2010-09-25T00:00:10 all of my games are ending like that 2010-09-25T00:00:47 I think the tcp program will normally tell you why a game ends early 2010-09-25T00:01:25 actually http://72.44.46.68/getplayer?player=sb8244 shows it as being a loss by illegal move 2010-09-25T00:01:56 is that like sending things to a null planet? 2010-09-25T00:02:25 or sending more ships than a planet has, etc. yeah 2010-09-25T00:02:38 sending from a planet you don't own 2010-09-25T00:02:45 which is weird, i don't konw why its doing that 2010-09-25T00:02:48 that NEVER happens against bots 2010-09-25T00:03:49 i'm confused as to why that's happening, it shouldnt be 2010-09-25T00:04:08 time to log the games and start debugging :/ 2010-09-25T00:04:21 *** rwa has joined #aichallenge 2010-09-25T00:05:23 !seen j3camero 2010-09-25T00:05:23 AlliedEnvy: j3camero was last seen in #aichallenge 1 week, 2 days, 1 hour, 15 minutes, and 13 seconds ago: Cyndre: are you interested in building it? That would be soooo bitchin. I have seen some simple online IDEs before, and I gotta say it's pretty cool. We could build something unprecedented! 2010-09-25T00:05:32 lmao 2010-09-25T00:05:38 every time you do that I swing back in 2010-09-25T00:05:53 Ahaha, sorry. 2010-09-25T00:06:06 oh dont worry abuot, just find it funny 2010-09-25T00:06:13 heh, I've just been doing it in pm's to contestbot 2010-09-25T00:08:42 i found the error and i think it's in the visualizer / tcp game code 2010-09-25T00:08:52 because it doesn't happen on the contest visualizer 2010-09-25T00:09:00 http://72.44.46.68/canvas?game_id=149690 2010-09-25T00:09:13 Right before i get my planet taken, i have ships from that planet out 2010-09-25T00:09:23 then i crash when the planet is taken, with ships still on the field 2010-09-25T00:09:38 the ships should'nt be affected by the defeated planet though 2010-09-25T00:10:16 visualizer can't affect result of your game 2010-09-25T00:10:33 well i mean the game code then 2010-09-25T00:10:38 that DOESNT happen in contest situation 2010-09-25T00:11:55 are you running identical version of your bot? 2010-09-25T00:12:13 http://72.44.46.68/getplayer?player=sb8244 2010-09-25T00:12:30 Every one of the games was lost due to invalid orders 2010-09-25T00:13:11 tcp server ignored invalid orders before 2010-09-25T00:13:22 does official server ignore them? 2010-09-25T00:13:32 idk? I never crash due to invalid orders 2010-09-25T00:13:37 i dont even know WHAT invalid order s i have 2010-09-25T00:13:55 log your orders :) 2010-09-25T00:14:03 that's why you really need to log the game and see what's actually happening 2010-09-25T00:14:17 sending less than one ship? sending from a planet you don't own? from/to the same planet? To a nonexistent planet? 2010-09-25T00:14:40 Also what's the best way to log in java 2010-09-25T00:14:42 write to a file? 2010-09-25T00:15:07 is that even legal? I know system.out.println() will crash with an invalid order 2010-09-25T00:15:14 as you can see from my round 0 loss 2010-09-25T00:15:27 Yup. Just make sure you remove the file printing code before submitting officially 2010-09-25T00:16:04 I think even writing to stderr with tcp server 2010-09-25T00:16:13 Man. StringTokenizer is so much faster than Scanner. 2010-09-25T00:16:58 If you wish to create a Planet Wars bot from scratch, you must first invent the universe. ;) 2010-09-25T00:17:00 will the tcp server only take my IssueOrder commands? 2010-09-25T00:17:05 it won't try to run my file code 2010-09-25T00:17:19 correct, sb8244 2010-09-25T00:17:20 hmm just thought of it, does the tee command flush often enough to be useful here? 2010-09-25T00:25:31 *** Baphomet has quit IRC (Ping timeout: 240 seconds) 2010-09-25T00:27:54 *** syntaxglitch has quit IRC (Quit: return ()) 2010-09-25T00:30:06 *** Appleman1234 has quit IRC (Read error: Connection reset by peer) 2010-09-25T00:30:20 *** sb8244 has quit IRC (Ping timeout: 252 seconds) 2010-09-25T00:30:26 *** Baphomet has joined #aichallenge 2010-09-25T00:57:15 Top 10 players: dmj111(3456), kcm1700(3438), felixcoto(3430), McLeopold(3390), EBraun(3355), wagstaff(3316), GreenTea(3302), adumlah(3284), Infinity(3276), Zilog8(3259) 2010-09-25T01:08:46 Janzert: i use tee with my bot output 2010-09-25T01:09:03 tee flushes when its input flushes; as long as you setlinebuf(stdout) you're good 2010-09-25T01:09:14 *** hellman has joined #aichallenge 2010-09-25T01:20:28 ahh, thanks a1kon that should be handy 2010-09-25T01:33:21 *** jimmerton has joined #aichallenge 2010-09-25T01:46:51 *** sinan has joined #aichallenge 2010-09-25T01:55:27 *** jimmerton_ has joined #aichallenge 2010-09-25T01:57:17 Top 10 players: dmj111(3510), kcm1700(3439), McLeopold(3395), EBraun(3352), felixcoto(3337), wagstaff(3323), GreenTea(3317), Infinity(3290), adumlah(3289), sighbot(3268) 2010-09-25T01:57:37 *** jimmerton has quit IRC (Ping timeout: 252 seconds) 2010-09-25T01:57:37 *** jimmerton_ is now known as jimmerton 2010-09-25T01:57:45 hey! back in the top 10! 2010-09-25T01:58:07 @yay 2010-09-25T01:58:07 Double rainbow all the way! 2010-09-25T01:58:12 :) 2010-09-25T01:58:22 Current Rank: 118 2010-09-25T01:58:24 ouch 2010-09-25T01:58:31 a1k0n: lol 2010-09-25T01:58:31 >< 2010-09-25T01:58:31 time to get with the program 2010-09-25T02:00:45 *** Appleman1234 has joined #aichallenge 2010-09-25T02:05:53 *** midorikid has joined #aichallenge 2010-09-25T02:12:29 *** alexcg has joined #aichallenge 2010-09-25T02:14:52 *** rtaycher has quit IRC (Remote host closed the connection) 2010-09-25T02:16:54 *** mega1 has joined #aichallenge 2010-09-25T02:22:59 *** mega1_ has joined #aichallenge 2010-09-25T02:23:51 *** Utkarsh has quit IRC (Ping timeout: 276 seconds) 2010-09-25T02:27:32 *** Utkarsh has joined #aichallenge 2010-09-25T02:28:41 *** millertime has quit IRC (Quit: WeeChat 0.3.2) 2010-09-25T02:36:56 *** other_donkey has joined #aichallenge 2010-09-25T02:55:03 anybody else have the tcp visualizer jump around a lot as the text wraps and unwraps? 2010-09-25T02:55:43 yeah, the ships/growth counts tend to do that 2010-09-25T02:57:17 Top 10 players: dmj111(3490), kcm1700(3490), McLeopold(3382), EBraun(3359), felixcoto(3335), wagstaff(3315), GreenTea(3314), sighbot(3261), Infinity(3260), Tracer(3255) 2010-09-25T02:59:32 hmmm... I don't have that problem, but if it is the text wrapping maybe Ctrl- to make the font size smaller? 2010-09-25T03:00:59 didn't seem to work 2010-09-25T03:02:50 the hardcore solution is to use greasemonkey (or equivalent for your browser) to change the visualiser :D 2010-09-25T03:03:00 or make a bookmarklet 2010-09-25T03:06:48 hmm, the official engine doesn't write async to the bots, right? what if its write buffer is too small for an entire input (many fleets...), and the first bot sets its read buffer to 1, read up to the 'g' in 'go', then blocks the engine doing calculations for 10 seconds? :> 2010-09-25T03:07:28 *** Theorevery has left #aichallenge ("Leaving...") 2010-09-25T03:07:37 i guess that would be cheating, but maybe the engine should do blocking io, to prevent blocking on the write to the first player. 2010-09-25T03:07:49 s/should/shouldn't/ 2010-09-25T03:08:33 dhartmei: interesting 2010-09-25T03:08:45 * sigh goes and looks at engine code 2010-09-25T03:09:50 *** Zeiris has quit IRC (Ping timeout: 240 seconds) 2010-09-25T03:10:29 http://code.google.com/p/ai-contest/source/browse/trunk/planet_wars/backend/engine.py?r=180 2010-09-25T03:10:42 looks like a blocking write, first to the first player, then to the second 2010-09-25T03:11:22 at least the code doesn't deal with incomplete writes, re-writing remainders, etc. 2010-09-25T03:38:42 *** boegel has joined #aichallenge 2010-09-25T03:44:19 *** Baphomet has quit IRC (Ping timeout: 252 seconds) 2010-09-25T03:47:06 *** Itkovian has quit IRC (Quit: Itkovian) 2010-09-25T03:48:19 *** Baphomet has joined #aichallenge 2010-09-25T03:56:48 *** Accoun has joined #aichallenge 2010-09-25T03:57:18 Top 10 players: dmj111(3489), kcm1700(3474), EBraun(3366), McLeopold(3360), felixcoto(3344), wagstaff(3310), GreenTea(3308), sighbot(3287), Zilog8(3286), Tracer(3256) 2010-09-25T03:59:58 *** delt0r has joined #aichallenge 2010-09-25T04:04:07 *** midorikid has quit IRC (Quit: Page closed) 2010-09-25T04:06:56 should def go to bed now to retain the illusion that my latest bot can be McLeopold.6 consistently 2010-09-25T04:15:23 *** Palmik has joined #aichallenge 2010-09-25T04:15:25 *** Baphomet has quit IRC (Ping timeout: 245 seconds) 2010-09-25T04:17:05 *** mega1_ has quit IRC (Ping timeout: 245 seconds) 2010-09-25T04:17:05 *** mega1 has quit IRC (Ping timeout: 245 seconds) 2010-09-25T04:19:23 *** RainCT has joined #aichallenge 2010-09-25T04:19:35 *** Baphomet has joined #aichallenge 2010-09-25T04:22:38 *** choas has joined #aichallenge 2010-09-25T04:27:10 *** mega1 has joined #aichallenge 2010-09-25T04:27:16 *** mega1_ has joined #aichallenge 2010-09-25T04:30:22 *** Baphomet has quit IRC (Read error: Connection reset by peer) 2010-09-25T04:30:45 *** Baphomet has joined #aichallenge 2010-09-25T04:30:48 *** mega1_ has quit IRC (Remote host closed the connection) 2010-09-25T04:32:24 *** kozlovsky has joined #aichallenge 2010-09-25T04:32:36 it seems to be longer than usual before a new submission is being picked up by the official server... is this known? 2010-09-25T04:32:54 *** Utkarsh has quit IRC (Ping timeout: 252 seconds) 2010-09-25T04:33:32 *** BtbN has joined #aichallenge 2010-09-25T04:35:15 it picked it up now, but it took a few minutes, whereas before it was almost immediately 2010-09-25T04:38:50 *** smellyhippy has quit IRC (Ping timeout: 240 seconds) 2010-09-25T04:39:25 *** Utkarsh has joined #aichallenge 2010-09-25T04:51:36 *** Utkarsh has quit IRC (Read error: Connection reset by peer) 2010-09-25T04:57:19 Top 10 players: kcm1700(3486), dmj111(3475), EBraun(3377), McLeopold(3372), felixcoto(3339), wagstaff(3321), GreenTea(3320), Zilog8(3299), Tracer(3269), Longing(3250) 2010-09-25T04:58:21 well that was an interesting grand final 2010-09-25T04:58:26 thanks for coming see you next week 2010-09-25T05:00:45 *** delt0r has quit IRC (Read error: Operation timed out) 2010-09-25T05:01:35 *** customKoncept has joined #aichallenge 2010-09-25T05:04:37 *** Utkarsh has joined #aichallenge 2010-09-25T05:06:06 *** customKoncept has quit IRC (Client Quit) 2010-09-25T05:13:01 *** sinan has quit IRC (Ping timeout: 276 seconds) 2010-09-25T05:14:34 *** Utkarsh has quit IRC (Ping timeout: 272 seconds) 2010-09-25T05:19:18 *** Itkovian has joined #aichallenge 2010-09-25T05:19:49 *** Utkarsh has joined #aichallenge 2010-09-25T05:31:49 *** xvinyl has quit IRC (Remote host closed the connection) 2010-09-25T05:32:45 dmj111 is moving up up up 2010-09-25T05:33:39 wow, a win/loss ratio of over 90% (72 wins, 7 losses) 2010-09-25T05:33:42 that's very nice 2010-09-25T05:41:12 *** smellyhippy has joined #aichallenge 2010-09-25T05:41:12 *** smellyhippy has joined #aichallenge 2010-09-25T05:46:35 smellyhippy: cool nick :) 2010-09-25T05:50:47 *** Itkovian has quit IRC (Quit: Itkovian) 2010-09-25T05:54:20 *** aerique has joined #aichallenge 2010-09-25T05:56:45 *** kjeldahl has joined #aichallenge 2010-09-25T05:57:20 Top 10 players: kcm1700(3481), dmj111(3472), EBraun(3393), McLeopold(3359), felixcoto(3316), wagstaff(3312), GreenTea(3306), Zilog8(3287), Tracer(3268), Longing(3243) 2010-09-25T05:57:32 boegel: thanks :P 2010-09-25T06:02:20 dmj111 is going to take over the lead 2010-09-25T06:02:48 I think 2010-09-25T06:02:49 soon 2010-09-25T06:02:50 ish 2010-09-25T06:05:17 *** tapwater has quit IRC (Quit: tapwater) 2010-09-25T06:05:41 *** customKoncept has joined #aichallenge 2010-09-25T06:12:13 *** jaspervdj has joined #aichallenge 2010-09-25T06:19:34 *** choas has quit IRC (Quit: leaving) 2010-09-25T06:20:22 *** Itkovian has joined #aichallenge 2010-09-25T06:23:21 *** sugarshark has joined #aichallenge 2010-09-25T06:29:17 *** otherAntimatroid has joined #aichallenge 2010-09-25T06:30:25 *** amstan has quit IRC (Ping timeout: 245 seconds) 2010-09-25T06:39:18 *** sinan has joined #aichallenge 2010-09-25T06:41:49 *** Itkovian has quit IRC (Quit: Itkovian) 2010-09-25T06:42:26 *** dhartmei_ has quit IRC (Quit: Leaving) 2010-09-25T06:57:19 Top 10 players: kcm1700(3476), dmj111(3463), EBraun(3389), McLeopold(3355), wagstaff(3308), GreenTea(3303), felixcoto(3290), Zilog8(3283), Tracer(3264), Longing(3245) 2010-09-25T06:58:01 hmm, ELO of top 3 is dropping 2010-09-25T06:58:30 seems like 7-8-9 are winning games 2010-09-25T07:03:46 *** aerique has quit IRC (Quit: ...) 2010-09-25T07:13:38 *** boegel has quit IRC (Quit: Leaving) 2010-09-25T07:13:54 *** sinan has quit IRC (Ping timeout: 252 seconds) 2010-09-25T07:16:16 *** Palmik has quit IRC (Read error: Operation timed out) 2010-09-25T07:19:22 dhartmei: The tournament manager isn't using blocking on game reads 2010-09-25T07:19:30 dhartmei: *player reads 2010-09-25T07:19:43 dhartmei: It kicks off one thread for each player 2010-09-25T07:19:49 dhartmei: and blocks there 2010-09-25T07:20:03 dhartmei: I'm going to be changing it to use select though 2010-09-25T07:28:19 *** u1tr0n has joined #aichallenge 2010-09-25T07:34:16 I had games in my profile, and now they have disappeared :( 2010-09-25T07:36:56 so did i 2010-09-25T07:37:00 but it improved my score :| 2010-09-25T07:37:31 antimatroid: I can see games in your profile 2010-09-25T07:37:44 mine has none: http://www.ai-contest.com/profile.php?user_id=6041 2010-09-25T07:38:11 I uploaded a new bot a couple of hours ago and it had played ~10 games 2010-09-25T07:38:47 yeah, but i saw games there that were new that are now gone 2010-09-25T07:39:01 ie. Beanhome beat me, that's gone 2010-09-25T07:39:08 something screwing is going on 2010-09-25T07:39:09 i thought he resubmitted and bumped my score, but he hadn't 2010-09-25T07:39:13 *screwy 2010-09-25T07:41:04 sigh: yes 2010-09-25T07:41:17 sign: Server had issues this morning 2010-09-25T07:41:32 sigh: spent from midnight to 6 overloaded 2010-09-25T07:41:51 ah, ok 2010-09-25T07:41:55 sigh: This probably caused most games to end in unreal timeouts 2010-09-25T07:41:55 so those games are lost? 2010-09-25T07:42:21 sigh: I tossed those couple hours of games 2010-09-25T07:42:35 ah, i see 2010-09-25T07:42:49 that would make sense 2010-09-25T07:42:53 thanks for the heads-up 2010-09-25T07:42:54 sigh: did you see the afl tie before? :P 2010-09-25T07:43:07 antimatroid: not really into footy :P 2010-09-25T07:43:24 neither, but hadn't seen some friends in a while so thought i had better go and say hello 2010-09-25T07:43:40 i have an epic beard going atm, i get referred to as jesus 2010-09-25T07:43:43 fair enough 2010-09-25T07:44:33 it's now taking like 3-4 minutes just to see if my changes do anything on a game against a default bot :| 2010-09-25T07:44:44 why? 2010-09-25T07:44:48 still not sure if i'm going to be able to get this approach below 1 second a move 2010-09-25T07:44:55 cause it's taking AGES to decide on some moves 2010-09-25T07:44:56 oh, i see 2010-09-25T07:44:59 *** wh1teside has joined #aichallenge 2010-09-25T07:45:13 what are you doing? 2010-09-25T07:45:20 :P that's secret 2010-09-25T07:45:20 or is that a secret :P 2010-09-25T07:45:31 still using less than 50mb memory most of the time 2010-09-25T07:46:21 *** Itkovian has joined #aichallenge 2010-09-25T07:46:39 I can't think of anything for my bot that would require a lot of processing yet 2010-09-25T07:46:46 *** Palmik has joined #aichallenge 2010-09-25T07:47:37 i/we have a couple of different ideas to try out/mash together 2010-09-25T07:47:40 and see what we can get 2010-09-25T07:56:03 *** Itkovian has quit IRC (Quit: Itkovian) 2010-09-25T07:57:21 Top 10 players: kcm1700(3451), dmj111(3444), EBraun(3371), McLeopold(3353), wagstaff(3309), felixcoto(3303), GreenTea(3280), Zilog8(3258), Tracer(3252), bix0r4ever(3232) 2010-09-25T08:07:32 *** jimmerton_ has joined #aichallenge 2010-09-25T08:10:16 *** jimmerton has quit IRC (Ping timeout: 265 seconds) 2010-09-25T08:10:16 *** jimmerton_ is now known as jimmerton 2010-09-25T08:18:58 hmmm... still no games 2010-09-25T08:19:30 maybe because I'm have no rank, and I'm not a new submission the tournament engine is ignoring me 2010-09-25T08:19:39 I'll try uploading my bot again 2010-09-25T08:19:48 Is there a better player than the java one? My bot seems to do an invalid move, but i can't get what's wrong with it 2010-09-25T08:22:07 BtbN: try looking in log.txt first, it logs all the commands that the engine sees... so you can check if one is invalid 2010-09-25T08:22:31 there is only one that can be invalid, but i don't find out why it is invalid 2010-09-25T08:22:56 sigh: definitely reupload if you haven't yet 2010-09-25T08:23:08 i wish i could stop a game mid way through and see what's happened 2010-09-25T08:23:18 it's at turn 93 and i expect it'll probably go to 200 2010-09-25T08:23:23 been going for a good 10 minutes aha 2010-09-25T08:23:33 why can't you stop it? 2010-09-25T08:24:08 cause then it'd close the viewer too? 2010-09-25T08:24:17 and i want to see what's happened 2010-09-25T08:24:31 it's onto move 95 2010-09-25T08:24:31 Woah 2010-09-25T08:24:41 kill the engine processs specifically? 2010-09-25T08:24:47 By some miracle I'm 23rd on the official server 2010-09-25T08:24:52 :) 2010-09-25T08:25:09 This won't last long 2010-09-25T08:25:12 my 11th seems like more of a lie 2010-09-25T08:25:26 especially after it had a loss to someone before that got removed 2010-09-25T08:25:40 alright, new submission got a game! 2010-09-25T08:26:05 now i have to wait for however long the time limit ended up being :| 2010-09-25T08:26:07 heeh 2010-09-25T08:26:07 against someone ranked 400 as well, so I don't have to start from 1200 like last time -__ 2010-09-25T08:26:08 Are you antimatroid there too? 2010-09-25T08:26:14 nope, "youwillfail" 2010-09-25T08:26:18 *** antimatroid is now known as youwillfail 2010-09-25T08:26:23 i was antimatroid last time sorry 2010-09-25T08:26:29 *** sigh is now known as antimatroid 2010-09-25T08:26:36 *** youwillfail is now known as sigh 2010-09-25T08:26:39 touche :P 2010-09-25T08:26:46 :P 2010-09-25T08:26:58 *** sigh is now known as youwillfail 2010-09-25T08:27:04 *** antimatroid is now known as sigh 2010-09-25T08:27:45 lol, i wouldn't want to come in now and try to work that out 2010-09-25T08:27:54 haha... I saw this: "antimatroid is now known as sigh"... and I was like, wtf how can he change to sigh when I just took sigh 2010-09-25T08:28:19 * sigh fail 2010-09-25T08:28:23 bit like futurama recently 2010-09-25T08:28:27 <3 that episode 2010-09-25T08:28:32 yup :) 2010-09-25T08:28:55 that proof was pretty interesting 2010-09-25T08:29:05 yeah, i didn't really look at it 2010-09-25T08:29:15 did a unit on group theory last semester though, that was cool 2010-09-25T08:29:20 not too hard to follow 2010-09-25T08:29:56 i keep seeing groups all over the place these days, it's like ahuh i can thrust with that 2010-09-25T08:30:20 I only have a basic understanding of group theory 2010-09-25T08:30:46 enough to follow thing when people talk about it, but not enough to know when to use it 2010-09-25T08:30:51 my topics teacher set an assignment the other week to basically show the symmetric difference operation on the set of allowable sets is a group 2010-09-25T08:31:42 we had an algebra assignment to show an operation defined on the fano plane forms a group, then we managed to extend that out and find group tables for fano planes up to like 8 dimensions 2010-09-25T08:31:43 what class is that for? 2010-09-25T08:31:49 topics in adv. maths 2010-09-25T08:32:04 the group table for the 2d case is so amazingly symmetric 2010-09-25T08:32:14 * sigh doesn't know what a fano plane is 2010-09-25T08:32:16 not even in just the normal symmetric ways 2010-09-25T08:32:27 http://en.wikipedia.org/wiki/Fano_plane 2010-09-25T08:32:53 that, basically opeation is on the points, and for any two, you get the third unique point on a line segment with the other two 2010-09-25T08:33:26 turns out you can extend that up dimensions and keep it still working, we managed to work out the geometry of the 3d one, but haven't been able to make a model of it yet, tried with paper but got too messy 2010-09-25T08:33:36 i see 2010-09-25T08:34:18 i wish i had done more maths 2010-09-25T08:34:34 *** zaphus has joined #aichallenge 2010-09-25T08:34:36 yeah, i can't get enough of it these days 2010-09-25T08:34:37 hello sir 2010-09-25T08:35:26 *** jaspervdj has quit IRC (Quit: NEVER GONNA GIVE YOU UP NEVER GONNA LET YOU DOWN) 2010-09-25T08:37:16 haha 2010-09-25T08:38:42 *** zaphus has quit IRC (Client Quit) 2010-09-25T08:46:46 *** customKoncept has quit IRC (Quit: Page closed) 2010-09-25T08:47:23 *** vikhyat has joined #aichallenge 2010-09-25T08:50:17 *** jimmerton has quit IRC (Quit: jimmerton) 2010-09-25T08:52:02 *** Olathe has joined #aichallenge 2010-09-25T08:57:20 Top 10 players: kcm1700(3469), dmj111(3454), EBraun(3387), adumlah(3372), McLeopold(3370), wagstaff(3324), felixcoto(3300), GreenTea(3295), Zilog8(3274), Tracer(3267) 2010-09-25T09:03:13 *** Appleman1234 has quit IRC (Quit: Leaving) 2010-09-25T09:04:56 sigh: ouch, you lost already :P 2010-09-25T09:05:13 haha... yeah, that games has a lot of fail 2010-09-25T09:06:47 opponent is ranked 51 so it's not *that* bad 2010-09-25T09:08:40 that's an annoying map :( 2010-09-25T09:08:55 yep 2010-09-25T09:09:07 but the close starts seem to lead to harder matches 2010-09-25T09:09:17 so i think they should perhaps be more abundant 2010-09-25T09:11:32 possibly 2010-09-25T09:11:42 they also makes draws more likely though 2010-09-25T09:18:16 *** sinan has joined #aichallenge 2010-09-25T09:19:45 depends, that's usually when the absolutely obvious best choice is to do nothing 2010-09-25T09:20:23 i have seen very few draws so far other than on maps like that 2010-09-25T09:20:33 although stalemate like maps are pretty boring too 2010-09-25T09:20:55 where one managed to claim slightly more, then they sit and wait indefinitely 2010-09-25T09:36:27 *** vikhyat has quit IRC (Quit: Leaving) 2010-09-25T09:38:31 Which version of the gmcs-compiler does the Server use? Does it support linq/.NET 3.5? 2010-09-25T09:57:20 Top 10 players: dmj111(3447), kcm1700(3446), adumlah(3385), McLeopold(3373), EBraun(3366), blubby(3336), wagstaff(3323), felixcoto(3292), GreenTea(3289), Tracer(3268) 2010-09-25T10:03:12 .NET 2.0 IIRC 2010-09-25T10:03:55 hm, i'll just try to upload my code, and see if he compiles it 2010-09-25T10:05:39 error while compiling submission, hrmpf 2010-09-25T10:12:42 *** Palmik has quit IRC (Read error: Operation timed out) 2010-09-25T10:17:46 *** kozlovsky has quit IRC (Ping timeout: 252 seconds) 2010-09-25T10:31:07 *** Cyndre has quit IRC (Read error: Connection reset by peer) 2010-09-25T10:31:22 *** Cyndre has joined #aichallenge 2010-09-25T10:34:13 *** Itkovian has joined #aichallenge 2010-09-25T10:36:03 *** otherAntimatroid has left #aichallenge 2010-09-25T10:38:35 *** Itkovian has quit IRC (Client Quit) 2010-09-25T10:57:21 Top 10 players: dmj111(3445), kcm1700(3435), adumlah(3395), McLeopold(3371), EBraun(3365), felixcoto(3317), wagstaff(3312), bix0r4ever(3285), GreenTea(3283), Tracer(3254) 2010-09-25T11:13:26 bartwe: http://72.44.46.68/canvas?game_id=155971 :P 2010-09-25T11:25:40 wow quick death :) 2010-09-25T11:28:22 nullkuhl: I've killed bartwe in 7 moves before :) 2010-09-25T11:29:56 I have never seen enemy planet so close :) 2010-09-25T11:30:48 yeah, those maps are hard to play 2010-09-25T11:30:57 I don't see a way to expand in this map 2010-09-25T11:31:05 my bot got owned on the main server because of a close map 2010-09-25T11:31:22 distance 2, no planets with num_ships <10 2010-09-25T11:32:58 yeah, doesn't look like there is anything to do 2010-09-25T11:33:08 looks like stalemate 2010-09-25T11:34:00 well, if one player sends 10 ships out, another can send up to 20 and capture a planet 2010-09-25T11:34:16 but probably a draw is the best result 2010-09-25T11:34:51 first mover loses 2010-09-25T11:35:16 Janzert: not if the first mover just attacks the opponent :P 2010-09-25T11:35:43 you'll send out 100 and he'll have 110 by the time the fleet gets there 2010-09-25T11:36:04 it's just another way to maintain the stalemate 2010-09-25T11:36:12 sigh: you're right still a draw in that case 2010-09-25T11:36:45 looks like best strategy is to wait, but things could change once you get enough ships on your home planet to expand 2010-09-25T11:37:11 jesionaj: no matter how many ships you have, your opponent will have the same amount 2010-09-25T11:37:15 assuming you both wait 2010-09-25T11:37:35 right 2010-09-25T11:39:29 but at some point, you might be able to expand to enough other planets that losing your homeworld is acceptable 2010-09-25T11:40:04 I don't think you can hold any long enough to make back the ships you'll spend conquering it 2010-09-25T11:40:13 yeah, that's the problem 2010-09-25T11:40:22 jesionaj: maybe, but your opponent can reach any planet you conquer in <= 2 moves 2010-09-25T11:40:40 true, and then they can capture it easily 2010-09-25T11:40:41 chances are they will be able to take it for cheaper 2010-09-25T11:41:01 looks like it is a stalemate 2010-09-25T11:41:39 yeah 2010-09-25T11:42:19 unless your opponent doesn't account for it, then it's a win :) 2010-09-25T11:42:37 hehe 2010-09-25T11:43:22 that's true for all maps 2010-09-25T11:44:07 every game is a stalemate, if both bots are perfect 2010-09-25T11:44:33 I don't think game has perfect play 2010-09-25T11:44:42 *this game 2010-09-25T11:44:56 only on some maps 2010-09-25T11:48:25 yeah, I think there are at least some rock-paper-scissors type situations 2010-09-25T11:48:40 not sure how common they are though 2010-09-25T11:49:21 me neither, but I'm leaning on the side of quite common 2010-09-25T11:57:22 Top 10 players: dmj111(3436), kcm1700(3415), adumlah(3389), McLeopold(3357), EBraun(3347), felixcoto(3303), wagstaff(3297), Tracer(3267), GreenTea(3257), Zilog8(3240) 2010-09-25T12:00:57 *** Naith has joined #aichallenge 2010-09-25T12:08:33 sigh: haven't competed in a few days now 2010-09-25T12:09:08 aw, that's too bad 2010-09-25T12:09:21 minecraft is it atm 2010-09-25T12:09:32 after that, multi homed task based orders 2010-09-25T12:11:20 :) 2010-09-25T12:11:44 *** coventry has joined #aichallenge 2010-09-25T12:13:03 *** jaspervdj has joined #aichallenge 2010-09-25T12:16:54 *** Migi32 has joined #aichallenge 2010-09-25T12:17:27 how do you easily view the run-time errors of your bot? I can't be bothered to type out a map + orders on the command-line 2010-09-25T12:17:58 and the visualizer just gives me "WARNING: player 1 timed out." 2010-09-25T12:18:52 *** nrub has joined #aichallenge 2010-09-25T12:19:05 Migi32: http://ai-contest.com/forum/viewtopic.php?p=3836#p3836 will give you any whole lines that print to stderr. 2010-09-25T12:20:14 thanks 2010-09-25T12:21:04 *** perror has joined #aichallenge 2010-09-25T12:21:11 *** sigh has quit IRC (Remote host closed the connection) 2010-09-25T12:22:17 Migi32: You're welcome. 2010-09-25T12:23:21 *** customKoncept has joined #aichallenge 2010-09-25T12:24:12 Is the main server having load issues again? 2010-09-25T12:25:07 ...lost a game on turn 1 2010-09-25T12:26:11 *** olexs has joined #aichallenge 2010-09-25T12:26:55 this TCP server, is it like the main server where it pits you against a random other player? 2010-09-25T12:28:57 against random available player 2010-09-25T12:29:05 ok 2010-09-25T12:29:07 cool :P 2010-09-25T12:31:25 Migi32: You can also see rankings for that server at http://72.44.46.68/getratings 2010-09-25T12:47:32 *** RainCT has quit IRC (Ping timeout: 245 seconds) 2010-09-25T12:48:33 *** boegel has joined #aichallenge 2010-09-25T12:48:38 yo 2010-09-25T12:48:52 can someone explain what happened on the official server a couple of hours ago? 2010-09-25T12:49:02 *** coventry has quit IRC (Ping timeout: 252 seconds) 2010-09-25T12:49:06 for some reason, the games that were played by my bot were erased (with resubmitting a new version) 2010-09-25T12:49:23 *** RainCT has joined #aichallenge 2010-09-25T12:49:23 *** RainCT has joined #aichallenge 2010-09-25T12:49:24 and now I notice that my bot played against itself, wtf (see http://ai-contest.com/profile.php?user_id=7846) 2010-09-25T12:49:49 lol 2010-09-25T12:50:06 who won? 2010-09-25T12:50:32 other boegel 2010-09-25T12:51:16 [12:34:44] sign: Server had issues this morning 2010-09-25T12:51:16 [12:35:00] sigh: spent from midnight to 6 overloaded 2010-09-25T12:51:24 Naktibalda: doesn't make sense :) 2010-09-25T12:51:30 [12:34:44] sign: Server had issues this morning 2010-09-25T12:51:30 [12:35:00] sigh: spent from midnight to 6 overloaded 2010-09-25T12:51:43 it was 5 hours ago 2010-09-25T12:52:00 I guess I should just resubmit to truly evaluate this version 2010-09-25T12:52:02 sucks though :( 2010-09-25T12:57:21 Top 10 players: dmj111(3448), kcm1700(3424), adumlah(3390), McLeopold(3367), r.nitsch(3366), EBraun(3355), felixcoto(3324), wagstaff(3308), Tracer(3290), GreenTea(3267) 2010-09-25T13:00:57 *** u1tr0n has quit IRC (Ping timeout: 252 seconds) 2010-09-25T13:01:16 *** boiserm has joined #aichallenge 2010-09-25T13:01:49 Is there a possibility that the end-of-game calculations are off? 2010-09-25T13:01:57 the game ends when the other player has no fleets at all, right? 2010-09-25T13:01:59 what the hell 2010-09-25T13:02:10 but I won this game too early it seems... http://ai-contest.com/visualizer.php?game_id=4716358 2010-09-25T13:02:12 even after a new submission, my bot is playing against itself! 2010-09-25T13:02:14 maybe the other bot timed out? 2010-09-25T13:02:19 again, see http://ai-contest.com/profile.php?user_id=7846 2010-09-25T13:02:41 someone introduced a bug in an update of the server, I'd say... :) 2010-09-25T13:02:50 can I report this somewhere 'official'? 2010-09-25T13:02:52 boegel: on the bright side, you won ;-) 2010-09-25T13:02:58 boiserm: just once :) 2010-09-25T13:03:10 I lost and won at the same time :) 2010-09-25T13:03:14 hehe 2010-09-25T13:05:02 People still seem to have bugs when considering 0 planet situations 2010-09-25T13:05:54 *** kozlovsky has joined #aichallenge 2010-09-25T13:06:22 meaning the bot I was playing just wasn't coded right? 2010-09-25T13:06:34 I guess that's probably a lot more likely, yeah 2010-09-25T13:06:50 probably 2010-09-25T13:07:11 at least the game started 2010-09-25T13:07:22 and at least I won :-P 2010-09-25T13:07:23 I think I got stung by an overloaded server 2010-09-25T13:07:24 http://ai-contest.com/visualizer.php?game_id=4740832 2010-09-25T13:07:40 ouch 2010-09-25T13:07:42 draw or loss? 2010-09-25T13:07:44 I lost 2010-09-25T13:07:47 :( 2010-09-25T13:08:24 I know I'm not going to win, but I want a top ten spot, even if its only just long enough to get a screenshot 2010-09-25T13:08:31 yikes, rank 22... double digits are still a dream for me 2010-09-25T13:08:36 meh 2010-09-25T13:08:47 bartwe: minecraft has also stolen me away for a couple days, but i'm making myself do more productive things again now i hope 2010-09-25T13:08:54 try thinking about it differently 2010-09-25T13:09:10 I look at the top few bots & just think WTF 2010-09-25T13:09:22 but even though I still have lots of bugs 2010-09-25T13:09:31 (logic, not code) 2010-09-25T13:09:36 a few things really help 2010-09-25T13:09:50 do you have some decent reinforcement code? 2010-09-25T13:09:51 yeah, I realized I was trying to just code up a bot every evening and upload it to see how it did... no surprise that I'm not very high in the rankings... I need to stop and think a bit. 2010-09-25T13:10:11 It takes about 1/2 day to a day to float 2010-09-25T13:10:15 The next version I'm going to upload does reinforcement much better 2010-09-25T13:10:23 :) 2010-09-25T13:10:26 :-P 2010-09-25T13:10:47 * Naktibalda is going to read Algorithms in a Nutshell. 2010-09-25T13:11:16 boiserm - are you a student? 2010-09-25T13:11:27 I tentatively uploaded an "improved" version, but I think I left in an experimental parameter, and it ended up 2x as bad as the previous version 2010-09-25T13:11:37 customKoncept: not currently 2010-09-25T13:11:55 Graph Algorithms, Pathfinding in AI and Network flow algorithms may be useful for this contest 2010-09-25T13:12:25 just wondering... I'd like the time of being a student again 2010-09-25T13:12:38 customKoncept: *me too* 2010-09-25T13:12:50 this is just too funny to be true: http://ai-contest.com/visualizer.php?game_id=4742099 :) 2010-09-25T13:12:52 gah - how do you whisper? 2010-09-25T13:13:07 actually, I'm surprised either instance of my bot wins, they should draw afaik 2010-09-25T13:13:17 yeah 2010-09-25T13:13:19 hmm... 2010-09-25T13:13:29 I've found small differences between players 2010-09-25T13:13:45 in that OCCASIONALLY 1 player get a different turns to planet rounding 2010-09-25T13:14:01 and... I start my turn with a Collections.shuffle() on my planets 2010-09-25T13:14:17 but not on enemy / neutral planets 2010-09-25T13:14:19 boegel: your bot is deterministic and still somehow beat itself? 2010-09-25T13:14:24 yeah 2010-09-25T13:14:31 might be a sign that there's a bug :) 2010-09-25T13:14:34 customKoncept: what advantage does that get you? 2010-09-25T13:14:35 list/array traversal starts at 0 2010-09-25T13:14:40 no advantage 2010-09-25T13:14:51 its a carry over from a slower bot 2010-09-25T13:14:54 well, I mean... then why do it? 2010-09-25T13:14:58 oh 2010-09-25T13:15:18 evaluating planets usually involves distance calculations 2010-09-25T13:15:18 * boiserm understand 2010-09-25T13:15:21 ...s 2010-09-25T13:15:29 AI is based per planet... kk 2010-09-25T13:22:00 *** u1tr0n has joined #aichallenge 2010-09-25T13:22:16 lol wtf? http://ai-contest.com/visualizer.php?game_id=4742487 2010-09-25T13:22:29 ...? I ported my bot from Java to Python, and the Python version only wins 44% of the time... 2010-09-25T13:22:41 I guess that's within the margin of error 2010-09-25T13:22:44 maybe 2010-09-25T13:22:55 against java version? 2010-09-25T13:23:17 right 2010-09-25T13:23:20 just watch a few games and find differences 2010-09-25T13:23:42 *** jaspervdj has quit IRC (Quit: NEVER GONNA GIVE YOU UP NEVER GONNA LET YOU DOWN) 2010-09-25T13:23:51 yeah, good call 2010-09-25T13:24:11 how it can be that bot fight with himself ? 2010-09-25T13:24:51 server bug 2010-09-25T13:25:21 uploaded a new game 2010-09-25T13:25:24 er a new bot 2010-09-25T13:25:25 lol 2010-09-25T13:25:41 http://ai-contest.com/visualizer.php?game_id=4737205 <- wtf? blue bot drunken? 2010-09-25T13:26:30 :D 2010-09-25T13:26:56 how long for games now 2010-09-25T13:27:00 blue bot is borring to fight normal way ;) 2010-09-25T13:27:44 http://ai-contest.com/visualizer.php?game_id=4743078 first game against deepblue 2010-09-25T13:28:50 *** Queue29 has joined #aichallenge 2010-09-25T13:28:54 not sure why I stock pile 1708 troops and dont use them though 2010-09-25T13:30:08 *** Queue29 has quit IRC (Client Quit) 2010-09-25T13:32:02 steps 76-145 your bot calculate something ? 2010-09-25T13:32:22 watching right now 2010-09-25T13:33:07 no, just an over reinforcement (my bot had more growth then) 2010-09-25T13:33:53 once my bot was ahead in growth it just waits 2010-09-25T13:34:04 which under a strict 200 turn limit is a bad thing 2010-09-25T13:34:27 *** KomarSerjio has joined #aichallenge 2010-09-25T13:34:32 jmcarthur: I don't think it's a bug 2010-09-25T13:34:45 jmcarthur: more like a dependence on the order in which stuff appears 2010-09-25T13:34:57 jmcarthur: for example, fleets or somesuch 2010-09-25T13:35:15 *** KomarSerjio has quit IRC (Client Quit) 2010-09-25T13:35:30 jmcarthur: alhthough, I can't think of a reason why the order of something would be different in different instances 2010-09-25T13:35:31 weird 2010-09-25T13:35:35 anyway 2010-09-25T13:35:52 *** mega1 has quit IRC (Ping timeout: 245 seconds) 2010-09-25T13:35:54 are games against yourself supposed to happen on the official server? 2010-09-25T13:37:13 I guess the good news is that games against yourself shouldn't affect your ELO rating (?) 2010-09-25T13:37:19 boegel: dunno but i found one: http://ai-contest.com/visualizer.php?game_id=4742487 2010-09-25T13:37:20 http://ai-contest.com/visualizer.php?game_id=4743445 new bot uploaded 2010-09-25T13:37:57 *** Bobng has joined #aichallenge 2010-09-25T13:38:03 games against yourself really screw up bayeselo 2010-09-25T13:38:23 a1k0n: I'd guess so too 2010-09-25T13:38:52 i was goofing around with matching algorithms with it and i accidentally allowed that and got some totally crazy results 2010-09-25T13:38:59 and in profile it writen like loss :D http://ai-contest.com/profile.php?user_id=8969 2010-09-25T13:39:32 http://ai-contest.com/profile.php?user_id=9558 2010-09-25T13:39:34 sometimes a win 2010-09-25T13:39:39 u1tr0n: in my case, I see two entries, one loss, one win 2010-09-25T13:39:43 yeah 2010-09-25T13:40:32 u1tr0n: see: http://ai-contest.com/profile.php?user_id=7846 2010-09-25T13:40:57 did you guys get new servers yet? 2010-09-25T13:41:58 *** Itkovian has joined #aichallenge 2010-09-25T13:45:30 *** Itkovian has quit IRC (Client Quit) 2010-09-25T13:49:31 jmcarthur: I think I know what happened... I sort things based on some double value, and the Haskell runtime probably isn't deterministic in that 2010-09-25T13:49:48 jmcarthur: comparing doubles never is a good idea is "equal" values matter 2010-09-25T13:50:10 jmcarthur: I think two things with the same double value are ordered differently in both instances 2010-09-25T13:50:17 jmcarthur: which explain the different behavior 2010-09-25T13:50:22 boegel: it should at least be deterministic... 2010-09-25T13:50:26 it all makes sense \o/ 2010-09-25T13:50:38 it would be really weird for haskell especially... 2010-09-25T13:50:42 jmcarthur: well, no, not necessarily, not with double values 2010-09-25T13:50:49 since pure functions are supposed to be referentially transparent 2010-09-25T13:50:59 jmcarthur: it really depends on how they are represented in memory, rounding errors and whatnot 2010-09-25T13:51:05 no, i can see things like commutativity failing, but not referential transparency 2010-09-25T13:51:26 jmcarthur: well, it's the best thing I can think of 2010-09-25T13:51:36 anyway, afk for a while 2010-09-25T13:51:57 I'll post a new topic on the forum about bots playing against themselves, this should be solved ASAP imho 2010-09-25T13:52:02 i think you can break RT if you are compiling with -fexcess-precision, but the server doesn't do that anyway 2010-09-25T13:57:05 *** u1tr0n has quit IRC (Quit: Page closed) 2010-09-25T13:57:22 Top 10 players: dmj111(3443), adumlah(3406), kcm1700(3390), EBraun(3351), felixcoto(3346), McLeopold(3345), wagstaff(3294), Tracer(3290), r.nitsch(3279), GreenTea(3255) 2010-09-25T13:58:38 *** elijahbal has joined #aichallenge 2010-09-25T13:58:40 hello. 2010-09-25T13:58:49 What's the size of the map ? 20x20 ? 2010-09-25T13:59:59 elijahbal: it varies, you're probably better off measuring 2010-09-25T14:02:10 how ? 2010-09-25T14:03:44 *** jmcarthur has quit IRC (Quit: WeeChat 0.3.0) 2010-09-25T14:05:56 *** kjeldahl has quit IRC (Remote host closed the connection) 2010-09-25T14:06:23 Is 0-1 knapsack problem NP-complete if weights are vectors (of integers)? 2010-09-25T14:06:59 and values real numbers 2010-09-25T14:12:18 *** jmcarthur has joined #aichallenge 2010-09-25T14:14:20 *** mewo2 has quit IRC (Remote host closed the connection) 2010-09-25T14:15:57 *** amstan has joined #aichallenge 2010-09-25T14:15:57 *** ChanServ sets mode: +o amstan 2010-09-25T14:16:54 bogel: I'll take a look shortly 2010-09-25T14:21:16 danielvf: cool, th 2010-09-25T14:21:18 *thx 2010-09-25T14:21:35 danielvf: define shortly? in a couple of minutes/hours/days? :) 2010-09-25T14:31:10 20 min 2010-09-25T14:31:12 or less 2010-09-25T14:31:14 for now 2010-09-25T14:31:16 *from 2010-09-25T14:33:12 *** perror has quit IRC (Quit: Bye all !) 2010-09-25T14:38:59 danielvf: cool! thanks a bunch 2010-09-25T14:42:07 *** Yoshi-TS4 has joined #aichallenge 2010-09-25T14:45:37 boegel: should be fixed. You were playing a previous version of yourself 2010-09-25T14:46:09 danielvf: ah, that explains the win/loss situation 2010-09-25T14:46:12 danielvf: thx! 2010-09-25T14:46:28 Np 2010-09-25T14:46:35 Thanks for letting me know. 2010-09-25T14:47:52 *** fry has joined #aichallenge 2010-09-25T14:52:04 *** Migi32 has quit IRC (Quit: http://boinc.berkeley.edu) 2010-09-25T14:54:33 danielvf: any idea how this was introduced? a pointer to an issue report or somesuch will do (if there's any) 2010-09-25T14:55:01 *** sinan has quit IRC (Quit: WeeChat 0.2.6.3) 2010-09-25T14:55:15 This was on the experimental new cloud game servers 2010-09-25T14:55:39 and it was playing the most recent submissions, and not checking for player uniqueness 2010-09-25T14:56:13 k, thx for the details : 2010-09-25T14:56:14 :) 2010-09-25T14:56:15 The game picking code for the new server is not finalized yet 2010-09-25T14:56:26 danielvf: clearly ;) 2010-09-25T14:56:35 It's in more just to test that the new servers can play games 2010-09-25T14:57:04 *** Itkovian has joined #aichallenge 2010-09-25T14:57:04 adding the new servers really made a difference 2010-09-25T14:57:16 a lot more games are being played now 2010-09-25T14:57:20 which is cool 2010-09-25T14:57:22 Top 10 players: dmj111(3476), kcm1700(3440), adumlah(3400), EBraun(3363), McLeopold(3361), felixcoto(3346), wagstaff(3305), Tracer(3284), GreenTea(3275), Zilog8(3262) 2010-09-25T14:57:30 you get disappointed more quickly about the quality of your bot :P 2010-09-25T14:57:34 boegel: lol 2010-09-25T15:11:50 ouch... definitely should have gotten automated local tournaments done way earlier :-/ 2010-09-25T15:13:17 turns out I haven't been as successful against the sample bots as I thought -_- 2010-09-25T15:15:53 :) 2010-09-25T15:16:03 boiserm: :) 2010-09-25T15:18:14 Are there any plans to capture why a game was terminated early, ie timeout, invalid move, etc? 2010-09-25T15:19:11 I've been playing hundreds of games on through the tcp server and locally, and I never run into any problems.. the two games i have on the live server prematurely terminate with my loss for some unknown reason.. 2010-09-25T15:19:52 anoek: what language are you using? 2010-09-25T15:20:11 anoek: is it possible that there's some reliance on a certain version? 2010-09-25T15:21:06 boiserm: precompiled c++ module through javascript 2010-09-25T15:21:26 so there shouldn't be any reliance on versions, since it's the same binary i test over here 2010-09-25T15:21:41 what's that? 2010-09-25T15:22:00 contest server doesn't accept compiled code 2010-09-25T15:22:14 it's a javascript module 2010-09-25T15:25:19 Anywho, would be really great if it had a little note saying why the game server terminated.. 2010-09-25T15:30:26 *** Strepto has joined #aichallenge 2010-09-25T15:30:43 yeah, i think someone should work on that 2010-09-25T15:31:24 anoek: is it a turn 1 loss, or midgame? 2010-09-25T15:31:51 Zannick: mid game 2010-09-25T15:32:05 Zannick: turn 50, 48, and 43 so far 2010-09-25T15:32:28 could you paste a link to your profile? 2010-09-25T15:33:23 http://ai-contest.com/profile.php?user_id=6859 2010-09-25T15:34:09 hmm.. i do spew out a ton of stuff to stderr, would that cause any problems maybe? 2010-09-25T15:35:46 i shouldn't be timing out or anything like that, i make decisions in ~50 ms or so .. max 2010-09-25T15:36:07 my guess is invalid order 2010-09-25T15:36:22 or not saying "go" 2010-09-25T15:36:55 Hm, i would have thought i would have seen that playing any of the hundreds of tcp based games, or any of the mini tournaments i do locally benching bots against eachother.. 2010-09-25T15:36:56 stderr output shouldn't be an issue, the engine ignores that 2010-09-25T15:37:08 or at least it should ignore it 2010-09-25T15:37:14 actually 2010-09-25T15:37:32 the engine doesn't read stderr; it doesn't go anywhere 2010-09-25T15:37:41 if you're outputting a lot, you might be filling the pipe 2010-09-25T15:37:59 Doh 2010-09-25T15:38:04 and then a write to stderr is waiting for the buffer to clear 2010-09-25T15:38:07 Yeah that'd explain it 2010-09-25T15:38:44 testing that theory now.. 2010-09-25T15:43:01 *** danielvf_ has joined #aichallenge 2010-09-25T15:43:01 *** danielvf has quit IRC (Read error: Connection reset by peer) 2010-09-25T15:43:02 *** danielvf_ is now known as danielvf 2010-09-25T15:43:19 Zannick: Looks like that was it :) 2010-09-25T15:43:59 :) 2010-09-25T15:44:27 I had presumed since the playgame.jar they shipped relays debug information to the console, they'd just pipe that out to /dev/null and be done with it.. bad assumption i guess 2010-09-25T15:46:32 *** rwa has quit IRC (Read error: Operation timed out) 2010-09-25T15:48:39 *** Accoun has quit IRC () 2010-09-25T15:52:17 *** rwa has joined #aichallenge 2010-09-25T15:57:23 Top 10 players: dmj111(3478), kcm1700(3444), adumlah(3392), McLeopold(3382), EBraun(3361), felixcoto(3351), wagstaff(3307), GreenTea(3272), Tracer(3269), r.nitsch(3264) 2010-09-25T16:02:03 *** Itkovian has quit IRC (Quit: Itkovian) 2010-09-25T16:08:29 *** BtbN has quit IRC (Quit: Verlassend) 2010-09-25T16:10:59 *** espes has quit IRC (Read error: Operation timed out) 2010-09-25T16:11:51 *** Accoun has joined #aichallenge 2010-09-25T16:14:13 *** aerique has joined #aichallenge 2010-09-25T16:29:04 *** Itkovian has joined #aichallenge 2010-09-25T16:33:53 :-/ next version stats: 2010-09-25T16:33:58 96/100 against DualBot 2010-09-25T16:34:04 85/100 against RageBot 2010-09-25T16:34:18 hmm 2010-09-25T16:36:36 improvement? 2010-09-25T16:37:09 lol sadly, yes :) 2010-09-25T16:37:23 I guess that was some necessary context, yeah 2010-09-25T16:38:24 then that's all that matters for now 2010-09-25T16:40:09 yay my bot doesn't segfault now 2010-09-25T16:41:25 next up: turn 1 2010-09-25T16:41:48 *** jaspervdj has joined #aichallenge 2010-09-25T16:47:58 hmm, I'm seeing 3 first-turn failed games in a row with my bot, while the first 2 games were OK 2010-09-25T16:48:06 can someone look into that? 2010-09-25T16:48:15 example of such a game: http://ai-contest.com/visualizer.php?game_id=4749082 2010-09-25T16:48:50 the same bot has played 2 earlier games just fine, so there's no reason why it should crash in the first turn (it would have with the other games to then) 2010-09-25T16:49:49 *** espes has joined #aichallenge 2010-09-25T16:51:09 *** jesionaj` has joined #aichallenge 2010-09-25T16:52:09 *** delt0r has joined #aichallenge 2010-09-25T16:55:05 *** jesionaj has quit IRC (Ping timeout: 272 seconds) 2010-09-25T16:57:23 Top 10 players: dmj111(3509), kcm1700(3428), adumlah(3377), EBraun(3363), McLeopold(3354), felixcoto(3340), wagstaff(3297), Tracer(3256), r.nitsch(3243), GreenTea(3226) 2010-09-25T16:59:01 danielvf: don't know if you have time, but could you look at http://ai-contest.com/visualizer.php?game_id=4749082 and also the games with ids 4749063 and 4749264? 2010-09-25T16:59:33 danielvf: I lost those in the first turn, while games 4747094 and 4743812 went just fine 2010-09-25T16:59:39 Okay 2010-09-25T16:59:57 danielvf: I can't see why my bot would not crash in the first two, but would crash (in the first turn) in the other 3 2010-09-25T17:00:24 doesn't make sense at all 2010-09-25T17:00:34 unless the server is causing time-outs or somein' 2010-09-25T17:00:51 It's likely not your fault. 2010-09-25T17:01:54 I'd guess so too 2010-09-25T17:01:58 just want to make sure 2010-09-25T17:02:21 and check whether this is a server problem similar to the boegel vs boegel stuff I was seeing earlier :) 2010-09-25T17:02:32 After we take care of more games, I'm going to be checking into the timeout issues. 2010-09-25T17:02:35 just want to help out, no critisism 2010-09-25T17:02:41 boegel: yeah, np 2010-09-25T17:05:29 * boegel resubmits to get rid of the time-outs 2010-09-25T17:05:39 I really want to figure out how high this bot can go :) 2010-09-25T17:05:53 I have a feeling I'm close to top 100 :) 2010-09-25T17:06:01 :D 2010-09-25T17:06:30 You haskellers are doing well 2010-09-25T17:06:45 * boegel bows 2010-09-25T17:06:56 Haskell is superb for doing stuff like this 2010-09-25T17:07:10 efficiency isn't too much of an issue (with my tactics at least) 2010-09-25T17:07:26 you're constantly working with fairly small lists of things (planets, fleets, ...) 2010-09-25T17:07:47 it's really easy to make a slight change to your code to adjust your strategy, ... 2010-09-25T17:07:51 I'm loving it 2010-09-25T17:08:01 *nods* 2010-09-25T17:08:08 I wanted to get back to Haskell with something fun, and I've definitely found it 2010-09-25T17:08:22 a buddy of mine is learning Haskell and trying to convince me it's the best thing since sliced bread 2010-09-25T17:08:42 I've started reading a few tutorials, and I'll admit it's pretty cool 2010-09-25T17:08:52 Heh 2010-09-25T17:09:13 but I don't know it well enough to do a bot (yet) and haven't had the right project idea to motivate me to learn it 2010-09-25T17:09:20 who knows, got a couple of months left 2010-09-25T17:09:48 does this count as a strategy, split up fleets as much as possible to get the other bot to time-out? :D http://ai-contest.com/visualizer.php?game_id=4731923 2010-09-25T17:10:22 boegel: lol 2010-09-25T17:10:33 this question should be in FAQ :) 2010-09-25T17:10:34 boiserm: there's a great starters package for Haskell 2010-09-25T17:10:46 boiserm: check it out, you'll get started in no time 2010-09-25T17:10:52 boiserm: also, check out learnyouahaskell.com 2010-09-25T17:11:09 boiserm: best tutorial *ever* (and not just Haskell) 2010-09-25T17:11:16 boegel: it won't help you to get into top 100 2010-09-25T17:11:27 Naktibalda: I'm not considering it 2010-09-25T17:11:45 Naktibalda: just wondering if this is allowed 2010-09-25T17:11:55 it is 2010-09-25T17:11:55 Naktibalda: I can imagine my bot would currently trip over it 2010-09-25T17:12:13 Naktibalda: but that's easily remedied by merging fleets prior to analysis 2010-09-25T17:22:45 *** lotek93 has joined #aichallenge 2010-09-25T17:23:04 hello 2010-09-25T17:24:24 *** Zeiris has joined #aichallenge 2010-09-25T17:25:28 can anybody explain why did bot lose? 2010-09-25T17:28:43 lotek93: which game? 2010-09-25T17:29:15 *** jaspervdj has quit IRC (Quit: NEVER GONNA GIVE YOU UP NEVER GONNA LET YOU DOWN) 2010-09-25T17:29:58 boegel: http://ai-contest.com/visualizer.php?game_id=4750592 2010-09-25T17:30:44 lotek93: two main possibilities 2010-09-25T17:31:15 lotek93: either your bot produced an invalid order (e.g. send more ships in total from planet P than it has) 2010-09-25T17:31:36 lotek93: or it failed to produce all orders within 1 second, i.e. a time-out 2010-09-25T17:31:56 lotek93: noticing the large number of fleets in play, I'd say the last option, i.e. a time-out 2010-09-25T17:32:22 boegel: and there is no error diagnostic from engine? 2010-09-25T17:32:56 lotek93: not yet, it's being worked on 2010-09-25T17:33:10 lotek93: there is a trick though, if you have debugging in your bot locally 2010-09-25T17:33:55 boegel: locally i haven't any problem 2010-09-25T17:34:37 lotek93: there's a way in which you can replay a game played on the server, to help figure out what might have gone wrong 2010-09-25T17:34:45 lotek93: see http://ai-contest.com/forum/viewtopic.php?f=17&t=628&start=0 2010-09-25T17:35:13 lotek93: if your bot sent an invalid order, you should be able to see it with the replay 2010-09-25T17:35:32 lotek93: read the entire topic, the last script is a lot easier to use 2010-09-25T17:36:06 thanks boegel! 2010-09-25T17:39:22 its very strange anyway.. 2010-09-25T17:39:58 lotek93: it *may* be due to a busy server too, they have seen problems like this in the past 2010-09-25T17:40:16 lotek93: but I'd really check if you handle lots of fleets efficiently 2010-09-25T17:42:18 boegel: btw how they can guarantee exactly 1 second for make orders? 2010-09-25T17:43:06 boegel: all my planets send only half of ships, not more 2010-09-25T17:46:42 *** sugarshark has quit IRC (Quit: ERC Version 5.3 (IRC client for Emacs)) 2010-09-25T17:47:17 lotek93: your bot is only giving 1 second by the server to create its orders and give a 'go' to actually execute them 2010-09-25T17:47:33 lotek93: if it takes longer, the server consideres it a time-out, and you lose 2010-09-25T17:47:51 lotek93: so, either you very sure you're way below that 1s limit, or you check yourself in your bot 2010-09-25T17:48:20 lotek93: I'd guess your code processing all the fleets that are in play is too slow, leading to a time-out 2010-09-25T17:50:39 *** aerique has quit IRC (Quit: ...) 2010-09-25T17:51:59 boegel: yes, i know, i mean - if there is many games simultaneously how can i be sure that server performance is enough for my bot? 2010-09-25T17:52:21 lotek93: well, I guess that's one of the problems they're having now 2010-09-25T17:52:38 lotek93: I'm guessing they only run one bot at a time or something? 2010-09-25T17:52:53 that's not very efficient though :) 2010-09-25T17:53:04 danielvf: how do you guys handle this stuff? 2010-09-25T17:53:06 dunno :) 2010-09-25T17:56:40 *** Bobng has quit IRC (Ping timeout: 276 seconds) 2010-09-25T17:56:57 boegel: they can run a bot per core. There's 200 turn max, so 203s max per game. I expect they have a lot of machines, no? 2010-09-25T17:57:25 Top 10 players: dmj111(3503), kcm1700(3416), adumlah(3414), EBraun(3382), McLeopold(3329), felixcoto(3316), wagstaff(3307), bix0r4ever(3286), Tracer(3261), r.nitsch(3259) 2010-09-25T17:57:33 Itkovian: 400s per game, there's two sides 2010-09-25T17:57:41 no, 1 core per player 2010-09-25T17:57:44 Itkovian: also, what about shared caches? 2010-09-25T17:57:49 true also 2010-09-25T17:57:51 Itkovian: ah, right 2010-09-25T17:58:05 1 core per pair mb? 2010-09-25T17:58:16 Itkovian: shared caches might have a *big* influence on time, especially on a time-scale of < 1s 2010-09-25T17:58:22 I think caches are no issue, unless you give the thing a lot of code, and you are allowed to upload 2MB zips? 2010-09-25T17:58:34 so yeah, in the end it might be a problem 2010-09-25T17:58:39 makes me wonder 2010-09-25T17:58:54 would you allowed to spawn threads to offload computation? 2010-09-25T17:59:15 Itkovian: what about the data this thing produces :) 2010-09-25T17:59:30 what do they keep? 2010-09-25T17:59:35 Itkovian: no, only single-threaded stuff imho 2010-09-25T17:59:41 Itkovian: depends on the bot, I guess :) 2010-09-25T17:59:42 they need not keep state, just order afaik 2010-09-25T17:59:57 Itkovian: some bots keep state across turns 2010-09-25T18:00:22 true, but I was referring to the logs/data 2010-09-25T18:00:33 or did you mean the data per bot in-game? 2010-09-25T18:02:29 Itkovian: of course 2010-09-25T18:05:38 ...was comparing fleet owners to planet IDs instead of to planet owners -_- 2010-09-25T18:05:52 ...fixed the bug, and the bot did *worse.* Make of that what you will. 2010-09-25T18:06:12 change it back then 2010-09-25T18:07:48 tests are still running... did worse against DualBot, but much better against RageBot 2010-09-25T18:07:57 eh, whatever. I was amused 2010-09-25T18:09:21 *** u1tr0n has joined #aichallenge 2010-09-25T18:10:42 boiserm: try and learn from that 2010-09-25T18:10:54 *** boiserm is now known as boiserm-AFK 2010-09-25T18:11:06 comparing to planetId means that there were less matches, since pids have a much bigger range 2010-09-25T18:11:21 less matches means (fill in) 2010-09-25T18:11:29 *** blbrown_win3_ has joined #aichallenge 2010-09-25T18:12:38 i win lol http://ai-contest.com/visualizer.php?game_id=4753601 2010-09-25T18:13:56 u1tr0n: erm, no, you lose :) 2010-09-25T18:15:13 *** blbrown_win3 has quit IRC (Ping timeout: 265 seconds) 2010-09-25T18:25:03 he won 2010-09-25T18:25:28 probably opponent issued an invalid order 2010-09-25T18:26:31 Tidan2 uploaded 6 times in 62 minutes :) 2010-09-25T18:27:12 Naktibalda: ah :D 2010-09-25T18:27:42 *** jimmerton has joined #aichallenge 2010-09-25T18:28:10 *** Appleman1234 has joined #aichallenge 2010-09-25T18:29:17 huh, games disappeared again 2010-09-25T18:29:24 is the server in trouble again? 2010-09-25T18:29:30 or did some reset occur? 2010-09-25T18:29:36 *** scottj has joined #aichallenge 2010-09-25T18:30:22 *** jimmerton has quit IRC (Read error: Connection reset by peer) 2010-09-25T18:31:22 what games? 2010-09-25T18:31:53 Naktibalda: my last submission played a game 2010-09-25T18:32:04 Naktibalda: and now it disappear from the list of played games 2010-09-25T18:32:52 ai-contest.com is having trouble on my end... 2010-09-25T18:33:12 boegel: with tron, that's what happened when the opponent resubmitted 2010-09-25T18:34:17 cmatei: I haven't seen games disappear in here 2010-09-25T18:34:24 cmatei: but you might be right actually 2010-09-25T18:34:37 *** jimmerton_ has joined #aichallenge 2010-09-25T18:34:38 cmatei: although... 2010-09-25T18:35:30 I'm having trouble viewing my own profile now :( 2010-09-25T18:35:42 *** hellman has quit IRC (Ping timeout: 240 seconds) 2010-09-25T18:35:48 anyone else able to view http://ai-contest.com/profile.php?user_id=7846 ? 2010-09-25T18:36:51 cmatei: I can imagine that game is no longer counter for the ELO/ranking 2010-09-25T18:37:03 cmatei: but it should disappear from the list, right? 2010-09-25T18:37:11 all the maps are symmetric right? 2010-09-25T18:37:12 *** Strepto has quit IRC (Ping timeout: 252 seconds) 2010-09-25T18:37:44 boegel: i really don't know the details. some pages may be cached and take a while to update, maybe ? 2010-09-25T18:38:34 *** blbrown_win3_ has quit IRC (Quit: Leaving) 2010-09-25T18:38:34 or some results, like rankings. i can see your profile just fine 2010-09-25T18:38:45 cmatei: I can too now 2010-09-25T18:38:51 *** RobotCaleb has quit IRC (Ping timeout: 252 seconds) 2010-09-25T18:39:50 *** RobotCaleb has joined #aichallenge 2010-09-25T18:40:10 *** zz_SophieB is now known as SophieB 2010-09-25T18:40:47 *** star_fire has joined #aichallenge 2010-09-25T18:41:02 hey guys 2010-09-25T18:41:07 I need a PHP starter pack 2010-09-25T18:41:12 but I can't seem to find it 2010-09-25T18:41:28 *** RainCT has quit IRC (Remote host closed the connection) 2010-09-25T18:41:47 any help would be deeply appreciated? 2010-09-25T18:41:57 star_fire: check the forums? 2010-09-25T18:42:33 maybe it's just me, idk 2010-09-25T18:42:33 but I couldn't find it.. 2010-09-25T18:42:33 I did 2010-09-25T18:42:44 boegel: i have the same problem today 2010-09-25T18:43:04 *** blbrown_win3 has joined #aichallenge 2010-09-25T18:43:08 boegel: about disappering games 2010-09-25T18:44:21 star_fire: http://code.google.com/p/ai-contest/issues/detail?id=104 2010-09-25T18:44:46 Oh thank you kind sir..! 2010-09-25T18:44:55 ^^ 2010-09-25T18:46:44 *** kozlovsky has quit IRC (Ping timeout: 252 seconds) 2010-09-25T18:47:17 boegel: yep tecnical im loose, but in my profile im win http://ai-contest.com/profile.php?user_id=9397 im still have some ships 2010-09-25T18:47:22 danielvf: any idea what's going on with disappearing games? 2010-09-25T18:47:43 u1tr0n: yeah, the other guy must have crashed :D 2010-09-25T18:47:50 u1tr0n: funny as hell :D 2010-09-25T18:48:15 boegel: yep 2010-09-25T18:50:02 boegel: im have good idea for bot but it didnt work and i dunnow why and i looking today some way to debug php bot 2010-09-25T18:50:02 *** customKoncept has quit IRC (Ping timeout: 252 seconds) 2010-09-25T18:50:17 boegel: need no make my own server :( 2010-09-25T18:51:43 have tried using tcp? 2010-09-25T18:51:49 tcp rocks 2010-09-25T18:52:10 except for the time-outs, but that's probably my current network talking 2010-09-25T18:54:48 dunno how to connect php bot to tcp 2010-09-25T18:55:47 u1tr0n: your bot runs locally 2010-09-25T18:56:30 see http://72.44.46.68/ 2010-09-25T18:57:00 *** Naith has quit IRC (Quit: Naith) 2010-09-25T18:57:06 u1tr0n: you'll probably need to replace "./MyBot" with "php MyBot.php" or something similar 2010-09-25T18:57:26 Top 10 players: dmj111(3515), adumlah(3420), kcm1700(3410), EBraun(3378), bix0r4ever(3360), wagstaff(3315), felixcoto(3314), McLeopold(3312), Loks1982(3256), Beanhome(3244) 2010-09-25T18:57:45 boegel: i will try tomorrow 2010-09-25T18:57:56 u1tr0n: k 2010-09-25T18:58:00 I'm off to bed too 2010-09-25T18:58:07 gn 2010-09-25T18:58:07 ttyl y'all 2010-09-25T18:58:12 *** boegel has quit IRC (Quit: Leaving) 2010-09-25T18:58:19 *** u1tr0n has quit IRC (Quit: Page closed) 2010-09-25T18:58:57 *** JCS^ has joined #aichallenge 2010-09-25T19:02:58 *** jimmerton_ has quit IRC (Ping timeout: 245 seconds) 2010-09-25T19:07:22 so has anyone downloaded the replay data from the top bots to build their bot from? 2010-09-25T19:09:16 *** alexcg has quit IRC (Quit: Leaving.) 2010-09-25T19:15:22 *** Zaphus has joined #aichallenge 2010-09-25T19:16:06 danielvf: you here ? 2010-09-25T19:26:42 *** star_fire has quit IRC (Ping timeout: 252 seconds) 2010-09-25T19:33:18 scottj: hmm, that's an interesting idea :-) 2010-09-25T19:33:23 but no. 2010-09-25T19:35:38 *** rogue780_ has joined #aichallenge 2010-09-25T19:48:07 *** migi_ has joined #aichallenge 2010-09-25T19:51:11 *** sigh has joined #aichallenge 2010-09-25T19:52:30 sigh: morning 2010-09-25T19:52:38 morning 2010-09-25T19:53:04 my bot is at 16 now, good chance it will make the top 10, I think 2010-09-25T19:53:52 especially if my thoughts here are correct: http://www.ai-contest.com/forum/viewtopic.php?f=18&t=566&p=4854#p4854 the non-compiled bots are getting an artificial boost by playing amongst themselves more 2010-09-25T19:54:10 *** SophieB is now known as CMI-DOS 2010-09-25T19:55:56 ah i see 2010-09-25T19:56:09 I'm not certain, but it's plausible 2010-09-25T19:56:27 @rankings 2010-09-25T19:56:28 sigh: Top 10 players: dmj111(3517), adumlah(3430), kcm1700(3405), bix0r4ever(3398), EBraun(3345), wagstaff(3343), felixcoto(3325), McLeopold(3286), sighbot(3255), Beanhome(3253) 2010-09-25T19:56:30 *** CMI-DOS is now known as cad 2010-09-25T19:56:36 there we go 2010-09-25T19:56:57 nice work ! 2010-09-25T19:57:01 :) 2010-09-25T19:57:26 Top 10 players: dmj111(3519), adumlah(3434), kcm1700(3408), bix0r4ever(3402), EBraun(3348), wagstaff(3345), felixcoto(3327), McLeopold(3289), davidjliu(3260), r.nitsch(3259) 2010-09-25T19:57:28 I havent touched my bot (not a euphamism) - too much footy to watch, plus Civ V :-) 2010-09-25T19:58:03 oh, thats cruel, the contestbot just updated without you in there again... 2010-09-25T19:59:34 haha 2010-09-25T19:59:35 *** rogue780_ has quit IRC (Read error: Connection reset by peer) 2010-09-25T20:00:49 5 python bots in the top 10 now 2010-09-25T20:01:15 only 2 C++ bots... feels wrong given the previous few weeks results 2010-09-25T20:03:01 oh well, it will work itself out 2010-09-25T20:03:27 yeah, it will 2010-09-25T20:03:56 adumlah's strategy is really effective 2010-09-25T20:04:04 I'll have to figure out a way to steal it 2010-09-25T20:04:07 I'm just making up excuses for why my bot that was in the top 10 a few days ago has now dropped to 44th :-) 2010-09-25T20:04:12 hehe :) 2010-09-25T20:05:28 adumlah's looks easy to emulate at a glance, might be worth building one to play against - I cant imagine its always strong 2010-09-25T20:06:04 a lot of strategies *look* easy to emulate :P 2010-09-25T20:07:05 yeah, actually looking at more of his games it isnt as obvious 2010-09-25T20:09:48 wow - his two most recent losses were illegal/timeouts as well, not an actual loss 2010-09-25T20:11:14 *** Titankiller has joined #aichallenge 2010-09-25T20:11:33 I've taken some wins off him on tcp, but I'm not sure if that's his latest bot 2010-09-25T20:11:42 it uses the same strategy though 2010-09-25T20:14:17 I wish they could have archived the top 10 bots from each week so we could see the change in strategies as the weeks progress, it's been interesting to see it get refined 2010-09-25T20:15:06 that won't be as useful until new submission don't put you back down to ~1000 2010-09-25T20:16:13 yeah, that annoys me :-) these arent new bots, they are incremental changes 2010-09-25T20:20:58 it's not so bad if you get a game every hour, but when it's 4 hours between games it can take 2-3 days to get back up to the top 50 2010-09-25T20:25:28 haha bartwe is in #minecraft too 2010-09-25T20:28:34 bartwe said last night that he wasn't working on his bot because of minecraft 2010-09-25T20:28:49 i figured that might be why he fell behind. same here 2010-09-25T20:28:55 haha 2010-09-25T20:29:12 but i did do some work last night. made an exhaustive alpha-beta search on limited instances of the game. 2010-09-25T20:29:27 turns out, knowing the move your opponent made before you make yours gives you a huuuuuge advantage 2010-09-25T20:29:30 did that pan out well? 2010-09-25T20:29:53 well no, basically for the 'minimax incorrect' reasons from the tron contest 2010-09-25T20:29:57 a1k0n: well der :P 2010-09-25T20:30:27 "'minimax incorrect' reasons from the tron contest"? 2010-09-25T20:30:33 i tried a very naive implementation at picking coordinated moves rather than individual ones, but way too slow, so going to need to chop that down a lot 2010-09-25T20:30:41 minimax basically assumes you're taking turns making moves 2010-09-25T20:30:41 I thouoght you guys used minimax in the tron contest? 2010-09-25T20:30:43 but i'm gonna need to stop working on it and do some uni work today 2010-09-25T20:30:47 a1k0n: yes, I know 2010-09-25T20:30:52 yeah. 2010-09-25T20:31:04 but in tron, it didn't matter so much, so there was a whole thread on the forum about how minimax shouldn't work but it does 2010-09-25T20:31:18 because the instances of tron maps in which there were mutually exlusive outcomes were rare 2010-09-25T20:31:46 in tron did you assume that you knew your opponent's move, or that he knew his? 2010-09-25T20:31:56 that your opponent knew yours 2010-09-25T20:32:09 so moves were slightly pessimized 2010-09-25T20:32:18 yup, figured... just checking 2010-09-25T20:32:29 I've never applied minimax to simultaneous games 2010-09-25T20:32:29 i just built the tree as thought it was sequential move, with them as player 2, asusming they would best counter our moves, then said what's the best minimum we can ensure, and picked that 2010-09-25T20:32:31 here, if you pessimize your moves, you just don't expand fast enough 2010-09-25T20:32:43 youwillfail: like a two-step lookahead? 2010-09-25T20:32:48 a1k0n: I agree 2010-09-25T20:32:55 i implemented something along those lines, also coordinating attacks, also way too slow 2010-09-25T20:33:01 and even when i ignored the timeout, it was completely ineffective 2010-09-25T20:33:15 a1k0n: it would generate for as long as it could, then afert each turn, throw out the part of the tree that wasn't followed, and continue calculating 2010-09-25T20:33:23 *** olexs has left #aichallenge 2010-09-25T20:33:36 I skimmed through my AI book for techniques for dealing with games with simultaneous books, but didn't find anything :( 2010-09-25T20:33:39 youwillfail: does it work? 2010-09-25T20:33:56 sigh: maybe we need to write a book after this :-) 2010-09-25T20:33:58 yeah i was hoping to find some axioms of the system or something with this minimax on limited instances 2010-09-25T20:33:58 it would also pick like a "draw with space left" over a "certain draw" to see if we can fill space better etc. 2010-09-25T20:34:02 Zaphus: hshs 2010-09-25T20:34:04 *haha 2010-09-25T20:34:16 youwillfail: oh you mean for tron, right 2010-09-25T20:34:26 yeah that was for tron 2010-09-25T20:34:33 thought you meant for planetwars 2010-09-25T20:34:38 if you want game theory books, look at myersons or fudenbergs 2010-09-25T20:34:54 i'm wondering whether there's any value in modelling this as a POMDP 2010-09-25T20:34:54 nah, way too hard to search 2010-09-25T20:35:05 although even looking one or two moves ahead with that might work? 2010-09-25T20:35:12 if you could get a payoff function working well 2010-09-25T20:35:27 yeah. well i tried it with two moves, and it sucks 2010-09-25T20:35:37 but my implementation may be buggy as hell and probably is 2010-09-25T20:35:49 actually searchin gall possible moves is really hard 2010-09-25T20:36:22 yep, i tried a different approach (sorry, we're keeping that secret) which tries to look across all possible moves and pick the best 2010-09-25T20:36:26 hell enumerating all possible moves is quite expensive :P 2010-09-25T20:36:32 was taking over like 30 seconds to make some moves when i only owned 2 planets 2010-09-25T20:36:34 indeed 2010-09-25T20:36:51 i would imagine for each src-dest pair you could solve a linear program though 2010-09-25T20:37:10 yeah, have considered that, but becomes messy when you have fleets of over this or 0 2010-09-25T20:37:12 all possible moves for the first turn would be huge, 22 destination planets, 1-100 ships each :-) 2010-09-25T20:37:14 so you need either or constraint binding 2010-09-25T20:37:16 yeah, have to make some approximations even then... as it's not really linear 2010-09-25T20:37:24 yeah, so it becomes integer programming, which is messy 2010-09-25T20:37:54 Zaphus: all possible moves once you've captured a few planets is just insane 2010-09-25T20:38:45 yep, basically number of possible moves to a single planet is the product of the number of ships at each planet you own (plus 1 to each) 2010-09-25T20:38:47 haha, even on this first move you are enumerating all integer partitions of 100 :P 2010-09-25T20:38:53 *the first 2010-09-25T20:38:55 yep 2010-09-25T20:38:57 although i think you could certianly reduce that with a bit of dark magic 2010-09-25T20:39:19 * sigh reaches for his eye o newt 2010-09-25T20:39:47 i'm really curious how your bot works, sigh, as i'm basically out of ideas. :) 2010-09-25T20:39:54 I'd be interested to see if anyone does any predicting of the oppositions move, whether they ever get it correct ! 2010-09-25T20:40:13 Zaphus: i'm pretty sure it's going to be needed after a certain point 2010-09-25T20:40:17 once people get it working well 2010-09-25T20:40:26 i've seen people doing it already 2010-09-25T20:40:28 a1k0n: score each destination on the best way to attack it, then attack the destinations with the best scores 2010-09-25T20:40:35 that's basically it 2010-09-25T20:40:44 oh. well that's bascially all i'm doing. that's reassuring. 2010-09-25T20:40:46 no lookahead? 2010-09-25T20:40:51 that is, no game tree searches 2010-09-25T20:40:55 none 2010-09-25T20:41:05 my bots run in less than 1ms 2010-09-25T20:41:10 per move 2010-09-25T20:41:16 i figure game trees will be useful, but not unless you iron out all the heuristic stuff first 2010-09-25T20:41:23 ok, then i'm going to explore reinforcement learning if i can come up with a value function representation 2010-09-25T20:41:36 sigh: impressive 2010-09-25T20:41:48 sigh: Imagine what you could do with a few more ms :-) 2010-09-25T20:42:00 i still wish i knew why a1k0n-0.2 is better than a1k0n-0.4.1, when the latter wipes the floor with the former 2010-09-25T20:42:28 because you build it to improve against your own strategy, but then it wasn't necessarily as strong? 2010-09-25T20:42:32 my current bot is like that 2010-09-25T20:42:34 Zaphus: extra ms might help my bot but it makes my testing cycles much longer 2010-09-25T20:42:43 although it seems to have slipped to a more reasonable position overnight 2010-09-25T20:42:48 so it might be detrimental in the short term 2010-09-25T20:43:01 youwillfail: well, both bots work fundamentally the same way, -0.4.1 is just more accurate in its calculations 2010-09-25T20:43:09 apparently 0.2's approximations cover more of reality 2010-09-25T20:43:19 is it not as aggressive though? 2010-09-25T20:43:26 it's more aggressive actually 2010-09-25T20:43:32 it's just more naïve about defense 2010-09-25T20:43:40 i made mine smarter with moves, but it lost the aggressiveness that won it some games, but stopped losing so many games to lower ranked players 2010-09-25T20:43:58 i've taken defense on as pretty much the more important part 2010-09-25T20:44:05 a1k0n, is your 0.2 doing the same as me and ignoring opponent's captures when it comes to defense? 2010-09-25T20:44:22 i find my bot often gets stuck trying to take an enemies planet when they can easily back it up 2010-09-25T20:44:25 ignoring opponent captures? 2010-09-25T20:44:26 migi_: it's counting incoming ships from both parties for each planet 2010-09-25T20:44:33 then it gets carried away while the other bot expands then goes in front and claims the win 2010-09-25T20:44:36 so it will try to defend if there is an attack 2010-09-25T20:44:39 yes that's what I'm doing too, a1kon 2010-09-25T20:44:46 youwillfail: yes me too 2010-09-25T20:44:53 youwillfail: I've got plans to mitigate that 2010-09-25T20:44:57 yep 2010-09-25T20:45:06 but I'm ignoring the possibility of the opponent first capturing another planet, then attacking with those ships 2010-09-25T20:45:24 and I'm considering if I should make those calculations or if they won't be worth it 2010-09-25T20:45:45 i have trouble trying to do that 2010-09-25T20:46:06 because then other parts using closest enemy used that 2010-09-25T20:46:07 hm, i could acount for that but you'd have to maintain a 2*universe diameter horizon on all future lookahead 2010-09-25T20:46:37 a1k0n: i've found generating such data is pretty much free with time 2010-09-25T20:46:48 yeah it's pretty low-complexity 2010-09-25T20:47:34 I *want* to do things based on my predictions of what the opponent will do 2010-09-25T20:47:41 but I don't know *how* 2010-09-25T20:47:47 it would be cool to have an ai comp with hundreds/thousands of planets and like 0.1 seconds for each move 2010-09-25T20:49:13 sigh, neural networks are pretty good at predicting stuff 2010-09-25T20:49:33 though I'm not sure they can be used here 2010-09-25T20:49:36 * a1k0n scoffs 2010-09-25T20:49:56 i have tried to find information on neural networks but it's all handwavy crap 2010-09-25T20:49:57 * sigh looks for his eye of newt again 2010-09-25T20:50:07 is there somewhere I can actually find the maths defining everything? 2010-09-25T20:50:28 got any ai textbooks? 2010-09-25T20:50:35 nope :P 2010-09-25T20:50:39 to predict stuff, you write down tentative equations of the probability that such-and-such happens given so-and-so, and then turn it around with bayes' theorem and measure so-and-so to figure out such-and-such 2010-09-25T20:51:01 we're intending to do something very much like that anyway 2010-09-25T20:51:09 i should really look into neural networks more 2010-09-25T20:51:15 but haven't even really started on that yet 2010-09-25T20:51:18 but to predict non-trivial bots with that, you'd have to write a million such equations I think 2010-09-25T20:51:44 not really. you just need to be clever about your representation. 2010-09-25T20:51:58 another possible strategy idea i had for this was the possibility of trying to set up a numerical algorithm for approximating a nash equilibria, then playing what that gets to in time 2010-09-25T20:52:08 but i'm not sure that'd work very well 2010-09-25T20:52:13 just saying 'neural network' doesn't solve anything: what are your inputs and what are you outputs? what is it really saying if the neural network outputs X? 2010-09-25T20:52:33 if you have those inputs and outputs you might just be able to write down something more straightforward. 2010-09-25T20:52:50 no, but one would ideally like to have a probability distribution over moves the enemy might make ;) 2010-09-25T20:53:06 youwillfail: hah, good luck with that 2010-09-25T20:53:25 indeed. but if it's parameterized by a funciton that looks at each planet, and the nearest enemy/friendly planets to that, it might be easily possible 2010-09-25T20:53:44 even a *really bad* approximation can get you somewhere 2010-09-25T20:54:24 a1k0n: I think that kind of what I was looking at doing 2010-09-25T20:54:32 like, have your bot be tunable by like three parameters, and then watch your opponents' moves and tune those three parameters to them 2010-09-25T20:54:43 it might be complete crap, but it might work 2010-09-25T20:54:49 "this bot seems partly defensive with a chance of rage" 2010-09-25T20:55:01 you could always try a tit for tat approach, but i think that'd be pretty crappy 2010-09-25T20:55:06 someone made a copybot 2010-09-25T20:55:09 it was pretty funny, and always lost 2010-09-25T20:55:11 Zannick: 2010-09-25T20:55:20 yes? 2010-09-25T20:55:20 Zannick made the copybot 2010-09-25T20:55:21 i want to see an optimal response bot 2010-09-25T20:55:30 so always assumes it has second mover advantage 2010-09-25T20:55:31 that's why i wrote the alpha-beta search 2010-09-25T20:55:32 and see how that goes 2010-09-25T20:55:38 sorry, Zannick was just responding to a1k0n 2010-09-25T20:55:53 it pretty much only won if the other player tried to take central planet and it was any good 2010-09-25T20:56:00 has anyone tried to turn it into a game of nim? 2010-09-25T20:56:06 man, me puncutation no; is good 2010-09-25T20:56:28 sigh: I am terrible at english, I just try to be unambiguous 2010-09-25T20:56:50 i didn't even notice you weren't a native speaker 2010-09-25T20:56:58 i am 2010-09-25T20:57:04 still terrible 2010-09-25T20:57:07 oh, heh. right, punctuation. 2010-09-25T20:57:11 i am antimatroid if you hadn't realised that 2010-09-25T20:57:21 i hadn't. right, from new zealand. 2010-09-25T20:57:24 :P 2010-09-25T20:57:26 Top 10 players: dmj111(3527), bix0r4ever(3422), adumlah(3401), EBraun(3375), kcm1700(3375), felixcoto(3352), wagstaff(3311), sighbot(3292), McLeopold(3286), davidjliu(3265) 2010-09-25T20:57:37 #8 2010-09-25T20:57:38 sigh: You have no gotten any error messages recently, so here's a random one just to let you know that we care. 2010-09-25T20:57:50 haha 2010-09-25T20:58:01 contestbot speaks scottish english 2010-09-25T20:58:02 a1k0n: I'm sorry Dave, I cannot speaks. 2010-09-25T20:58:11 clearly. 2010-09-25T20:58:21 haha 2010-09-25T20:58:29 i feel like this is poor design, to have a bot respond to everything starting with a command character 2010-09-25T20:58:42 contestbot function 2010-09-25T20:58:43 Zannick: You have no gotten any error messages recently, so here's a random one just to let you know that we care. 2010-09-25T20:58:50 amstan set it up so that #!@ were all command characters 2010-09-25T20:58:50 :P 2010-09-25T20:58:52 contestbot function 2010-09-25T20:58:53 Zannick: You have no gotten any error messages recently, so here's a random one just to let you know that we care. 2010-09-25T20:59:01 oh well 2010-09-25T20:59:07 yes 2010-09-25T20:59:18 my favorite function is still: 2010-09-25T20:59:24 contestbot: sigh is awesome 2010-09-25T20:59:24 i know... 2010-09-25T20:59:25 yay? 2010-09-25T20:59:29 oh, ha 2010-09-25T20:59:32 *** davidjliu has joined #aichallenge 2010-09-25T21:00:14 a1k0n: it's not even a basic integer programming problem :( 2010-09-25T21:00:29 its a complex integer problem! 2010-09-25T21:00:37 but i don't think you really need to reduce it to integer programming 2010-09-25T21:00:37 what, optimizing moves to a cost function? 2010-09-25T21:00:50 provided you can meet the linear constraints, can't you just take the integer floors? 2010-09-25T21:01:13 a1k0n: assuming you either want to take a planet or leave it along, fleets to planet j are either above say x or 0 2010-09-25T21:01:19 that generally gets you a starting point in integer programming, from what i understand 2010-09-25T21:01:20 so you need either or constraint to bind 2010-09-25T21:01:44 yeah, in integer programming you use linear programming for bounds 2010-09-25T21:01:45 not either or, one of (strictly) 2010-09-25T21:01:45 oh. yeah. it's an integer programming problem, given you want to take planets x, y, and z 2010-09-25T21:01:57 figuring out your givens there is the hard part 2010-09-25T21:03:01 i think i know a way to use simplex, just not sure if it'd be fast enough either 2010-09-25T21:03:12 how fast do you think phase I would run on these maps? 2010-09-25T21:03:16 i'm just doing a greedy "planet 0 has 50 extra ships, send those first" 2010-09-25T21:03:31 a1k0n: same :D 2010-09-25T21:03:35 a1k0n: i've decided i shouldn't take a planet unless i can back up my claim :P 2010-09-25T21:03:44 youwillfail: same :D 2010-09-25T21:04:18 basically track to the middle (not if it's extremely sub optimal), then if they expanded to fast, ruin them, otherwise fight over most planets 2010-09-25T21:04:39 that's my current thoughts on a heuristic in a box 2010-09-25T21:04:39 although me bots idea of "back up a claim" doesn't assume that my opponent will attempt to defend the planet 2010-09-25T21:04:47 i've never implemented simplex myself, so i have no idea how long it would take to run or even what equations you're thinking of using 2010-09-25T21:04:56 *** MarcAndrePC has joined #aichallenge 2010-09-25T21:05:01 I've run simplex by hand! 2010-09-25T21:05:06 ... and that's about it 2010-09-25T21:05:07 i guess i have in college 2010-09-25T21:05:13 *** Naith has joined #aichallenge 2010-09-25T21:05:24 i did an or course last year, then did a very bad self implementation later last year playing around 2010-09-25T21:05:30 it's a bit messy, but gets there in the end 2010-09-25T21:06:02 hi 2010-09-25T21:06:12 hi 2010-09-25T21:06:27 just to know, are most of you guys student or you graduate 2010-09-25T21:06:42 i'm last year undergrad 2010-09-25T21:06:48 I've graduated 2010-09-25T21:07:10 3rd year student 2010-09-25T21:07:24 what are you studying? 2010-09-25T21:08:16 software engineering 2010-09-25T21:08:35 and you? 2010-09-25T21:08:45 i'm a dropout 2010-09-25T21:08:48 cs, economics and maths (I'm a sucker for academics these days) 2010-09-25T21:09:21 nobody would have picked that from me prior to uni, didn't do the top maths before uni and hadn't done any programming aha 2010-09-25T21:09:40 I hadn't really done any programming before uni either 2010-09-25T21:09:52 I graduated in '89 :-) 2010-09-25T21:09:55 except mucking around on my TI83 :P 2010-09-25T21:09:59 i was born in 88 :P 2010-09-25T21:10:07 i.. started programming when i was 5 years old on a TRS-80 CoCo 2010-09-25T21:10:23 (for some loose definition of programming) 2010-09-25T21:10:39 :) 2010-09-25T21:12:09 *** Naith has quit IRC (Quit: Naith) 2010-09-25T21:13:14 gosh darn, i have such a reasonable idea to try now, but it's big and i don't have time to try it till like next weekend :( 2010-09-25T21:13:16 if even then 2010-09-25T21:21:42 *** Zaphus has quit IRC (Quit: Page closed) 2010-09-25T21:31:39 *** migi_ has quit IRC (Quit: http://boinc.berkeley.edu) 2010-09-25T21:33:07 oh crud, i see why you can't just floor non-integer solutions, hmmm 2010-09-25T21:34:55 I need 10.2 ships to take over this planet! I'll just send 10 2010-09-25T21:38:29 nearly in top 10 sigh :) 2010-09-25T21:38:57 haha, i've been in and out of the top 10 over the last couple of hours 2010-09-25T21:39:39 *** boiserm-AFK has quit IRC (Quit: Page closed) 2010-09-25T21:40:34 hmm, maybe you're right 2010-09-25T21:40:45 I wonder if it will get really harder later in the competition 2010-09-25T21:40:45 i was thinking i had problems if it spat out like 3.5 and 2.5 to make 6 2010-09-25T21:40:50 but i think the constraints eliminate that 2010-09-25T21:40:59 MarcAndrePC: almost definitely 2010-09-25T21:41:10 lol 2010-09-25T21:41:37 it's a better learning experience than most courses if you do it right though 2010-09-25T21:42:21 hmmm 2010-09-25T21:42:26 maybe 2010-09-25T21:42:58 everyone is still learning about what works and what doesn't 2010-09-25T21:43:34 I'm doing an AI but I don't know anything about doing AI or the specific technique to do AI 2010-09-25T21:43:52 kind of strange 2010-09-25T21:44:07 neither really 2010-09-25T21:44:20 just play around with different ideas for picking moves, implement them and see how they work out 2010-09-25T21:44:33 i would suggest doing as everyone else did and doing strictly heuristic based stuff to get a feel first 2010-09-25T21:44:45 wikipedia is helpful as always <3 2010-09-25T21:45:27 ya min/max tree... 2010-09-25T21:45:37 search space is way too large 2010-09-25T21:45:40 I thought of using it but didn't see the point so far 2010-09-25T21:45:50 you either need to work out a way to reduce that or use other methods 2010-09-25T21:46:22 you can reduce it by making "intelligent move'' like : to take a neutral he will send at least that much... 2010-09-25T21:46:55 yeah 2010-09-25T21:47:04 the only work I did so far is a defensive function 2010-09-25T21:47:16 but even then the problem of working out feasible "pseudo optimal" moves isn't so trivial in this instance 2010-09-25T21:47:18 at least imo 2010-09-25T21:47:39 even if you knew that you needed x ships to take over a neutral planet, you still have a huge combinatorial search for the best planets to send those ships from 2010-09-25T21:48:04 usually the closest 2010-09-25T21:48:05 it gets messier and messier the more heuristic stuff you try to throw in 2010-09-25T21:48:10 and pruning those moves are much less obvious 2010-09-25T21:48:26 MarcAndrePC: yeah, but you might want to use those units for something else 2010-09-25T21:49:30 well you separate the unit needed to defend the planet from the unit that are free (that's how I work) 2010-09-25T21:50:03 MarcAndrePC: that's what I'm doing at the moment, but I'm finding that's overly restrictive 2010-09-25T21:50:13 needed to defend based on incoming ships 2010-09-25T21:50:40 Is there really an alternative to that? if you need 10 to defend leave at least 10... 2010-09-25T21:50:55 there is, you can decide to flee 2010-09-25T21:50:55 if your fleet will make it in time to save, it isn't necessarily optimal to send from the closest planets 2010-09-25T21:51:08 as those ships might be better used elsewhere, that's what makes it so difficult ;) 2010-09-25T21:51:50 better use somewhere : defend against another attack, take a new neutral or attack enemy planet 2010-09-25T21:51:54 MarcAndrePC: also, just defending based on incoming units is not enough... play ragebot and you will see :) 2010-09-25T21:51:56 for example, it would be better to send planets from a ship further away from the enemy to save incoming fleets from the enemy rather than leave new ships grown at that planet 2010-09-25T21:52:00 sigh: if you have a good weighting function, 0-1 knapsack makes the combinatorial search easy 2010-09-25T21:52:27 i'm going to go out on a limb and say i think dp is out 2010-09-25T21:52:38 unless you work some really good dark magic in there 2010-09-25T21:53:03 I have a function that move useless ship in the back to the "front" whatever the front might be... be the front is like where the things are happening 2010-09-25T21:53:08 once you get more than a couple of planets, it's ridiculous 2010-09-25T21:53:39 sigh: cough cough :P 2010-09-25T21:53:58 jesionaj`: that's like saying that if I had a good enough evaluation function I would just need to search one move deep :P 2010-09-25T21:54:06 youwillfail: ? 2010-09-25T21:54:07 *** MarcAndrePC has quit IRC (Quit: Page closed) 2010-09-25T21:54:11 haha, true 2010-09-25T21:54:19 that's what i tried this weekend and have been bitching about run times for 2010-09-25T21:54:35 i think i'm going to move away from it anyway 2010-09-25T21:54:43 *** MarcAndrePC has joined #aichallenge 2010-09-25T21:54:54 ah ok 2010-09-25T21:55:15 dp is pretty broad, what were you using it for? optimising ship allocation for attacks? 2010-09-25T21:55:30 what is dp? 2010-09-25T21:55:35 dynamic programming 2010-09-25T21:55:36 double penetration 2010-09-25T21:55:39 oh... 2010-09-25T21:55:42 lol 2010-09-25T21:56:06 btw sigh I beat rage bot every time, I'm rank 69 2010-09-25T21:56:26 we found that we could beat dual bot on map 44 as player 1, but not as player 2 2010-09-25T21:56:27 go figure 2010-09-25T21:56:35 MarcAndrePC: then you are using more defence than just looking at incoming ships :) 2010-09-25T21:56:50 nah 2010-09-25T21:56:58 well 1 more 2010-09-25T21:57:17 my bot is plumeting today 2010-09-25T21:57:25 MarcAndrePC: what does you bot do on map30 against ragebot? 2010-09-25T21:57:27 Top 10 players: dmj111(3525), EBraun(3427), adumlah(3403), kcm1700(3378), bix0r4ever(3369), felixcoto(3350), davidjliu(3322), McLeopold(3299), wagstaff(3284), Loks1982(3275) 2010-09-25T21:57:29 when I have only 1 planet, I don't send more fleet that I need to defend if the oponent send all his fleet to attack me 2010-09-25T21:57:41 does it wait, attack opponent, or capture 20-ship neutral? 2010-09-25T21:58:14 MarcAndrePC: ah yup, I do that too, but I generalised it to more planets 2010-09-25T21:58:21 never tried this one, I don't have a script for testing all map, but I will try it now 2010-09-25T21:59:19 from what you said, you should be fine on that map 2010-09-25T21:59:36 hmmm.. would the starter packages ALWAYS draw with themselves? 2010-09-25T21:59:43 that might be a way of eliminating starter package bots 2010-09-25T21:59:55 I take 20 neutral, win in 34 turn 2010-09-25T22:00:11 nah they dont draw 2010-09-25T22:00:24 they attack the same planet so they are really stupid, one always win 2010-09-25T22:00:46 is it possible to construct a map on which they always draw and wouldn't normally? 2010-09-25T22:00:58 yup 2010-09-25T22:01:00 so they pick symmetric moves but not trivially nothing 2010-09-25T22:01:06 make weakest the center one 2010-09-25T22:01:11 they will always draw 2010-09-25T22:01:28 because there are a LOT of starter package bots on the main server, ideally it would be nice to disable them from playing as many games as all the other bots 2010-09-25T22:01:29 they will always send ship at the same time to the same planet and the same amount and never take any planets 2010-09-25T22:01:59 they dont pick symmetric move, they attack the same planet 2010-09-25T22:02:20 planet have an ID, 1 , 2 ,3 4, they run down the list, pick the first weakest and attack it 2010-09-25T22:03:00 if the weakest is at the same distance from both side they will never be able to take it, and continue attacking it with equivalent attack 2010-09-25T22:03:25 ragebot will always draw against itself though 2010-09-25T22:03:38 really? 2010-09-25T22:03:44 why not? 2010-09-25T22:03:52 yeah okay, i think that'd be a reasonable way to remove starter bots, get 2 or 3 maps like that, run the lower ranked bots against a starter package bot on them, if it's all draws, no more games for them? 2010-09-25T22:03:55 it just sends all ships to weakest opponent 2010-09-25T22:04:05 didn't ever check how rage bot work 2010-09-25T22:04:27 do you guys think that'd work? 2010-09-25T22:04:37 you don't want to disable any legitimate accounts 2010-09-25T22:04:39 ANY 2010-09-25T22:05:13 if they dont take any planet for all the game, it's 2 starter probably... 2010-09-25T22:05:50 but if you can just judge on draw, win or lose to determine that's a problem 2010-09-25T22:05:51 the maps would also want to have obvious better moves for any other bots 2010-09-25T22:07:05 in the tron contest they just did a file comparison to the starter packages 2010-09-25T22:07:48 is there any way to test your bot against other player without using tcp server, I can't seems to make them work 2010-09-25T22:08:02 my bot always timed out for no reason... 2010-09-25T22:10:12 Help me understand why a linear regression based on the data in the game replays of the top bots using a model like Y (some representation of moves source/destination/quantity) = X (map state + binary won/lost) B is a bad idea? map state would be a series of matrices representing the map chopped up into cells and representing growth rates, {enemy, my, neutral} forces, reinforcement arrivals at time t+1...n*, and then a binary whether the player one or lo 2010-09-25T22:11:15 in english? 2010-09-25T22:11:28 o espanol 2010-09-25T22:13:00 nah your strat just seems impossible to understand 2010-09-25T22:13:46 yeah say what? what are you trying to understand? 2010-09-25T22:14:05 try it out and see if it's a bad idea 2010-09-25T22:14:24 you aren't necessarily going to know which map you're playing on in the final tournament though, nor are you guaranteed those will be the only maps 2010-09-25T22:15:40 hey everyone, I was just wondering if there is any way to debug games previously played online 2010-09-25T22:16:13 on tcp? you run locally, so set up your own debugging 2010-09-25T22:16:24 I mean from the actual online contest 2010-09-25T22:16:30 if you want to debug on the main server, it's going to have to be in the form of sending signals 2010-09-25T22:16:40 like something fucked up, send 1 ship 2010-09-25T22:16:44 davidjliu: http://www.ai-contest.com/forum/viewtopic.php?f=17&t=628 2010-09-25T22:17:01 k, thanks 2010-09-25T22:17:04 that will let you parse the playback string, and feed the state into your bot locally 2010-09-25T22:17:33 cool 2010-09-25T22:19:07 *** Utkarsh has quit IRC (Read error: Connection reset by peer) 2010-09-25T22:19:25 *** Utkarsh has joined #aichallenge 2010-09-25T22:24:54 anyone is using parasite attack : wait for enemy to attack a neutral then send some ship to take it after? 2010-09-25T22:25:04 does it do a big difference in game? 2010-09-25T22:25:58 MarcAndrePC: I don't have specific code for it, but my bot does enjoy sending fleets to arrive the turn after it's opponents. 2010-09-25T22:26:07 mine too 2010-09-25T22:26:07 *** bss03-AFK is now known as bss03 2010-09-25T22:26:50 my code is not specific either 2010-09-25T22:26:54 Somtimes that hurts me though, 'cause my bot is still a bit stupid. 2010-09-25T22:27:00 my bot resend ship to retake it currently so I wonder of the utility 2010-09-25T22:27:30 *** boiserm has joined #aichallenge 2010-09-25T22:27:32 It depends on the distance the neutral is from your respective planets. 2010-09-25T22:27:43 yup 2010-09-25T22:27:57 If it's a neutral that is closer to you, waiting for you opp to send ships is good, because you'll be able to retake and hold at a discount. 2010-09-25T22:28:00 i will do that to a nearest neutral if they are stupid enough 2010-09-25T22:28:12 and if it is possible to take a planet, but let your opponent cover the initial cost, that good in general 2010-09-25T22:28:19 but you don't want to wait and them expand and grow 2010-09-25T22:28:26 It the other way, you might get it for 1-2 turns, but they'll be able to retake and hold and you'll have lost ships for nothing. 2010-09-25T22:28:30 but you also don't want to expand so fast that they can ruin you 2010-09-25T22:28:53 it's such an awesome game 2010-09-25T22:28:55 * bss03 Is "DaTwinkDady" on the official server and TCP server. 2010-09-25T22:28:58 so I'm considering playing around with making a Haskell bot :-P Does anyone have a favorite unofficial starter pack? 2010-09-25T22:29:10 thats an..... interesting name 2010-09-25T22:29:24 boiserm: I sorta use the one from jaspervdj. 2010-09-25T22:29:46 bss03 sorta as in with custom modifications? 2010-09-25T22:29:58 boiserm: MY github project might also be a good place to start; my bot is under LGPL. 2010-09-25T22:30:36 I would like the first ranked bot to be open source :) 2010-09-25T22:30:43 :) 2010-09-25T22:30:49 MarcAndrePC: most people relesae their code at the end 2010-09-25T22:30:49 boiserm: Yeah. I didn't change his main IO loop, or the GameState/Planet/Fleet/Order data structures, but I added a simulation engine of my own and dropped whatever he had. 2010-09-25T22:31:27 http://github.com/stephen-smith/ai-contest-2010 (I think; that's from memory.) 2010-09-25T22:31:42 hey sigh, are you sighbot? 2010-09-25T22:31:48 yup 2010-09-25T22:31:48 he is 2010-09-25T22:31:55 heh, we're literally tied right now 2010-09-25T22:31:57 in the rankings 2010-09-25T22:32:02 Last I checked I was still in the Top 100. I have 3-4 modifications to make tonight, too. 2010-09-25T22:32:07 oops, not anymore 2010-09-25T22:32:09 nope :P 2010-09-25T22:32:11 I lost, I guess 2010-09-25T22:32:35 :) 2010-09-25T22:33:46 i love being able to use the arrow keys on canvas visualiser 2010-09-25T22:34:34 is there a way to use that on ones own computer yet? 2010-09-25T22:34:43 sure, let me link you 2010-09-25T22:34:51 :) 2010-09-25T22:34:59 http://github.com/DanielVF/Planet-Wars-Canvas-Visualizer 2010-09-25T22:35:27 use visualise_locally.py as the visualiser 2010-09-25T22:36:45 do i need to download python? 2010-09-25T22:37:07 you will need to have python installed, yes 2010-09-25T22:37:51 okay, downloading 2010-09-25T22:37:53 cheers 2010-09-25T22:38:02 the python code is very simple, if you have some aversion to python and want to port to a different language :P 2010-09-25T22:38:22 i have no aversion, haskell and python are on the to learn list 2010-09-25T22:38:26 but so are a lot of things 2010-09-25T22:38:29 :) 2010-09-25T22:42:15 *** boiserm has left #aichallenge ("but I don't *wanna* sleep :( I wanna take over planets!") 2010-09-25T22:48:46 my bots mean http://ai-contest.com/visualizer.php?game_id=4765257 2010-09-25T22:50:32 haha, move 54 looks so funny... with blue attacking the 43-ship neutral 2010-09-25T22:51:17 lmao 2010-09-25T22:51:54 only slightly over the top aa 2010-09-25T22:53:53 Cyndre: you have a bug on move 26? attacking the 20 ship neutral? 2010-09-25T22:54:01 same game? 2010-09-25T22:54:05 yup 2010-09-25T22:54:53 hmmm 2010-09-25T22:55:12 it looks like its next planet choice didnt count the incomming fleet (or didnt get the time right) 2010-09-25T22:55:34 looks like a timing issue 2010-09-25T22:56:12 light blue didn't have any defensive function 2010-09-25T22:56:33 easy win 2010-09-25T22:56:36 yeah well, Cyndre is picking on people ranked ~1100 2010-09-25T22:56:57 5 games already - much nicer 2010-09-25T22:57:15 I do have a bug though 2010-09-25T22:57:22 Cyndre: do you test on tcp? 2010-09-25T22:57:28 Top 10 players: dmj111(3529), EBraun(3433), adumlah(3391), kcm1700(3388), bix0r4ever(3388), felixcoto(3348), McLeopold(3306), wagstaff(3293), DocBot(3289), sighbot(3288) 2010-09-25T22:57:36 beat whiteside-cobalt 2010-09-25T22:58:02 http://72.44.46.68/canvas?game_id=163674 2010-09-25T22:59:12 dam it why I cant play on tcp :( 2010-09-25T22:59:43 Im just using the windows one someone made in here (sorry I cant remember his name right now) 2010-09-25T22:59:47 MarcAndrePC: what happens? 2010-09-25T23:00:10 always got an error : bot timed out 2010-09-25T23:00:19 :( 2010-09-25T23:00:20 do you need a different code to play on tcp? 2010-09-25T23:00:43 no, should it should just work 2010-09-25T23:01:04 work locally, work on main... 2010-09-25T23:01:42 so it connects to the tcp server, starts a game and times out on the first move? 2010-09-25T23:01:52 if your bot isn't an executable you need to make a script to invoke it 2010-09-25T23:02:13 eg ./MyBot.sh instead of trying to pass "java MyBot" into tcp as one argument 2010-09-25T23:02:37 exactly 2010-09-25T23:03:05 oh really? I use jave and was passing java MyBot... 2010-09-25T23:03:32 that's probably it 2010-09-25T23:03:37 do you have the code for the script? 2010-09-25T23:04:42 put "#!/bin/sh" as the first line and "java MyBot" as the second 2010-09-25T23:04:52 then chmod +x the script file 2010-09-25T23:07:00 will it work on windows? 2010-09-25T23:07:26 oh, windows 2010-09-25T23:07:31 that's linux right 2010-09-25T23:07:36 yeah 2010-09-25T23:07:52 MarcAndrePC: Try http://ai-contest.com/forum/viewtopic.php?p=3836#p3836 2010-09-25T23:08:05 MarcAndrePC: It'll tell you if it can't run your bot. 2010-09-25T23:08:14 sorry, i'm not really experienced programmatically with windows 2010-09-25T23:09:29 hmmm this damn visualiser wont load :) 2010-09-25T23:09:29 Ouch. 8-8Bot is a lot better than 8-7Bot. 2010-09-25T23:09:32 :(* 2010-09-25T23:10:00 youwillfail: what's happening? 2010-09-25T23:10:10 I try, bss03 2010-09-25T23:10:12 Olathe: that jar what the thing I was using for connecting to the tcp actually 2010-09-25T23:10:38 will check the -d debug mode option to see... 2010-09-25T23:10:47 i run the bat file and it just closes, perhaps i should try from command line and see what it says 2010-09-25T23:10:59 MarcAndrePC: Ahh, OK :) 2010-09-25T23:12:01 MarcAndrePC: Yeah, debug will give you timestamps, so you can see if the server actually gives your bot any time before it says you lose. 2010-09-25T23:12:13 claims python isn't a recognized internal or external command 2010-09-25T23:12:24 i added the python folder to the environment variables 2010-09-25T23:12:41 youwillfail: well, for now just give the full to the python exe 2010-09-25T23:12:49 I don't know much about windows 2010-09-25T23:12:55 *full path 2010-09-25T23:13:40 *** Hicklc01 has joined #aichallenge 2010-09-25T23:13:44 works, thanks 2010-09-25T23:13:57 cool 2010-09-25T23:14:11 You currently have -352 Elo There are currently 33 connections and 16 games running Your map is maps/map366.txt Your opponent is miscui with -253 Elo This is game_id=163915 I javaexample timed out You LOSE against miscui Exiting because server closed connection 2010-09-25T23:14:27 that's all I get 2010-09-25T23:14:42 what's the ping time like to the server for you? 2010-09-25T23:14:49 i get like 500ms and i'm usaully fine 2010-09-25T23:14:54 can't ping the new server 2010-09-25T23:14:56 how do I get it? 2010-09-25T23:15:01 ... 2010-09-25T23:15:02 hmm okay 2010-09-25T23:15:42 it's afraid of ping :) 2010-09-25T23:15:46 MarcAndrePC: What are the timestamps on those (you can paste at codepad.org)? 2010-09-25T23:16:05 I dont get any other message in my console 2010-09-25T23:16:32 how long does your bot usually take to decide on its move? 2010-09-25T23:16:37 are the timestamp send to a file? 2010-09-25T23:16:41 my usual bot is usualyl under 0.1 seconds most of the time 2010-09-25T23:16:51 0.01* 2010-09-25T23:17:21 maybe 100 ms max. Didn't make a function to calculate it 2010-09-25T23:17:22 but 2010-09-25T23:17:28 There should be something like [3492708.0979970080] on the left of each line. 2010-09-25T23:17:29 I will try connecting with starter package 2010-09-25T23:17:47 we will see where the problem is like that 2010-09-25T23:17:50 you WILL want a timer for the final tournament 2010-09-25T23:18:00 really? 2010-09-25T23:18:00 if you exceed the time limit on a turn you get disqualified 2010-09-25T23:18:20 yeah that's harsh 2010-09-25T23:18:31 super harsh, and imo entirely unecessary 2010-09-25T23:18:34 but those are the rules 2010-09-25T23:18:46 could just call it a lose... 2010-09-25T23:18:59 either way, you will fail 2010-09-25T23:19:04 i had always expected that until i got disqualfied last tournament :P 2010-09-25T23:20:26 Who's bot is vortex? 2010-09-25T23:20:32 rwa 2010-09-25T23:20:39 hi 2010-09-25T23:20:49 oh yeah, those are mine 2010-09-25T23:20:54 yay, I can summon people 2010-09-25T23:21:04 Oh, just curious, I've been playing you a lot 2010-09-25T23:21:38 which one is yours? 2010-09-25T23:21:42 on main site or tcp? 2010-09-25T23:21:53 Anything in the format 8-*Bot 2010-09-25T23:21:56 ah 2010-09-25T23:21:56 tcp 2010-09-25T23:21:58 yeah 2010-09-25T23:22:10 there's no concept of "playing a lot" on the main site :P 2010-09-25T23:22:18 true dat 2010-09-25T23:22:50 i have no idea how they plan to run the final tournament in 5 days 2010-09-25T23:23:24 i'd rather they just let it run for a week or two with a live leaderboard 2010-09-25T23:23:44 (last time they closed off viewing the leaderboard while the final tournament ran, removed lots of the fun i think) 2010-09-25T23:24:09 yeah, that would suck 2010-09-25T23:24:31 I hope google can step up for the final tournament with some compute resources 2010-09-25T23:25:00 I mean it's kind of ridiculous to have a contest sponsored by google running on a single shitty machine 2010-09-25T23:25:13 that is woefully, woefully underpowered 2010-09-25T23:25:29 well, they are starting to run on more servers now 2010-09-25T23:25:45 cool, when did that start 2010-09-25T23:25:50 google seems like just giving their name, and nothing else 2010-09-25T23:25:50 yeah, the google thing is more of it looks good for both parties 2010-09-25T23:26:24 rwa: http://www.ai-contest.com/forum/viewtopic.php?f=18&t=566&start=30 2010-09-25T23:27:09 DanielVonFange's post says: "I need to finish up getting games and bots distributed to cloud servers. Today I got the first cloud server up and running - although at a duct tape level." 2010-09-25T23:28:04 *** McLeopold has joined #aichallenge 2010-09-25T23:31:20 is the tournie a round robin or a bracket style 2010-09-25T23:33:41 *** iFire has quit IRC (Read error: Connection reset by peer) 2010-09-25T23:35:06 couldn't do round robin, too many bots 2010-09-25T23:35:25 i would personally like to see one with the top 25 bots or so at the end, but i don't think it'll happen 2010-09-25T23:37:56 MarcAndrePC: You can try the Windows C client: http://ai-contest.com/forum/viewtopic.php?p=3786#p3786 2010-09-25T23:38:02 MarcAndrePC: If it works, please let me know. 2010-09-25T23:41:01 *** scottj has quit IRC (Quit: leaving) 2010-09-25T23:41:33 *** FlagCapper has quit IRC (Ping timeout: 252 seconds) 2010-09-25T23:41:42 *** iFire has joined #aichallenge 2010-09-25T23:42:30 *** FlagCapper has joined #aichallenge 2010-09-25T23:44:44 it would only take about 32 years for a round robin with 2249 bots 2010-09-25T23:45:34 ish 2010-09-25T23:46:29 If Google is actually sponsoring this, I'm sure they can use some huge cluster or another. 2010-09-25T23:48:01 damn, dodgy loss: http://www.ai-contest.com/visualizer.php?game_id=4767240 2010-09-25T23:48:51 woah, my bot also bugged out against Polski 2010-09-25T23:49:04 ha, on the same map 2010-09-25T23:49:20 jk, different map 2010-09-25T23:49:47 *** iFire has quit IRC (Read error: Connection reset by peer) 2010-09-25T23:51:45 *** iFire has joined #aichallenge 2010-09-25T23:52:11 checked it locally, no illegal move... must have been timeout... I blame the server 2010-09-25T23:57:28 Top 10 players: dmj111(3545), adumlah(3426), bix0r4ever(3402), EBraun(3399), kcm1700(3394), felixcoto(3365), McLeopold(3312), davidjliu(3305), Zilog8(3272), wagstaff(3265) 2010-09-25T23:58:13 davidjliu: we swapped places :P 2010-09-25T23:59:35 *** delt0r_ has joined #aichallenge