enemy 2011-12-10T16:55:49 yes 2011-12-10T16:55:50 but that fighting array 2011-12-10T16:55:57 in his last post hi says 2011-12-10T16:56:08 I didn't read the last post, sec 2011-12-10T16:56:18 rwest kk 2011-12-10T16:57:06 i thought i understood the fighting thing but still there seems something wrong in my implementation of it 2011-12-10T16:57:16 so i probably didnt 2011-12-10T16:57:27 I don't really get what he is saying there 2011-12-10T16:57:32 me neither 2011-12-10T16:57:33 *** Elvorn has joined #aichallenge 2011-12-10T16:57:33 however I am not using that method exactly 2011-12-10T16:57:39 it's just very similar 2011-12-10T16:57:45 since my whole bot uses an influence map 2011-12-10T16:57:56 rwest_ diffusion like? 2011-12-10T16:58:00 yes 2011-12-10T16:58:07 thats what im using for movement 2011-12-10T16:58:14 but cant make it work with combat 2011-12-10T16:59:46 http://ants.fluxid.pl/replay.24216 2011-12-10T16:59:52 vs xathis for good comparison 2011-12-10T16:59:56 (he wins of course 2011-12-10T17:00:17 rossxwest2 is my bot 2011-12-10T17:01:32 holds pretty well 2011-12-10T17:02:08 tip: don't diffuse combat scores 2011-12-10T17:02:13 overlay them after diffusion 2011-12-10T17:02:47 i think i get it 2011-12-10T17:03:09 with influence 2011-12-10T17:03:09 because the scores shouldn't affect anything else 2011-12-10T17:03:21 and I actually make my ants attracted to enemy ants 2011-12-10T17:03:34 I already tried that too 2011-12-10T17:03:40 but i was difusing 2011-12-10T17:03:57 so if my ants had numerical advantage i was sending them to enemys 2011-12-10T17:04:12 I make them attracted, but then they get close and the tiles close to the ants have combat scores 2011-12-10T17:04:14 a stupid way and worked pretty bad 2011-12-10T17:04:22 so they won't suicide on them 1 by 1 2011-12-10T17:04:37 i see 2011-12-10T17:04:46 so thats why the overlay 2011-12-10T17:04:51 yea 2011-12-10T17:05:04 interesting 2011-12-10T17:06:57 i think i understand what memetix says 2011-12-10T17:07:18 back to some schematics 2011-12-10T17:10:35 *** kilae has quit IRC (Quit: ChatZilla 0.9.87 [Firefox 8.0/20111104165243]) 2011-12-10T17:14:26 McLeopold: amstan told me about the post-contest contest you were going to do to develop/debug stuff 2011-12-10T17:25:43 *** g0llum has quit IRC (Read error: Connection reset by peer) 2011-12-10T17:28:51 *** UncleVasya has quit IRC (Read error: Connection reset by peer) 2011-12-10T17:32:11 *** nickjohnson has joined #aichallenge 2011-12-10T17:34:00 *** amstan_ is now known as amstan 2011-12-10T17:35:12 *** bmh has joined #aichallenge 2011-12-10T17:42:34 *** Antimony has joined #aichallenge 2011-12-10T17:45:07 http://ants.fluxid.pl/replay.24312 timeout ftl 2011-12-10T17:46:07 rwest_: faster than light? 2011-12-10T17:46:29 haha 2011-12-10T17:46:32 for the lose! 2011-12-10T17:47:07 I guess recalculating weights for 329 ants is slow 2011-12-10T17:47:37 easily optimized to only be ants near combat 2011-12-10T17:48:39 *** sigh has joined #aichallenge 2011-12-10T17:55:21 rwest_: should be checking time and dropping out if you get too close to the time while iterating over your ants. 2011-12-10T17:58:44 *** Paradoxial has joined #aichallenge 2011-12-10T17:58:45 Yea I guess I can remove the comment from it 2011-12-10T18:03:28 *** pguillory has joined #aichallenge 2011-12-10T18:04:47 wow xathis has been busy, he's totally outclassing me on fluxid 2011-12-10T18:06:32 hahaha 2011-12-10T18:06:37 we'll see 2011-12-10T18:06:47 it's very hard to enter the top 100 now. how many games from upload do you usually have to play before entering the top100 again? 2011-12-10T18:07:26 someone posted a list 2011-12-10T18:07:31 6 or 7 if you're good 2011-12-10T18:08:02 pguillory: you should look the game we were in where I "timed out" 2011-12-10T18:08:17 you eroded my unmoving ants back to what your ants considered a good defense 2011-12-10T18:08:24 I thought it was instructional in that way 2011-12-10T18:11:10 don't see it, have the link? 2011-12-10T18:11:26 http://aichallenge.org/visualizer.php?game=181039&user=3473 2011-12-10T18:12:11 pguillory: I like your fighting :) 2011-12-10T18:12:21 BenJackson: i'm not that good then. lol. only top 120 after 14 games. hope to end up at the top 60 2011-12-10T18:12:43 I'm hoping to reenter the top 10 2011-12-10T18:13:00 my combat may just be too ant-inefficient though 2011-12-10T18:13:25 BenJackson: do you have a bot better than v4 but still not your best that you can run on my server? :) 2011-12-10T18:13:32 hahah 2011-12-10T18:13:48 I would like that too :) 2011-12-10T18:13:55 that was interesting 2011-12-10T18:14:08 V4 to V5 on aichallenge was score 64.64 to score 78.19 2011-12-10T18:14:15 that's a pretty big jump 2011-12-10T18:14:15 my problem is not combat at the moment. my ants keep clumping together on defending 2011-12-10T18:14:19 <_flag> pguillory: I noticed the same thing, xathis' new bot has been giving me a beating as well 2011-12-10T18:14:28 can't find a way to de-clump them 2011-12-10T18:14:35 BenJackson: run em both? :) 2011-12-10T18:14:44 BenJackson: 'eh, whatever. Can you run v5? v4 is no longer a threat 2011-12-10T18:14:59 the finals are going to be interesting 2011-12-10T18:15:19 ok, v4 and v5 both playing 2011-12-10T18:15:26 BenJackson: did i convince you to implement possible enemies? :P\ 2011-12-10T18:15:34 * antimatroid awakes 2011-12-10T18:15:36 antimatroid: it's an interesting idea 2011-12-10T18:15:54 i'm actually quite surprised you aren't doing that 2011-12-10T18:16:11 i would have expected that to be standard near the top? 2011-12-10T18:16:34 mostly you need vision in your territory anyway just for food 2011-12-10T18:16:49 but I can see it being useful in early games and low food games when you do lose sight of things 2011-12-10T18:17:16 it makes you care less about some not visible squares 2011-12-10T18:17:25 so i think you could wait longer before reexploring for food 2011-12-10T18:17:41 well my working version is less dense than the one on aichallenge 2011-12-10T18:17:50 so ants for visibility isn't a big deal 2011-12-10T18:18:04 I was inspired by someone's endgame 2011-12-10T18:18:07 not sure if chrish or pguillory 2011-12-10T18:19:53 *** Kommander has quit IRC (Ping timeout: 240 seconds) 2011-12-10T18:20:01 BenJackson: oh yeah, that game was kind of embarrassing actually, because of the way it fed single ants into your wall instead of assaulting it en masse 2011-12-10T18:20:24 in reference to the game you linked 2011-12-10T18:20:27 yeah 2011-12-10T18:20:42 I've noticed that a lot of games would end differently if they'd go a bit longer 2011-12-10T18:20:54 someone does have the upper hand but it's taking time 2011-12-10T18:23:53 bmh: rwest: now my two versions are only playing each other :) 2011-12-10T18:24:03 good news is v5 beats v4 every time 2011-12-10T18:24:16 right now I'm stuck trying to make v11 beat v10 similarly 2011-12-10T18:24:58 BenJackson: making a couple quick changes 2011-12-10T18:25:25 running again 2011-12-10T18:25:58 had to fix my timeout and optimize recalculating battle map after each ant move 2011-12-10T18:31:41 *** bravewarrior has quit IRC (Quit: Page closed) 2011-12-10T18:35:16 *** Antimony has quit IRC (Ping timeout: 240 seconds) 2011-12-10T18:36:26 java as resurged 2011-12-10T18:36:27 what is this 2011-12-10T18:40:09 wow, competition is kicking in as the tournament is near. some of the bots on the top 50 a month ago are in the top 150 now. 2011-12-10T18:40:28 I don't get why --html sometimes doesn't work for playgame 2011-12-10T18:42:01 *** shleb has joined #aichallenge 2011-12-10T18:42:52 *** Antimony has joined #aichallenge 2011-12-10T18:46:32 hul 2011-12-10T18:46:44 *** shleb has quit IRC (Client Quit) 2011-12-10T18:55:25 *** bluegaspode has joined #aichallenge 2011-12-10T18:58:38 AntDroid: the competition is always much fiercer right towards the end 2011-12-10T19:01:00 *** Vaenom has joined #aichallenge 2011-12-10T19:01:20 *** Antimony_ has joined #aichallenge 2011-12-10T19:01:35 *** Antimony has quit IRC (Ping timeout: 252 seconds) 2011-12-10T19:01:46 *** Antimony_ is now known as Antimony 2011-12-10T19:01:55 *** amstan_ has joined #aichallenge 2011-12-10T19:01:55 *** ChanServ sets mode: +o amstan_ 2011-12-10T19:01:56 *** bluegaspode has quit IRC (Quit: Page closed) 2011-12-10T19:03:00 *** amstan__ has joined #aichallenge 2011-12-10T19:03:00 *** ChanServ sets mode: +o amstan__ 2011-12-10T19:05:03 If anyone feels like trying to figure out what's wrong in this: http://pastebin.com/tSAApf4s 2011-12-10T19:05:39 my current tcpclient seems not to end my bot process properly. any ideas? the previous tcpclient i've downloaded on the forums was working fine before but I've downloaded the latest tools and i've replaced it with the new one from fluxid 2011-12-10T19:06:07 Given an origin/radius, both functions doesn't always consider the same tiles as being in the 'circle' and I can't figure out why 2011-12-10T19:06:17 Vaenom: modulus on negative numbers probably won't do what you want 2011-12-10T19:06:31 try point.Y + y + Smurfer.Height) % Smurfer.Height 2011-12-10T19:06:33 and for width 2011-12-10T19:06:35 *** Antimony has quit IRC (Ping timeout: 252 seconds) 2011-12-10T19:06:47 *** amstan_ has quit IRC (Ping timeout: 240 seconds) 2011-12-10T19:06:54 AntDroid: same here. you can just run it in a loop with a limited game count 2011-12-10T19:07:00 when it fully exits it will clean up 2011-12-10T19:07:18 I'll try that Ben, thanks 2011-12-10T19:08:12 BenJackson: i'm not good in python. care to post you script for tcp? 2011-12-10T19:08:36 I just do it in shell 2011-12-10T19:08:49 while :; do [your tcp command here] 5; done 2011-12-10T19:08:57 the '5' is the game limit for each loop 2011-12-10T19:09:07 *** amstan__ has quit IRC (Ping timeout: 240 seconds) 2011-12-10T19:11:26 *** Anilm3 has quit IRC (Quit: Lost terminal) 2011-12-10T19:14:05 *** foRei has quit IRC (Read error: Connection reset by peer) 2011-12-10T19:14:34 can't seem to make it work on windows batchfile. i would assume for shell you're using linux 2011-12-10T19:15:38 haha 20 games of old v new 2011-12-10T19:15:41 first round, new loses all 20 2011-12-10T19:15:56 second round I turn a heuristic back on 2011-12-10T19:16:00 new wins 1/10 2011-12-10T19:17:14 *** Anilm3 has joined #aichallenge 2011-12-10T19:20:22 *** TheLinker has quit IRC (Remote host closed the connection) 2011-12-10T19:21:55 *** bretep has quit IRC (Ping timeout: 258 seconds) 2011-12-10T19:21:59 oohhh crap 2011-12-10T19:22:04 now I see where this bug comes from 2011-12-10T19:24:07 http://sjeng.org/ftp/play/tcpclient.py 2011-12-10T19:24:25 this seems to solve my problem 2011-12-10T19:25:35 aichallenge: janzert epsilon * rb3eeace / (website/css/typo.css website/server_stats.php): Show bot failure rate per worker - http://git.io/dXGSLQ 2011-12-10T19:31:22 *** Redgis has quit IRC (Ping timeout: 268 seconds) 2011-12-10T19:31:34 @rankings 2011-12-10T19:31:35 Anilm3: Top 10 players: pguillory(92.8), xathis(92.0), ChrisH(90.1), lazarant(89.0), fourmidable(88.9), GreenTea(88.9), delineate(88.4), Memetix(87.9), RVeerdonk(87.3), a1k0n(86.2) 2011-12-10T19:32:03 aichallenge: janzert epsilon * ra06cedc / website/server_stats.php : Shorten label to fit - http://git.io/r_B9gA 2011-12-10T19:33:35 *** antimatroid1 has joined #aichallenge 2011-12-10T19:34:05 *** flowenol has quit IRC (Ping timeout: 252 seconds) 2011-12-10T19:34:07 *** antimatroid has quit IRC (Read error: No route to host) 2011-12-10T19:36:26 BenJackson: hmm, the modulus things seems to work pretty good aside from this issue. In this http://imgur.com/kivB7 , the AddPresence is used to select the tile that fits in the radius2, while the other function is used to determine ant proximity (to collate them in group). As you can see at the top-right, the 2 functions doesn't agree about one of the ants and the collation system fails to group the 4 ants into a battle... 2011-12-10T19:36:29 *** bmh has quit IRC (Ping timeout: 258 seconds) 2011-12-10T19:38:37 sorry, not following that 2011-12-10T19:38:47 any other idea? :\ don't mind me if you're busy with more interesting stuff. Also, congrats on your ranking Ben. I wish I'd do that well before the finals kick in, but that's not likely for now. :P 2011-12-10T19:39:59 I meant top-left. 2011-12-10T19:40:32 ah, didn't notice your += width/height below the mod already 2011-12-10T19:41:34 yea 2011-12-10T19:42:25 *** Anilm3 has quit IRC (Quit: leaving) 2011-12-10T19:42:46 *** lilEzek has quit IRC (Quit: Page closed) 2011-12-10T19:44:55 *** arscan has quit IRC (Ping timeout: 258 seconds) 2011-12-10T19:45:32 looks fine to me 2011-12-10T19:45:35 I have a similar function 2011-12-10T19:45:43 btw the combat radius isn't combatradius + 2 2011-12-10T19:45:58 it's combatradius + 2 euclidean steps 2011-12-10T19:47:19 yea? thanks for the review. I began to feel the problem is in the rendering of the visuals... can't be elsewhere at that point. stupid timewasting bug 2011-12-10T19:48:17 *** Antimony has joined #aichallenge 2011-12-10T19:48:49 not sure what you mean by that. Perhaps a google of "euclidean steps" will enlight my poor-math ass :S 2011-12-10T19:49:04 er manhattan steps, sorry 2011-12-10T19:49:20 combat and view radius are euclidean (sqrt(x^2+y^2) distances) 2011-12-10T19:49:26 movement is manhattan (steps n/s/e/w) 2011-12-10T19:49:38 when people say "combat radius + 2" 2011-12-10T19:49:44 the 2 is because they could move and you could move 2011-12-10T19:50:04 so a...b could become a.b 2011-12-10T19:50:12 attack radius = 5 2011-12-10T19:50:22 potential attack radius = 10 2011-12-10T19:50:35 potential conflict radius = 17 2011-12-10T19:51:50 aw, I see. You saw the large radius on the pic and guessed I used it for combat. 2011-12-10T19:52:11 but it's only for pre-combat aggregation 2011-12-10T19:52:49 right 2011-12-10T19:53:44 I get what you're saying and I'll apply that in my combat code when I'll soon get there 2011-12-10T19:53:48 Hey McLeopold could you review some sql changes 2011-12-10T19:53:53 sure 2011-12-10T19:54:01 already committed? 2011-12-10T19:54:06 the first one is 2011-12-10T19:54:22 https://github.com/aichallenge/aichallenge/commit/b3eeacece8c645c033c13483484367bdb0123a4a 2011-12-10T19:54:30 does that do what I think it does ;) 2011-12-10T19:54:42 and is there a better faster way to do that query 2011-12-10T19:55:03 specifically I'm trying to get the percentage of bots that timeout or crash on each worker 2011-12-10T19:55:10 can we move that to the sql.php file? 2011-12-10T19:55:25 :( I'd rather not, sql.php is such a pain 2011-12-10T19:55:31 I know I didn't get everything over there yet, but the more the merrier 2011-12-10T19:55:36 but if you really want to, I don't expect to touch it again 2011-12-10T19:55:56 I did it so that when we move to django, it is easier to do the orm 2011-12-10T19:56:08 it can wait 2011-12-10T19:56:22 ok reading... 2011-12-10T19:56:45 I really don't like the idea of sql.php since it breaks locality of code and ends up as a huge mixed up file of queries 2011-12-10T19:57:17 when you move to django you can do some pretty complex queries right on a QuerySet 2011-12-10T19:57:41 aichallenge: janzert epsilon * r82bd445 / sql/2_generate_matchup.sql : Shorten or eliminate race in choosing a seed - http://git.io/-3BoKg 2011-12-10T19:57:43 it is ugly, but at some point you just want the sql to go away and be able to write a getData() call 2011-12-10T19:57:47 and that is the second change 2011-12-10T19:58:42 janzert: I think you can get the player counts per game from the map table instead 2011-12-10T19:58:56 then you don't need a group by 2011-12-10T19:59:10 ahh, probably 2011-12-10T19:59:31 really I don't even need the players per game at all 2011-12-10T19:59:37 just the total for that worker 2011-12-10T20:00:56 I think the datetime check is probably the slow part, but there is no way around that 2011-12-10T20:01:12 we can add an index with status and timestamp to make it faster 2011-12-10T20:01:54 is there a way to get the min game_id that is within 30 minutes then use that everywhere instead of the explicit timestamp check? 2011-12-10T20:02:19 yeah 2011-12-10T20:02:35 if I had a game_id then the inner queries wouldn't even have to join against the game table 2011-12-10T20:02:55 I was doing some stuff in api_game_result.php to avoid some of that stuff per player, but I dropped its use 2011-12-10T20:03:42 we have an index for timestamp on game, just do a limit 1 query and save it to a var 2011-12-10T20:03:48 I think that should be fast 2011-12-10T20:04:05 mysql is horrible at recognizing inner queries that don't change 2011-12-10T20:04:27 I was so used to mssql before this where everything "just works" 2011-12-10T20:04:43 can the var be set in the outer query and be accessible to the inner ones? 2011-12-10T20:05:35 *** ikaros has quit IRC (Quit: Ex-Chat) 2011-12-10T20:05:36 set @max_game_id = (select game_id from game where timestamp > timestampadd(blah) order by game_id limit 1); 2011-12-10T20:05:48 then use @max_game_id in any subsequent query 2011-12-10T20:05:51 err... 2011-12-10T20:05:59 you need a single query from php 2011-12-10T20:06:12 yeah, select it as a fake table 2011-12-10T20:06:54 select * from table where game_id >= @max_game_id, (select @max_game_id = (query)) max_query 2011-12-10T20:07:34 the query is run once as a table a joined, then you use the value everywhere 2011-12-10T20:07:42 I used it in other places 2011-12-10T20:08:54 https://github.com/aichallenge/aichallenge/blob/epsilon/sql/2_generate_matchup.sql#L80 2011-12-10T20:09:13 the t table is an example of a query to select a value once, then use it other places 2011-12-10T20:09:43 * McLeopold starts writing a bot in mysql... 2011-12-10T20:10:03 except the value is only used in the outer query there 2011-12-10T20:10:30 anyway the matchup insert change is the more important one 2011-12-10T20:10:50 we're getting almost 20% duplicated seeds now 2011-12-10T20:11:54 due to race condition? 2011-12-10T20:12:15 pretty sure, yes 2011-12-10T20:12:56 janzert: I think the real fix is transactions 2011-12-10T20:13:06 I just don't know if that is going to do it 2011-12-10T20:13:44 why don't we instead write a server site script to keep a constant connection and keep the matchup count high 2011-12-10T20:14:19 I would feel more confidant doing that than changing the sql 2011-12-10T20:14:22 certainly the real fix is transactions but that entails a database change 2011-12-10T20:14:41 right, lets put effort into a manager instead of tweaking this 2011-12-10T20:14:51 this puts the seed select and instert in one statement 2011-12-10T20:14:57 insert 2011-12-10T20:15:23 yeah, it should work still 2011-12-10T20:15:49 I have to make sql changes anyway soon 2011-12-10T20:16:00 I need to go. I can look at stuff more carefully in about 5-6 hours 2011-12-10T20:16:02 to the pairing that is 2011-12-10T20:16:05 ok 2011-12-10T20:16:26 I'll probably take it live and then back it out if there is a problem 2011-12-10T20:24:27 *** jstrong has joined #aichallenge 2011-12-10T20:24:30 *** jstrong is now known as roflmao 2011-12-10T20:31:17 aichallenge: janzert epsilon * r465ae1c / website/server_stats.php : Speed up worker stats query - http://git.io/9rvSBA 2011-12-10T20:31:53 LAST WEEKEDN! 2011-12-10T20:32:05 shhh 2011-12-10T20:32:06 it's sunday 2011-12-10T20:45:23 last weekend and i haven't tried some of the ideas i've wanted to implement yet. oh well. 2011-12-10T20:51:04 last weekend and I was finally getting decent with my bot 2011-12-10T20:51:04 oh well 2011-12-10T20:51:08 heh 2011-12-10T20:53:41 tryin somethin new 2011-12-10T20:53:46 isntead of choosing best min 2011-12-10T20:53:49 lets try best max 2011-12-10T21:03:47 *** grc has joined #aichallenge 2011-12-10T21:15:48 *** TheLinker has joined #aichallenge 2011-12-10T21:21:10 *** roflmao has quit IRC (Ping timeout: 248 seconds) 2011-12-10T21:21:12 *** Conorach has quit IRC (Ping timeout: 244 seconds) 2011-12-10T21:25:32 *** jstrong has joined #aichallenge 2011-12-10T21:31:06 *** grc has quit IRC (Ping timeout: 258 seconds) 2011-12-10T21:31:27 so 2011-12-10T21:31:42 a friend of mine was hacked today (drive-by browser exploit) 2011-12-10T21:32:04 wasted all day figuring out how it worked and what the payload/keylogger did (yay for VMs!) 2011-12-10T21:32:18 anyway it's a java 7 exploit, and it worked on windows 7 with chrome as the browser 2011-12-10T21:33:24 kind of scary.... it looks like they bought the exploit or something and you just provide code to run in the html of the page 2011-12-10T21:34:13 anyway that's why you don't click random links you get over IM (that's what he did...) 2011-12-10T21:39:21 I should try plugging in a different payload to see if this works in a linux VM 2011-12-10T21:39:36 lol, i swear you have to try and get an im virus 2011-12-10T21:43:26 well 2011-12-10T21:43:29 it's a drive-by exploit 2011-12-10T21:43:54 and it worked in fully up to date java 7, windows 7 and chrome 2011-12-10T21:44:14 the .jar was already on VirusTotal and detected by most of the AV scanners 2011-12-10T21:45:37 but AV doesn't stop your browser from running something in the java plugin :P 2011-12-10T21:49:09 *** Antimony has quit IRC (Ping timeout: 252 seconds) 2011-12-10T22:06:58 *** Antimony has joined #aichallenge 2011-12-10T22:10:21 *** TheLinker has quit IRC (Read error: Connection reset by peer) 2011-12-10T22:11:20 potential conflict radius = 17 2011-12-10T22:11:32 How do you come up with the 17? 2011-12-10T22:12:20 *** TheLinker has joined #aichallenge 2011-12-10T22:12:39 *** ChrisH_ has joined #aichallenge 2011-12-10T22:13:44 janzert, amstan, McLeopold: I have some doubts about the performance of some of the servers running games, I'd like to discuss it if any of you are online. 2011-12-10T22:13:45 *** nha has quit IRC (Ping timeout: 252 seconds) 2011-12-10T22:16:52 ChrisH_: afk for a few minutes, but you can now see the percentage of bots on each worker that are timing out or crashing (in the past half hour) 2011-12-10T22:17:07 my concern is not about timeouts actually 2011-12-10T22:17:52 janzert: ping me when you are back 2011-12-10T22:20:29 Q: is there any maximum players per game specified? 2011-12-10T22:20:51 10 2011-12-10T22:21:07 *** thestinger has quit IRC (Quit: brb) 2011-12-10T22:21:59 and you beat all of them hey? :) cool bot. good luck for the finals when you get there 2011-12-10T22:23:18 *** jstrong has quit IRC (Quit: Leaving.) 2011-12-10T22:23:39 ChrisH_: back 2011-12-10T22:23:43 ok 2011-12-10T22:23:45 *** thestinger has joined #aichallenge 2011-12-10T22:23:58 take a look at this game: http://aichallenge.org/visualizer.php?game=186986&user=10 2011-12-10T22:24:39 there were no timeouts, but I see very suspicious behavior of my bot. 2011-12-10T22:24:44 go to turn 17 2011-12-10T22:25:00 and step to turn 18 2011-12-10T22:25:18 notice that my ant at 17,27 doesn't move 2011-12-10T22:25:36 that shouldn't happen given the board situation 2011-12-10T22:26:14 with only 6 ants and no combat yet my bot should finish a turn in around 30 ms 2011-12-10T22:26:28 and with that much unexplored space every ant should move 2011-12-10T22:27:50 and if you play the game forward you will notice a lot of my ants pausing for a turn or two, and they are always the ants closest to hills, which are the last ones to get their moves computed, so the only rational explanation I have is that my bot isn't getting computing resources similar to what i have come to expect 2011-12-10T22:28:30 and yet it is still avoiding a timeout, which means it is getting enough time to check the clock and bail out with <40ms to go 2011-12-10T22:29:00 not sure what to make of it, but something isn't right 2011-12-10T22:29:25 could you give me some sort of test program based off your bot to benchmark the workers? 2011-12-10T22:29:37 I could try comparing some of the workers with that 2011-12-10T22:29:52 what do you have in mind? 2011-12-10T22:30:34 I could make my bot spit out some stats to stderr if that's what you want, it could print out how long it thinks each turn is taking 2011-12-10T22:31:06 basically your bot with some static input built in and it can output how long it took to compute all the mvoes 2011-12-10T22:31:43 sure, we could use my existing code with a couple command line options I have for testing 2011-12-10T22:31:58 and just pipe in the input for any game 2011-12-10T22:32:46 *** zyberkiddy has joined #aichallenge 2011-12-10T22:32:46 look at turn 444 on that game 2011-12-10T22:33:18 err nm 2011-12-10T22:33:37 was just scanning for other cases where they were getting stuck 2011-12-10T22:34:37 do you really bail before issuing orders if time is almost up? that's intriguing :) 2011-12-10T22:35:21 so get the bot input for my bot from the game I just linked and run my bot with: MyBot -test -v=1 test.txt >o.txt and grep through test.txt for "Orders sent in" to get ms timings of each turn 2011-12-10T22:35:46 pguillory: no I don't bail, I just stop computations and send the orders decided on so far 2011-12-10T22:36:03 ah 2011-12-10T22:37:02 pguillory: another example is turn 59-60 2011-12-10T22:37:11 i mean 58-59 2011-12-10T22:37:47 *** TheLinker has quit IRC (Quit: Bye) 2011-12-10T22:37:51 the three ants near my hill in the upper left all pause 2011-12-10T22:38:18 and there are several more pauses over the next several turns 2011-12-10T22:41:36 noticed it near enemy hills as well, that a factor? 2011-12-10T22:41:42 but what's really telling about the turn 17-18 pause is my bot hasn't learned very much new on turn 18 that would cause the paused ant to move again on turn 19, no new hills, food, or enemy ants are discovered, just a little fog of war pushed back, so it doesn't make sense for the ant to stop moving and the suddenly restart 2011-12-10T22:42:23 pguillory: on what turn? 2011-12-10T22:42:33 550 2011-12-10T22:43:09 at location 47:49 2011-12-10T22:43:15 47:79 i mean 2011-12-10T22:43:38 the stretch of turn 125 - 140 are really noticable 2011-12-10T22:44:27 would seem odd to do any kind of ordering by distance from a razed enemy hill 2011-12-10T22:46:55 no, those three ants should be headed straight back to fight the blue ants at the end of the hallway where it is currently a 4-on-2 in favor of blue 2011-12-10T22:47:45 my bot finds the closest ant-target pairs first, so the farther an ant is from a suitable target, the later in the processing it's move gets decided 2011-12-10T22:48:03 gotcha 2011-12-10T22:49:07 so I'm used to seeing newly spawned ants far from the action do this sort of thing on a big map with lots of combats and ants to move, but this game is not on a big board and the pauses are happening early in the game, so it is unexpected 2011-12-10T22:49:53 I will have to download the game input and verify that these ants are actually getting orders if CPU time is available, that would be telling 2011-12-10T22:51:20 pguillory: BTW, I'm really curious about your combat logic, I hope you share it once the competition is over. 2011-12-10T22:52:20 sure i'll do that, probably post everything on the last day 2011-12-10T22:52:25 hopefully others will as well 2011-12-10T22:52:47 it would be neat to read other people's techniques while watching them be put into action in the finals 2011-12-10T22:52:52 after submissions close and while the finals are running is probably a good time to discuss bot workings 2011-12-10T22:53:46 i expect interest in after-action reports to dwindle after the finals end 2011-12-10T22:59:18 janzert: I downloaded the game input for the one I linked and ran it through the version of my bot my current submission uses and indeed it emits orders to move the ants that I see not moving in the game. In particular on turn 17 it emits the order "17 27 W" for an ant that doesn't move in the replay and completes that turn in 33ms on my machine. 2011-12-10T23:01:11 janzert: if you are going to do the test you were suggesting, then I have a correction, grep for "orders sent in" on the stderr stream of my bot running with the -v=1 option to see my turn timings 2011-12-10T23:01:23 *** thestinger has quit IRC (Quit: WeeChat 0.3.6) 2011-12-10T23:03:49 *** Fandekasp has joined #aichallenge 2011-12-10T23:05:52 *** zyberkiddy has quit IRC (Ping timeout: 240 seconds) 2011-12-10T23:06:01 *** Antimony has quit IRC (Ping timeout: 255 seconds) 2011-12-10T23:06:25 Do you think it'd be worth it to have more than one ant contest? Do you guys think there could be enough bot evolution/tuning potential for a second and a third round? 2011-12-10T23:08:21 I think a lot of people could make their bots better given more time, but personally I've hit a bit of a wall. I have one idea I'm working on right now that might have some benefit on certain maps, but I haven't been able to make any big leaps in my ant's quality over the last two weeks. 2011-12-10T23:08:36 I'm not suggesting that org shouldn't develop other challenges following this one, but since everything to run ants contest is already developed and running, it could get a life of its own? 2011-12-10T23:09:43 So, I'm currently in the top 5, and I see other bots catching up to the top, but I don't know how much more improvement the top bots can make without significant effort. I think (but could be wrong) that the top bots are hitting the diminishing returns wall. 2011-12-10T23:09:59 usually tcp server usage peters out after about a month 2011-12-10T23:10:10 ChrisH_, I see 2011-12-10T23:12:03 So many of the contest #1 tops wouldn't be that much interested in a second contest. But since the challenge is about learning and since there's always new comers, it be great if the contest could become permanent with 4 seasons a year. Think of it as the "Basic" of AI :) 2011-12-10T23:12:47 janzert: yea, but they aren't as exciting as the center stage.. 2011-12-10T23:13:04 very rapidly diminishing returns at this point 2011-12-10T23:13:08 I was analyzing a one-on-one game I had against BenJackson recently in which my bot won, and in the end it looked decisive, but when you step through the game and look for what my bot did better, it boiled down to a couple of lucky food spawns that I was closer to and a few combat situations that my bot handled better. Those deltas resulted in about a 7-8 ant advantage for me in the mid game and my bot used that advantage to gain 2011-12-10T23:13:47 there's so much effort going on to make this contest possible.. I really think it would deserve permanence on the web. 2011-12-10T23:14:08 janzert: do you see any reason to not do it? 2011-12-10T23:14:25 it really is a fun event and a great experience for everyone involved 2011-12-10T23:14:30 I think that if it didn't have an end we would all burn out and it would just fizzle instead of building to an exiciting finals. 2011-12-10T23:14:32 *** bmh has joined #aichallenge 2011-12-10T23:14:45 just lack of admins to do it and lack of resources to run it, but otherwise not really no 2011-12-10T23:15:50 ChrisH, I agree, except a "Season" setup would fix that. Hockey, Baseball, Basketball, Football ... they all never ends, but they do.. 2011-12-10T23:17:12 *** zyberkiddy has joined #aichallenge 2011-12-10T23:17:35 janzert: any action on becoming a non-profit? 2011-12-10T23:17:39 janzert: right. Well, the admins could rotate as the new comers gets interested in running a season or 2 after participating. 2011-12-10T23:17:42 janzert: how do you want to proceed related to the "buggy" performance of the game I pointed out? 2011-12-10T23:18:01 *** thestinger has joined #aichallenge 2011-12-10T23:18:29 still trying to setup so I can test it on a worker or two 2011-12-10T23:18:40 although I don't really expect it to show much 2011-12-10T23:19:18 since if anything it's probably temporary changes in performance is my guess 2011-12-10T23:20:11 janzert: as for the resources, I feel this could be worked out but I obviously don't know all the metrics and challenges, so.. 2011-12-10T23:20:55 would it be possible to make an AWS image that people could start to contribute computing resources? 2011-12-10T23:20:57 janzert: I was wondering if there could be anything odd about the behavior of the system timer on any of the boxes. the fact that my bot didn't timeout suggests to me that it thought it had less time than it really did and sent orders too early, if it truely wasn't getting CPU time I would expect a timeout because I only allow 40ms of overhead time 2011-12-10T23:21:59 pguillory: only if they withdrew from the contest and we trusted them completely to not change results or leak contestant code 2011-12-10T23:22:00 40ms of overhead is a lot of time. 2011-12-10T23:22:43 ChrisH_: not that I know of but *shrug* 2011-12-10T23:22:56 The workers could detect performance problems and end/void the games that suffered from them. It probably would be easy to implement. but maybe I drank a Red Bull too much. think I'll go back to my rank 600 code .. :P 2011-12-10T23:23:22 In python, in a iterative recursive algorithm I can get 1.2 million calls to the function in 400ms, 2011-12-10T23:23:41 Wraithan: yes, but on we are analyzing a situation in which a game my bot played on the server resulted in ants that my bot wanted to move didn't get moved. 2011-12-10T23:23:49 janzert: i was thinking with SSH disabled, not sure how tamper proof it's possible to make such a thing though 2011-12-10T23:24:23 And when I run the same game on my machine the whole turn took 33ms, and it tried to move the ant that the server didn't move. 2011-12-10T23:25:10 pguillory: since we update worker code semi-regularly and unpredictably we also still need a way to reset the VM basically 2011-12-10T23:25:22 So, knowing my bot's code I know that the ant in question would be the last one my code would figure the moves for, but it is also checking the system time between each ant, so it could do that and think time was up and just send the orders for the first ants. 2011-12-10T23:25:38 bj_v5 frags me right and proper. 2011-12-10T23:26:15 But that should only happen if my bot has taken 460 ms so far, and there are only two explanations for that, either it didn't get the CPU time it should have for 430 ms, or the system timer lied. 2011-12-10T23:26:36 ChrisH_: "orders sent in" doesn't appear anywher in the stderr output 2011-12-10T23:26:57 janzert: did you give the command line options "-test -v=1" ? 2011-12-10T23:27:11 I copied and pasted the line above 2011-12-10T23:27:11 *** Antimony has joined #aichallenge 2011-12-10T23:27:57 janzert: show me 2011-12-10T23:28:20 MyBot -test -v=1 test.txt >o.txt 2011-12-10T23:28:50 can you paste the beginning of the test.txt file somewhere for me to see? 2011-12-10T23:29:05 well, actually needed ./ at the beginning as well 2011-12-10T23:29:11 yeah 2011-12-10T23:30:29 http://pastebin.com/EPh65G71 2011-12-10T23:31:41 *** avdg has quit IRC (Quit: Leaving.) 2011-12-10T23:32:13 ok, that looks right, but it's a capitalization issue I think, try "Orders sent in" I forgot that I changed that spelling recently from what it was with my V5. 2011-12-10T23:32:51 or you can just grep for "sent in" 2011-12-10T23:33:22 ok, no easy way to tell what turn is what. but the first 100 turns appear to be under 100ms 2011-12-10T23:33:59 yeah, I added the turn number on that line since the version online, but you could check the line number if you open it in an editor 2011-12-10T23:34:06 range seems to be 28-75ms 2011-12-10T23:34:50 that's sounds reasonable, the first turn is usually a bit longer and then it is in the 30-50ms range until combat occurs 2011-12-10T23:36:02 hmm, first turn is 58, second is 28, 74 was down near 100 2011-12-10T23:36:27 now, open the o.txt file and find the 17th "go" line and show me the orders that follow 2011-12-10T23:36:59 janzert: that sounds completely normal 2011-12-10T23:38:22 http://pastebin.com/jBNHgZhD 2011-12-10T23:39:49 same as the actual game right? 2011-12-10T23:40:03 yeah, that's not what I expected 2011-12-10T23:41:06 I'm pretty sure I've got my versioning right, but maybe not, let me try one more time on my machine to make sure I didn't use the wrong version here. 2011-12-10T23:42:07 btw, turn 17 in the test was 47ms 2011-12-10T23:44:01 aichallenge: janzert epsilon * r7317514 / worker/worker.py : Fix worker compile option and remove hash option handling that no longer exists - http://git.io/tqkEjg 2011-12-10T23:47:32 janzert: can you try this for me: "6g -V" to check the go compiler version? 2011-12-10T23:49:03 sure, it's still the same as copying off the starter_packages page though 2011-12-10T23:49:04 6g version release.r60.1 9753+ 2011-12-10T23:49:15 Ok 2011-12-10T23:50:09 I tried bot my V5 code and my V4 code, and both issue six orders on turn 17, with the extra one being o 17 27 W, I can't explain this. 2011-12-10T23:51:16 the uploaded zip is ~30kb if you want it emailed to you 2011-12-10T23:51:30 sure 2011-12-10T23:53:48 Extrarius: you around? 2011-12-10T23:54:13 *** Antimony has quit IRC (Ping timeout: 244 seconds) 2011-12-10T23:56:14 janzert: I think we did eliminate the CPU or system timer as culprits at least, but I'm at a loss now. 2011-12-10T23:56:43 Especially because I don't see it in every game, or even every turn of the one game we were looking at. 2011-12-10T23:58:14 ok, should be sent to your account email 2011-12-10T23:58:47 yeah, intermittent things that only happen on the deployed version are a real pain :/ 2011-12-10T23:59:32 especially when the part of the code in question is some of the oldest and most tested in my bot