2010-10-31T00:00:57 *** dmj111 has quit IRC (Ping timeout: 245 seconds) 2010-10-31T00:02:03 I have 59 versions of my bot saved that I can use for test games, and often older ones beat newer ones 2010-10-31T00:02:37 jesus 2010-10-31T00:02:44 i guess i should start branching 2010-10-31T00:04:16 i am just now realizing how much i bit off with this implementation lol 2010-10-31T00:07:52 Yo yo! 2010-10-31T00:08:05 hey jeff 2010-10-31T00:08:10 long time no see 2010-10-31T00:08:10 I just copy and rename my main program so I can keep all the versions runnable 2010-10-31T00:09:50 Yeah I know. I've been busy. 2010-10-31T00:10:04 Happy Halloween! 2010-10-31T00:10:15 i just had an invalid order problem and realize it could have happened in like 6 places :( gonna be a long night 2010-10-31T00:10:32 Anybody have a bot that draws a pumpkin or bat? 2010-10-31T00:12:51 Top 10 players: GreenTea(4165), bocsimacko(4158), felixcoto(4151), Slin-.-(4132), Accoun(4113), dmj111(4106), rsergio(4068), Hazard(4033), joakim_mjardner(4022), rs5147(4013) 2010-10-31T00:16:04 *** davidjliu has quit IRC (Quit: Page closed) 2010-10-31T00:16:36 So who is GreenTea? Are they here? 2010-10-31T00:17:27 i think i will submit my mostly terrible bot now and be done for the contest 2010-10-31T00:19:29 a month early? 2010-10-31T00:20:09 i have other work i need to do for about all of november 2010-10-31T00:20:40 *** espes has quit IRC (Ping timeout: 265 seconds) 2010-10-31T00:28:49 noooo Zannick! 2010-10-31T00:29:35 I know that you will be back haha. I have carefully designed the contest to be as addictive as possible. "The first time is free" :-D 2010-10-31T00:29:36 *** espes has joined #aichallenge 2010-10-31T00:29:43 i'll still be here to help out 2010-10-31T00:30:05 and i'd love to help prepare/run the next one 2010-10-31T00:30:43 but november is nanowrimo so my free time won't be all spent programming 2010-10-31T00:35:42 *** janzert has joined #aichallenge 2010-10-31T00:35:43 j3camero: zomg! you're alive 2010-10-31T00:35:53 what, where 2010-10-31T00:36:03 I can't believe it :P 2010-10-31T00:36:16 j3camero: it's a little too addictive if you ask me 2010-10-31T00:36:24 i'm constantly up till sunrise 2010-10-31T00:36:44 i just did this: http://cubicmatter.net/forums/viewtopic.php?f=10&t=54&p=442#p442 2010-10-31T00:39:54 btw, golang is working on the servers. just waiting for a working starter package from lavalamp before announcing it 2010-10-31T00:40:23 double rainbow.. 2010-10-31T00:40:28 Although I may just announce it with starter package to follow 2010-10-31T00:40:29 antimatroid1: oh wow sunrise. Yikes what have I done? 2010-10-31T00:42:55 janzert: you are my hero! 2010-10-31T00:43:06 janzert: we had lots of good entries in Go during the Tron contest 2010-10-31T00:43:15 great 2010-10-31T00:43:27 janzert: and its very appropriate to have Go since it's Google-sponsored! 2010-10-31T00:43:29 lavalamp_: did most of the footwork actually 2010-10-31T00:43:45 amstan: what is that link you posted? Looks likke 3D modeling? 2010-10-31T00:44:11 janzert: oh nice. That's great that you're roping in more volunteers! 2010-10-31T00:44:14 j3camero: it's a game server, and you can edit multiplayer, i just added a way to procedurally generate maps 2010-10-31T00:44:25 amstan: sick. 2010-10-31T00:44:41 janzert: what's the status of C# right now? 2010-10-31T00:44:55 working through a work around :P 2010-10-31T00:45:12 still leaking semaphores like crazy but there are now scripts on the servers to clean up after it 2010-10-31T00:45:33 that took me 5 days or something to sort out :( 2010-10-31T00:46:07 I just added a script yesterday to monitor ec2 instances and reboot them 2010-10-31T00:46:13 *** superflit has joined #aichallenge 2010-10-31T00:46:22 you'll want to run it locally somewhere you trust putting your ec2 credentials 2010-10-31T00:46:49 janzert: oh that's awesome! we should get that set up asap. I am free basically all day tomorrow. You down? 2010-10-31T00:47:03 It's at /trunk/planet_wars/backend/ec2-check.py but with no isntructions currently 2010-10-31T00:47:19 yeah, I won't be around till the afternoon for a few hours 2010-10-31T00:47:30 then after around 8pm anytime 2010-10-31T00:47:35 Sweeeet I see we have around 37 games/minute average right now! That's probably doing a lot to reduce user frustration, eh amstan? 2010-10-31T00:47:36 too bad all your dev is in python i have too much free time ;) 2010-10-31T00:47:54 janzert: are you eastern time? 2010-10-31T00:47:58 yep 2010-10-31T00:48:02 Michigan 2010-10-31T00:48:07 Nice game: http://72.44.46.68/canvas?game_id=787507 2010-10-31T00:48:27 janzert: it's variable though 2010-10-31T00:48:40 j3camero: it's variable though 2010-10-31T00:49:14 I was without power for about 48 hours so I my instances went down during that time :/ 2010-10-31T00:49:34 eburnette: v nice 2010-10-31T00:50:38 ooh look at this one, go to the last frame and try to guess who won http://72.44.46.68/canvas?game_id=786559 2010-10-31T00:51:09 janzert: so are those your own servers running games right now? 2010-10-31T00:51:29 yes, the others went down at the beginning of the week 2010-10-31T00:51:49 I can try to kick them again now but I noticed they're just not working anymore. 2010-10-31T00:51:51 they tend to need a hard reboot through ec2 every 12 hours or so 2010-10-31T00:52:10 your's will really need to start up fresh so they pull down the latest svn code 2010-10-31T00:53:20 oh right right. 2010-10-31T00:53:36 So I'll have to completely decommission them and create new spot instances you think? 2010-10-31T00:53:41 I am not so familiar with EC2. 2010-10-31T00:53:54 rebooting the EC2 instances.... 2010-10-31T00:53:56 I think if you just terminate them 2010-10-31T00:54:04 ec2 will start fresh instances 2010-10-31T00:54:12 I see I see... 2010-10-31T00:54:33 I'll try that if they dont start playing games in the next 10 minutes. 2010-10-31T00:55:06 the problem is that even if they play games they won't have the c# updates and the golang installation 2010-10-31T00:55:43 I need to sort out a better way to automatically update them 2010-10-31T00:56:00 the easiest might be to just add it to the ec2-check script 2010-10-31T00:56:55 hopefully I can get away from fighting instances soon, I really want to work on the match pairing and some other stuff 2010-10-31T00:57:26 janzert: I have an idea for match pairing 2010-10-31T00:57:37 yes? 2010-10-31T00:57:54 would it help save time if you took a pool of 5-10 bots and did a round robin? 2010-10-31T00:58:01 not really 2010-10-31T00:58:20 I though that might make cleaning up a less frequent operation 2010-10-31T00:58:59 nah, there isn't much spent in clean up time or really even setup time after a worker has been up for an hour or two 2010-10-31T00:59:26 isn't the cleanup, or lack of, what is bringing down the servers? 2010-10-31T00:59:49 no 2010-10-31T01:00:00 at this point it's bugs in hardy on ec2 2010-10-31T01:00:17 where the machines randomly freeze up 2010-10-31T01:00:40 ah 2010-10-31T01:00:46 one problem with the current matchup code is it starts from scratch for every pairing request 2010-10-31T01:01:28 the process of making the matchups needs to be separated from the code that hands them out so it can prepare them in batches with much less database load 2010-10-31T01:01:32 you mean recomputing who's next instead of using a queue? 2010-10-31T01:01:38 right 2010-10-31T01:02:09 when we had 5 workers running at one point it was overloading the server just handing out and processing the results of matchups 2010-10-31T01:02:41 *** iris1 has quit IRC (Read error: Connection timed out) 2010-10-31T01:02:52 that seems odd, you mean just the code to give out 2 user ids? 2010-10-31T01:02:55 also by splitting it up we can easily do a little more intelligent things such as pairing new submissions with someone near their old rank 2010-10-31T01:03:17 because of all the database queries mostly 2010-10-31T01:03:58 that would be great 2010-10-31T01:04:05 then people's rankings would converge way faster 2010-10-31T01:04:12 have any of you guys tried having your bot and it's opponent repeatedly make best responses to each others moves and just hoped that went to something decent? 2010-10-31T01:04:24 i tried to hack it together last night, but i don't think i did a very good job of it 2010-10-31T01:05:55 janzert: what file does the matchmaking? 2010-10-31T01:06:02 antimatroid: are you talking about some sort of iterative algorithm? 2010-10-31T01:06:18 also am going to poke the mathematician who worked on the tron bot with me to try and optimize the pairings to stabilize and improve the ranking as quickly as possible 2010-10-31T01:06:41 http://code.google.com/p/ai-contest/source/browse/trunk/planet_wars/www/api_get_matchup.php 2010-10-31T01:06:48 is the one that hands out to workers 2010-10-31T01:07:54 the other problem with the current method is that one player keeps getting chosen for the first player until a game result is recorded which means people tend to get 2 or 3 games right in a row 2010-10-31T01:08:12 janzert i can tell you're doing a lot of great work here. You should focus on what you're interested in. If you would rather work on the mathcing policy than the server troubles maybe just do that. 2010-10-31T01:08:43 well, I think we have to have servers up though or people aren't going to bother staying around 2010-10-31T01:09:00 but I think I'm really close to those staying up fairly consistently now 2010-10-31T01:09:07 so I will be able to move on 2010-10-31T01:09:52 nice nice thats good to know 2010-10-31T01:10:04 How's it going to work when submissions close - reset scores and go until... what? 2010-10-31T01:10:21 I know I am always glad to be done with puttign out fires so I can work on the cooler algorithmic aspects. 2010-10-31T01:10:36 eburnette: until we run out of time most likely :P 2010-10-31T01:10:52 j3camero: yep, will feel good :) 2010-10-31T01:11:00 dang, nodejs is out? 2010-10-31T01:11:01 eburnette_: dunno we'll just wing it like last time. probably wipe the slate clean, create some new maps, and run people's bots until the rankigns more or less converge. 2010-10-31T01:12:05 how do you tell if map creation is fair? bots tend to have favorite maps 2010-10-31T01:12:07 McLeopold: if you're looking at what I think you are it's in just only running on the main server right now 2010-10-31T01:12:31 eburnette_: haven't thought much about it. what do you think? 2010-10-31T01:12:51 Top 10 players: GreenTea(4187), bocsimacko(4186), felixcoto(4153), Slin-.-(4137), Accoun(4117), dmj111(4102), rsergio(4067), joakim_mjardner(4065), protocolocon(4030), Hazard(4027) 2010-10-31T01:12:53 k going to bed. I'll be around tomorrow during the afternoon. 2010-10-31T01:13:01 see you then 2010-10-31T01:13:47 how about creating a brand new map for each match, using map gen program that is known a couple weeks in advance? 2010-10-31T01:14:21 janzert: if you get around to it, a nicer map choosing algo would be nice to 2010-10-31T01:14:36 McLeopold: yep, I've thought about that as well 2010-10-31T01:14:53 bots are very fickle about the map type 2010-10-31T01:15:06 should avoid repeated matches on the same map if at all possible 2010-10-31T01:15:35 that way you wouldn't get caught with an accidental set of maps that favors one bot 2010-10-31T01:15:55 yeah 2010-10-31T01:15:56 I think the best guard against that is just a rather large pool 2010-10-31T01:16:14 janzert: a large pool of maps 2010-10-31T01:16:20 right 2010-10-31T01:16:22 large pool might work too, say 10k or more 2010-10-31T01:16:26 j3camero: yep 2010-10-31T01:16:28 somebody help me look at this real quick before my eyes cross thinking too hard about turns 2010-10-31T01:16:35 i've tried a whole bunch of things and nothing seems to wor 2010-10-31T01:16:41 http://72.44.46.68/canvas?game_id=787923 invalid order was sending 9 from 16 - the one with 10 on it at the end 2010-10-31T01:16:44 i thinks it's my move selection though, it needs a lot more though 2010-10-31T01:16:44 t 2010-10-31T01:17:01 the details of the game are a lot more complicated than i originally thought 2010-10-31T01:17:13 the way we're doing neutral conflict resolution complicates things A LOT 2010-10-31T01:17:19 i spent hours and hours debugging bot turn vs game turn 2010-10-31T01:17:28 i get easily confused lol 2010-10-31T01:17:46 imo there's nothing wrong with sending 9 from 16 because there's 10 there 2010-10-31T01:17:48 but then, i'm sure any of the alternative resolution methods would have made it any beter 2010-10-31T01:18:12 not sure** 2010-10-31T01:18:14 *** delt0r has joined #aichallenge 2010-10-31T01:18:21 my brain hurts 2010-10-31T01:18:26 my bot sniped! tho :) 2010-10-31T01:18:40 *** delt0r___ has quit IRC (Ping timeout: 255 seconds) 2010-10-31T01:18:40 what is sniping? 2010-10-31T01:18:49 what i did to the bottom planet 2010-10-31T01:18:55 grabbed the neutral right after he conq'd it for me 2010-10-31T01:18:59 demerzel_: kinda, but you wasted 7 just before the snipe 2010-10-31T01:19:00 isn't that just attacking? 2010-10-31T01:19:22 McLeopold: enemy fleet calc - can you tell me why 16 17 9 would be my invalid order? 2010-10-31T01:19:28 bah, yep. 5 workers and the main server is up to a load average of 8 2010-10-31T01:19:38 drop one and it's now dropping back down 2010-10-31T01:19:51 demerzel_: because you did it twice? on turn 15, not 16? 2010-10-31T01:20:05 no i only did it once 2010-10-31T01:20:11 i have it stderr all moves 2010-10-31T01:20:13 antimatroid1: it's attacking at the right time 2010-10-31T01:20:32 McLeopold: your'e right about wasting 7 ships actually - i'm trying to coordinate launches obv it's not working right 2010-10-31T01:20:45 McLeopold: no it's not :P and my offensive moves get done as soon as possible, so it just does that 2010-10-31T01:21:00 demerel_: you need to not take a neutral the same time as the enemy unless you 2010-10-31T01:21:04 re intending to send more than them 2010-10-31T01:21:07 no i know 2010-10-31T01:21:09 it's a bug for sure 2010-10-31T01:21:21 i'm still tyring to figure out why that was invalid 2010-10-31T01:21:30 got the bot paused with all vars in front of me 2010-10-31T01:21:42 also, if you are taking a neutral, it has to be not the future state+1 but the future state + 1 + incoming fleets from them (if your future state has already subtracted that) 2010-10-31T01:21:47 i didn't realise that till a few days agoo 2010-10-31T01:21:51 yup 2010-10-31T01:21:53 demerzel_ maybe the error message is wrong 2010-10-31T01:22:13 i just calc what's left over + 1 2010-10-31T01:22:19 my stderr? 2010-10-31T01:22:27 more correctly max(future_state+1, incoming_fleets+1) 2010-10-31T01:22:29 i'm sure it's invalid tcp has been good that way i just don't know what i did wrong 2010-10-31T01:22:41 whoever says 16 17 9 is invalid 2010-10-31T01:22:43 or if you just run game turns like i do ships[turn] + 1 ;) 2010-10-31T01:22:54 that's the last move i made before it kicked me for invalid 2010-10-31T01:23:05 are you sure you didn't send on turn 15? when it had 8? 2010-10-31T01:23:11 that's what i'm wondering 2010-10-31T01:23:18 if tcp stops at 16 what does that mean tho 2010-10-31T01:23:25 that i gave invalid on 15 or 16 2010-10-31T01:23:58 i like that sometimes it's best to send everything at a planet at equal distance 2010-10-31T01:23:59 i think 15 2010-10-31T01:24:13 turn 15 does not have any moves from you 2010-10-31T01:24:13 ugh i hate gameturn vs bot turn bugs 2010-10-31T01:24:15 i haven't got that in my bot yet though 2010-10-31T01:24:15 so hard to track down 2010-10-31T01:24:31 demerzel_, just start counting at 1 2010-10-31T01:24:32 McLeopold: ok ty 2010-10-31T01:25:01 bot turn = game turn but pre-resolution 2010-10-31T01:25:07 in my implementations anyway 2010-10-31T01:25:13 i do actually run turn logic etc so i have to track it 2010-10-31T01:25:32 what's the difference between game turn and bot turn? aren't they the same? 2010-10-31T01:25:40 hehehe 2010-10-31T01:25:41 no 2010-10-31T01:25:49 I spent a day reviewing the turn orders. 2010-10-31T01:25:52 check all your commands since the last go, any of them could be wrong 2010-10-31T01:26:05 i am only allowing it 1 per turn atm 2010-10-31T01:26:19 virdo: on turn 1 you will have 105 ships 2010-10-31T01:26:27 get game state at check my state, inc turn num, give orders, perform battles, repeat 2010-10-31T01:26:28 virdo: but you can't use them, you're stuck pre-resolution 2010-10-31T01:26:50 so i just implement botturn concept, which is game turn, but prior to growth/fights 2010-10-31T01:26:56 how do you have 105 ships on turn 1? you start with 100 ships. 2010-10-31T01:26:57 I put in assertions in my issue_order function, to verify the moves were valid before issuing them just in case - found a buncha bugs 2010-10-31T01:27:02 technically there is no game/bot turn number 2010-10-31T01:27:21 you get a game state followed by a go, then you send orders followed by a go 2010-10-31T01:27:32 virdo: it does growth and battles AFTER your orders, but the results can't be used until next turn 2010-10-31T01:27:43 virdo: as in you will have 105 ships at the end of the turn but you can't use them till next turn 2010-10-31T01:28:05 if you have state prediction like most of the bots do you know you will have 105 ships at the end of turn 1 but you can't look at it like that you have to say it will be turn 2 2010-10-31T01:28:14 demerzel_: it's splitting hairs, but I would say you have 105 ships at the start of you turn 2010-10-31T01:28:21 next turn 2010-10-31T01:28:24 yes 2010-10-31T01:28:29 that's the best way to implement it i would say 2010-10-31T01:28:29 agreed here, that's turn 2 :) 2010-10-31T01:28:41 but then you have to adjust fleet etas 2010-10-31T01:28:44 you can't start by sending a fleet of 105 2010-10-31T01:28:52 for example 2010-10-31T01:29:01 turn 1 you launch to a 50 ship planet with an eta of 5 2010-10-31T01:29:08 you just have to remember that stuff happens between your go and the next game state 2010-10-31T01:29:12 on turn 6 you will have 1 ship on that planet but you can't use it till turn 7 2010-10-31T01:29:27 demerzel_: you can use the 1 right away 2010-10-31T01:29:27 it's either adjust the etas or adjust all ship calcs back a turn 2010-10-31T01:29:36 no not exaclty 2010-10-31T01:29:40 the eta is what turn it will arrive 2010-10-31T01:29:43 which is after you give orders 2010-10-31T01:29:49 hope 2010-10-31T01:29:51 if it's eta is turn 6 and you try to use it on turn 6 you will invalide 2010-10-31T01:29:55 i mean, nope 2010-10-31T01:29:59 you can't use it till turn 7 2010-10-31T01:30:12 demerzel_: I think you're doing your turns in some whacky way :) 2010-10-31T01:30:20 you can use it at turn 6 2010-10-31T01:30:22 virdo: could be ;) 2010-10-31T01:31:18 McLeopold: I'ma make some interesting maps with small planets and actually type them up, are you keen for some? and I would be interested in other people coming up with some too 2010-10-31T01:31:30 sure 2010-10-31T01:31:52 demerzel_: I hope you caught that you can use ships as soon as they arrive. 2010-10-31T01:32:17 *** eburnette_ has quit IRC (Quit: Page closed) 2010-10-31T01:32:49 McLeopold: i'm probably just not explaining it right - i have predicted all 200 turns accurately for almost a month 2010-10-31T01:33:03 i'm just recoding it and forgot my standards and overthinking it 2010-10-31T01:35:00 *** Rubicon-|-Cross has quit IRC (Quit: Leaving) 2010-10-31T01:35:13 okay, but if distance is 5, send at turn 1, arrive at turn 6, send 1 at turn 6 2010-10-31T01:35:42 ah holy shit i skipped a turn somewhere lol 2010-10-31T01:35:48 that's interesting 2010-10-31T01:35:53 probably didn't get a reply from tcp fast enough 2010-10-31T01:35:53 that logic doesn't prevent you from predicting game state, just holding on to fleets an extra turn 2010-10-31T01:36:15 well you would logically deduct 1 from your eta 2010-10-31T01:36:21 since it's moving this turn 2010-10-31T01:36:30 you just don't do that i guess 2010-10-31T01:36:58 wow this is going to be fun to track down 2010-10-31T01:37:05 you would logically deduct 1 from distance between turns 1 and 2 2010-10-31T01:37:21 so that the first time the opponent sees the fleet it would have eta 4 2010-10-31T01:38:04 orders, advance/growth, battle, repeat 2010-10-31T01:38:40 antimatroid1: http://ai-contest.com/forum/viewtopic.php?f=17&t=1038&p=6835#p6835 2010-10-31T01:38:50 there's my small map 2010-10-31T01:40:29 where's that site where you can paste game results? 2010-10-31T01:41:04 is there a URL for viewing all of your game results? 2010-10-31T01:41:40 there are two -- http://spz.nofate.com/planetwars/visualizer/ and pastebin like here: http://visualizer.naktibalda.lt/ 2010-10-31T01:41:52 coofamani: like the rankings page? 2010-10-31T01:42:13 virdo: no, a history of games for your profile 2010-10-31T01:42:25 virdo: do you know the format of the game file? 2010-10-31T01:42:28 contestbot: click on "My Profile" on the website. 2010-10-31T01:42:29 virdo: User error, it's not my fault. 2010-10-31T01:42:53 nm 2010-10-31T01:43:02 McLeopold: for the spz one? or pastebin? it's the replay file. 2010-10-31T01:43:16 I just read the part of the website where it tells me. :) 2010-10-31T01:45:40 demerzel_: http://visualizer.naktibalda.lt/506 2010-10-31T01:45:53 demerzel_: just to make sure we are on the same page 2010-10-31T01:46:27 demerzel_: notice how the 70 fleet avoids the 105 fleet? 2010-10-31T01:47:22 yes 2010-10-31T01:47:43 my bot has got multi launch down at least 2010-10-31T01:47:43 http://72.44.46.68/canvas?game_id=788111 2010-10-31T01:48:02 it's beautiful! watch from 13 on 2010-10-31T01:48:06 negative growth planet? wonder how different engines handle that ;) (negative ship counts?) 2010-10-31T01:48:22 got 3 separate launches to the neutral on the left :) 2010-10-31T01:48:32 now to figure out why i tried to launch more than i had lol 2010-10-31T01:48:57 bot 13 or game 13 :P 2010-10-31T01:49:03 turn 13 on the one i linked 2010-10-31T01:49:25 still got a 1-off problem tho 2010-10-31T01:49:36 2 fleeets land at the same time but the last one is one behind 2010-10-31T01:50:40 *** coofamani has quit IRC (Quit: leaving) 2010-10-31T01:53:08 hehe, my bot does so much multi launch I'm trying to stop it ;) -- if it doesn't use the fleet for defence, it always sends it out, either routes it, or attacks neutrals / enemy... have to make it delay attacks / coordinate more 2010-10-31T01:55:45 i haven't put in defense yet lol - it's mostly just automatic in that it sees a gain in defending atm 2010-10-31T01:57:44 well, mine in brain dead too then... if I see it I lose it in the future, I check if I can get to it in time to defend it, and I just send towards it the amount needed. 2010-10-31T01:58:02 if i could get this damn invalid order bug tracked down argh 2010-10-31T01:58:52 demerzel_: you got some games with 4-5 turns and only 1 planet. It should be easy to track down from them. 2010-10-31T01:59:01 http://72.44.46.68/canvas?game_id=787763 2010-10-31T01:59:08 you log for yourself the state of the game / fleet orders? so you log sending X amount of ships from planet Y that has Z ships available still? :) 2010-10-31T01:59:59 are you sending 0 size fleets? 2010-10-31T02:00:05 are you sending to the same planet? 2010-10-31T02:00:14 no i know what orders i'm sending 2010-10-31T02:00:15 are you sending from a planet you don't own? 2010-10-31T02:00:25 nope just sending more than i have in 2 different missions 2010-10-31T02:00:35 ah 2010-10-31T02:00:37 stupid thing is both planet state and supply state both know i don't have that much 2010-10-31T02:00:42 so it's difficult to figure out where it's not catching it 2010-10-31T02:01:45 in my bot, sometimes I started adding min(planet.ships_available, ships_i_want_to_send) 2010-10-31T02:01:52 is it only happening when you send 2 fleets from the same planet? 2010-10-31T02:03:07 it's sending fleets too early basically 2010-10-31T02:03:36 i'll find it ;) it knows the turn it's supposed to get there and it knows how much it really has 2010-10-31T02:03:40 but for whatever reason it's sending off anyway 2010-10-31T02:05:01 guess i just need to dump it all too much for stderr 2010-10-31T02:07:04 dump it into a file 2010-10-31T02:07:36 update visualize_localy.py and integrate it with the visualizer.js ;) 2010-10-31T02:08:06 at least that's the best method I've found so far 2010-10-31T02:08:16 but definitely interested in ways everyone else debugs their bots 2010-10-31T02:12:17 i used to dump all info but now i am thinking i have an ide just dump game state and then recreate and step throug 2010-10-31T02:12:54 Top 10 players: bocsimacko(4190), GreenTea(4189), Slin-.-(4141), felixcoto(4139), dmj111(4126), Accoun(4122), rsergio(4072), joakim_mjardner(4070), protocolocon(4040), rs5147(4035) 2010-10-31T02:13:21 demerzel_: I totaly need that too 2010-10-31T02:13:27 http://codepad.org/OmCCZZDq 2010-10-31T02:13:31 there's 4 maps to start with 2010-10-31T02:13:36 an ide just to review my logs files 2010-10-31T02:13:38 they're not very hard to work out 2010-10-31T02:13:44 i don't think 2010-10-31T02:13:54 now i'ma move to 2d 2010-10-31T02:13:55 well you might as ewll use the ide 2010-10-31T02:14:04 lot more work to make a bunch of customized data dumps 2010-10-31T02:16:59 my current setup: http://spz.nofate.com/planetwars/debug.png ... the debug info on the right side is per turn... how do you guys do it? I may want to copy your methods ;) 2010-10-31T02:17:53 let me know what you guys think are the best moves on each of those maps if you can be bothered 2010-10-31T02:18:38 mcleopold will probably find i got numbers wrong again, i think i got it right though 2010-10-31T02:20:09 antimatroid1: the second one should be attack the center 2010-10-31T02:20:13 the first, wait 2010-10-31T02:20:44 the 3rd and 4th, wait 2010-10-31T02:21:27 i think you should do your neutral on the third, hang on i'll check again 2010-10-31T02:22:18 yeah, pretty sure you can defend your claim of your neutral in 3 2010-10-31T02:22:26 yes, your right 2010-10-31T02:22:42 if opponent lanches far first turn, you can hop and gain 5 2010-10-31T02:22:53 doesn't even matter 2010-10-31T02:22:56 you can back it up 2010-10-31T02:23:09 if you into that... 2010-10-31T02:23:26 i'm not sure hoping is better for you 2010-10-31T02:23:47 cause you incure the 10 cost to take the neutral 2010-10-31T02:23:54 and you already lost it on your neutral 2010-10-31T02:24:46 okay, down 10, up 5 2010-10-31T02:25:24 i think best strategy is to take your neutral first move with 11, then send back up second move if they attacked 2010-10-31T02:26:03 well, since there is nothing interesting by your starting planet, you might as well send it forward 2010-10-31T02:26:10 The full 100 2010-10-31T02:26:17 and stream the rest 2010-10-31T02:26:23 yes, but assuming there were other choices, and this was a subgame 2010-10-31T02:27:05 only if those other choices are better than the 2 center planets combined 2010-10-31T02:27:08 i'm trying to work out specific details of the game better, i'm convinced that's why my bot sucks 2010-10-31T02:27:29 these types of maps are excatly what i've been working on for a month 2010-10-31T02:27:39 but they might expand to other places later, enabling you to spend the other 89 ships expanding elsewhere too 2010-10-31T02:27:54 so you don't want to send everything right away 2010-10-31T02:28:02 only what you HAVE to 2010-10-31T02:28:04 *** Rubicon-|-Cross has joined #aichallenge 2010-10-31T02:28:35 even if it is a sub-game, you still want to neutralize the enemy's ships 2010-10-31T02:29:15 yeah, but there might be something you can expand to with 15 ships near your home planet, you can't take that now, but if they send that much out to somewhere nowhere near the centre first turn, then you can expand with it 2010-10-31T02:29:20 if that makes sense 2010-10-31T02:30:55 it does, but it depends on where the front line is 2010-10-31T02:31:11 not moving to the center is like retreating 2010-10-31T02:31:38 what's behind you has too be worth 21 growth or more, and the enemy has to have nothing behind them 2010-10-31T02:35:32 i think the information advantage might be better, remember you only lose one turn 2010-10-31T02:37:20 the closer planet is even, the furthur is 2, due to reaction time 2010-10-31T02:37:34 you only control 1 planet at the start 2010-10-31T02:37:53 if you can move to the close neutral, and your opponent doesn't, you control 2 2010-10-31T02:37:57 gtg 2010-10-31T02:38:15 * raedwulf needs to find that sheet of paper with his strategy on 2010-10-31T02:42:42 *** virdo has quit IRC (Ping timeout: 240 seconds) 2010-10-31T02:52:15 *** Appleman1234 has joined #aichallenge 2010-10-31T02:56:55 *** McLeopold has quit IRC (Ping timeout: 240 seconds) 2010-10-31T03:09:58 that wasnt' too bad to set up 2010-10-31T03:10:08 streams can be useful lol 2010-10-31T03:12:52 Top 10 players: bocsimacko(4213), GreenTea(4202), felixcoto(4135), dmj111(4119), Accoun(4103), Slin-.-(4098), rsergio(4079), joakim_mjardner(4075), Hazard(4035), protocolocon(4024) 2010-10-31T03:14:59 *** iNtERrUpT has joined #aichallenge 2010-10-31T03:23:18 janzert, I'm around, still want me? I just updated the go starter package. 2010-10-31T03:24:37 *** Utkarsh has quit IRC (Ping timeout: 264 seconds) 2010-10-31T03:25:46 I'm back from a go tournament, ready to go write some go code, hehe... 2010-10-31T03:30:45 *** Utkarsh has joined #aichallenge 2010-10-31T03:37:34 lavalamp_: go? damn 2010-10-31T03:37:42 isn't that like high end ai shit nowadays lol 2010-10-31T03:37:51 is there a go competition after this or something 2010-10-31T03:40:47 no I play go competitively 2010-10-31T03:41:04 where do people find these ai things for go? 2010-10-31T03:41:13 i'd be keen to have a play 2010-10-31T03:41:16 language or game? 2010-10-31T03:41:24 game 2010-10-31T03:42:39 I don't know if there is a contest running for a go ai or not 2010-10-31T03:43:02 at some points there has been prize money available for making bots of a certain level 2010-10-31T03:43:19 surely there's a place people writing bots test against each other 2010-10-31T03:43:47 oh, yeah, there's the equivilent of the tcp server 2010-10-31T03:44:14 there's also a script so you can hook up a go bot to the KGS go server (gokgs.com) 2010-10-31T03:45:15 where would one find the equivalent of the tcp server is my question? :P 2010-10-31T03:45:42 janzert didn't say if golang was enabled or not, I just uploaded a submission to find out... 2010-10-31T03:46:05 http://www.cgl.ucsf.edu/go/ladder.html 2010-10-31T03:47:05 cheers 2010-10-31T03:51:26 grr it says there were errors compiling but it's not emailing me 2010-10-31T03:51:35 at least it identified the language 2010-10-31T03:52:15 hmm, lavalamp_ it did work for me earlier 2010-10-31T03:52:28