2010-10-10T00:00:19 Top 10 players: felixcoto(3803), ofrias(3802), MariusCG(3800), rsergio(3780), Hazard(3774), EBraun(3741), cfaftw(3735), adumlah(3696), dmj111(3685), sequoh(3678) 2010-10-10T00:04:19 *** Arthur_ has quit IRC (Ping timeout: 264 seconds) 2010-10-10T00:05:33 pgpaskar_: what did he do? 2010-10-10T00:05:56 *** Arthur has joined #aichallenge 2010-10-10T00:07:03 *** jesionaj has joined #aichallenge 2010-10-10T00:12:24 *** delt0r_ has joined #aichallenge 2010-10-10T00:12:45 http://72.44.46.68/canvas?game_id=400733 2010-10-10T00:14:14 *** delt0r___ has quit IRC (Ping timeout: 265 seconds) 2010-10-10T00:24:29 *** oldman has joined #aichallenge 2010-10-10T00:24:43 *** jesionaj has quit IRC (Read error: Connection reset by peer) 2010-10-10T00:25:25 *** agweber has quit IRC (Quit: Page closed) 2010-10-10T00:26:16 *** jesionaj has joined #aichallenge 2010-10-10T00:36:46 *** hellman has joined #aichallenge 2010-10-10T00:55:10 is there a time limit on how long it takes to compile. I am using Java interfaces and there may be an issue where those take longer to compile than classes 2010-10-10T00:55:43 it has to compile by the end of the contest 2010-10-10T00:56:05 blbrown_win3_: i don't think there is 2010-10-10T00:56:13 amstan cool 2010-10-10T00:56:15 how long? 2010-10-10T00:56:51 It takes a couple of minutes on my machine. Many classes/interfaces 2010-10-10T01:00:17 contestbot: repo 2010-10-10T01:00:17 amstan: repo = http://code.google.com/p/ai-contest/ 2010-10-10T01:00:20 Top 10 players: felixcoto(3806), ofrias(3795), rsergio(3782), Hazard(3763), cfaftw(3744), MariusCG(3730), EBraun(3720), adumlah(3696), dmj111(3682), sequoh(3673) 2010-10-10T01:02:10 *** oldman has quit IRC (Quit: Page closed) 2010-10-10T01:03:33 hm 2010-10-10T01:03:56 debating implementing an algorithm that will allocate about a megabyte on the stack 2010-10-10T01:04:09 in some cases 2010-10-10T01:04:35 Zannick, what language, C? 2010-10-10T01:04:39 yes 2010-10-10T01:06:40 *** wh1teside has quit IRC (Quit: Leaving) 2010-10-10T01:07:01 I think the default stack size on linux is 8MB so you should be ok 2010-10-10T01:10:42 why on the stack though? recursive? 2010-10-10T01:11:12 no, rather than invoke malloc 2010-10-10T01:11:24 why? 2010-10-10T01:11:31 * Zannick shrug 2010-10-10T01:11:40 *** Titankiller has joined #aichallenge 2010-10-10T01:11:46 just put it on the heap? and don't need to worry about blowing the stack 2010-10-10T01:11:47 i'm still trying to figure out a good way to do this 2010-10-10T01:16:36 *** krokokrusa has joined #aichallenge 2010-10-10T01:19:31 *** Titankiller has quit IRC (Ping timeout: 272 seconds) 2010-10-10T01:23:25 *** yasith has joined #aichallenge 2010-10-10T01:39:18 Oh you cannot call virtual functions in C++-constructors. Thats opposite to what i'm used from Delphi and Java, but good to know I'll need some init()-function 2010-10-10T01:41:30 is virtual function a non final function? 2010-10-10T01:42:00 Yes, it is a function you can 'modify' in derieved classes. 2010-10-10T01:44:51 I often do some initialization in the base class constructor using data i get from functions of the sub-classes I create. That doesn't work in c++. In the base class constructor the object you create is always treated as if it was of the base class. 2010-10-10T01:45:32 *** Mathnerd314 has quit IRC (Ping timeout: 245 seconds) 2010-10-10T01:46:20 in Java, they generally discourage calling other methods in the constructor because the calling method could blow up 2010-10-10T01:46:28 and then the object won't be created. 2010-10-10T01:47:09 frontier you can 2010-10-10T01:47:24 you can't call them because its the constructor that sets up the vtable 2010-10-10T01:50:33 *** Utkarsh has quit IRC (Ping timeout: 272 seconds) 2010-10-10T01:52:43 blbrown_win3_: I'm aware of that, and usually call functions that return constant stuff, that doesn't depend on the constructor. 2010-10-10T01:54:18 *** Utkarsh has joined #aichallenge 2010-10-10T01:55:43 I have some static objects in my sub-classes that provide default values, and in the base class constructor I tried to query these static objects from the sub classes and initialize some members of the sub-class with the default values they provide. 2010-10-10T01:56:59 The following tunable_count() and tunable_declaration() are virtual funcitons: 2010-10-10T01:56:59 for (int i = tunable_count() - 1; i >= 0; --i) { 2010-10-10T01:56:59 tunable(i) = tunable_declaration(i).default_value(); 2010-10-10T01:56:59 } 2010-10-10T02:00:19 Top 10 players: felixcoto(3791), rsergio(3784), ofrias(3778), Hazard(3759), cfaftw(3750), MariusCG(3746), EBraun(3726), sighbot(3686), adumlah(3683), dmj111(3678) 2010-10-10T02:01:48 yay, 8th 2010-10-10T02:03:58 I think I can safely say my new bot is better now :) 2010-10-10T02:12:14 looks like dmj111 is going to stabilize a little lower than he was before 2010-10-10T02:13:31 yeah, this bot was just an experiment for him, I think 2010-10-10T02:13:38 he wasn't sure how it would do 2010-10-10T02:14:18 ahh 2010-10-10T02:26:12 *** blbrown_win3_ has quit IRC (Quit: Leaving) 2010-10-10T02:36:37 *** Titankiller has joined #aichallenge 2010-10-10T02:43:55 http://ai-contest.com/visualizer.php?game_id=5705089 <= massiou crashed / did not start / timeout. WIN massiou 2010-10-10T02:45:31 *** Palmik has joined #aichallenge 2010-10-10T02:47:07 *** tapwater has quit IRC (Quit: tapwater) 2010-10-10T03:00:20 Top 10 players: rsergio(3815), felixcoto(3799), ofrias(3795), Hazard(3781), cfaftw(3765), MariusCG(3738), EBraun(3737), dmj111(3717), adumlah(3695), _iouri_(3686) 2010-10-10T03:00:51 pretty much known bug that the visualizer always shows the first player as the one to have the error 2010-10-10T03:01:43 http://code.google.com/p/ai-contest/issues/detail?id=189 2010-10-10T03:05:23 Could it be that the maps in the starter pack aren't exactly symmetric? It seems there are little differences in the distances, but I'm not sure yet. 2010-10-10T03:07:11 not that I've heard of before. I do know the java visualizer view starts up "squished" though 2010-10-10T03:07:16 Hmm, they should have centered them all around 0,0 to go sure there will be no rounding errors. 2010-10-10T03:08:15 I ran my bot against the same version on map 2 and on one turn the game becomes unsymmetrical although the bot is 100% deterministic. 2010-10-10T03:08:27 not surprising 2010-10-10T03:08:49 janzert: You are NOT surprised? 2010-10-10T03:08:54 nope :) 2010-10-10T03:09:06 May I ask why? 2010-10-10T03:09:10 I've seen and heard that a lot 2010-10-10T03:09:15 * krokokrusa nods 2010-10-10T03:09:38 the positions of the planets are symettric, but the map as a whole is not 2010-10-10T03:09:40 one reason that can cause it are the planets are fed to both bots in the same order 2010-10-10T03:09:48 one possible reason is that when you sort planets by some criteria, some planets have same value 2010-10-10T03:09:54 because of planet ids, yes 2010-10-10T03:10:24 *** Titankiller has quit IRC (Quit: Nettalk6 - www.ntalk.de) 2010-10-10T03:10:29 krokokrusa: That's it! 2010-10-10T03:11:37 you can see the behavior just by fighting bullybot against itself btw 2010-10-10T03:11:47 at least I don't reorder the planets for player 2, so the order in which they appear in the list has an influence. 2010-10-10T03:14:11 I will still check the distances of the planets, just to go sure. 2010-10-10T03:42:04 Distances were ok, krokokrusa was right about the order of planets, I corrected the ids for player 2 and have equal games now on all maps :) 2010-10-10T03:45:48 *** mega1 has joined #aichallenge 2010-10-10T03:45:58 *** krokokrusa has quit IRC (Ping timeout: 264 seconds) 2010-10-10T03:50:51 *** sigh has quit IRC (Remote host closed the connection) 2010-10-10T04:00:20 Top 10 players: rsergio(3813), felixcoto(3797), ofrias(3793), Hazard(3778), cfaftw(3763), MariusCG(3736), EBraun(3736), dmj111(3715), sighbot(3701), _iouri_(3690) 2010-10-10T04:04:20 *** Accoun has quit IRC (Ping timeout: 276 seconds) 2010-10-10T04:04:37 *** Accoun has joined #aichallenge 2010-10-10T04:17:11 *** epona has quit IRC (Ping timeout: 276 seconds) 2010-10-10T04:26:58 hi Developers 2010-10-10T04:30:17 hi 2010-10-10T04:39:28 *** Accoun has quit IRC (Ping timeout: 260 seconds) 2010-10-10T04:40:17 hi 2010-10-10T04:53:15 *** noid30188 has joined #aichallenge 2010-10-10T04:53:32 *** noid30188 is now known as Acount 2010-10-10T05:00:21 Top 10 players: rsergio(3836), felixcoto(3821), ofrias(3811), Hazard(3788), MariusCG(3773), cfaftw(3770), EBraun(3742), dmj111(3723), sighbot(3707), _iouri_(3698) 2010-10-10T05:01:11 *** fawek has joined #aichallenge 2010-10-10T05:05:46 *** fawek has quit IRC (Ping timeout: 252 seconds) 2010-10-10T05:12:16 *** fawek has joined #aichallenge 2010-10-10T05:13:36 *** sigh has joined #aichallenge 2010-10-10T05:22:39 *** Blkt has joined #aichallenge 2010-10-10T05:24:32 good day everyone 2010-10-10T05:36:03 happy 42 day 2010-10-10T05:37:21 *** Blkt has quit IRC (Ping timeout: 240 seconds) 2010-10-10T05:38:25 oPless: ? 2010-10-10T05:46:32 *** Meatkat has joined #aichallenge 2010-10-10T05:55:58 *** BtbN has joined #aichallenge 2010-10-10T05:57:47 *** BtbN has quit IRC (Client Quit) 2010-10-10T05:59:54 *** BtbN has joined #aichallenge 2010-10-10T06:00:21 Top 10 players: rsergio(3835), felixcoto(3820), ofrias(3810), Hazard(3787), MariusCG(3771), cfaftw(3769), EBraun(3740), dmj111(3721), sighbot(3706), _iouri_(3695) 2010-10-10T06:04:39 *** antimatroid has joined #aichallenge 2010-10-10T06:06:02 *** antimatroid1 has quit IRC (Ping timeout: 255 seconds) 2010-10-10T06:08:33 *** vizier has joined #aichallenge 2010-10-10T06:08:59 *** Acount has quit IRC (Ping timeout: 276 seconds) 2010-10-10T06:13:11 *** Acount has joined #aichallenge 2010-10-10T06:18:44 well that bot i put up right before leaving friday sucked 2010-10-10T06:19:45 wtf guys, how did java get 1,2 and 3? 2010-10-10T06:39:25 *** Arthur has quit IRC (Ping timeout: 260 seconds) 2010-10-10T06:40:52 I like Java 2010-10-10T06:41:32 But I expected C++ to win, because of the additional free time the bot can use to think 2010-10-10T06:41:37 *** Arthur has joined #aichallenge 2010-10-10T06:42:53 I am currently "auto-tuning" some parameters of my bot. It is fun to see where my 'human' tuning was quite good and where I was far off. 2010-10-10T06:49:08 *** mega1 has quit IRC (Ping timeout: 265 seconds) 2010-10-10T06:51:16 java is not slow 2010-10-10T06:51:20 bad java is... 2010-10-10T06:51:28 just as bad C/C++ 2010-10-10T06:52:05 if the sand box doesn't cause undue performance resstrictions there is no reason to expect java to have a performance penatly compared to C++ 2010-10-10T06:52:05 Frontier: 1 sec is plenty of time for every language 2010-10-10T06:52:28 It is not slow, but does it compile to native machine code? 2010-10-10T06:52:37 yes java does 2010-10-10T06:52:39 jit 2010-10-10T06:52:57 for the win... sure java is not the right tool for every job 2010-10-10T06:53:09 but in this case its as good a match as C++ 2010-10-10T06:53:19 I see trouble coming when everyone uses the 1 second. Did you check? 1 second * 200 rounds * 100 players = ??? hours of CPU time 2010-10-10T06:53:31 yes 2010-10-10T06:53:52 I have said that the current number of games per min are very high considering that 2010-10-10T06:54:35 but thats not a launage issue... unless its PHP :) 2010-10-10T06:55:04 i tried one function ported from c++ to php 2010-10-10T06:55:10 and php was just 15% slower 2010-10-10T06:55:17 lol awsome 2010-10-10T06:55:56 these days memeory access patents seem to have a larger effect that anything that can sort of fit in level 1cache 2010-10-10T06:56:01 kinda only two dimensional vectors vs two dimensional arrays 2010-10-10T06:56:36 i was expecting something 50000% speed difference 2010-10-10T06:56:43 yea 2010-10-10T06:56:45 me too 2010-10-10T06:57:00 thats often what i see between java/C++ and python 2010-10-10T06:57:20 but i am not good at python... so it could just be me 2010-10-10T06:57:34 *** epona has joined #aichallenge 2010-10-10T07:00:22 Top 10 players: rsergio(3832), felixcoto(3817), ofrias(3807), Hazard(3784), MariusCG(3768), cfaftw(3766), EBraun(3737), dmj111(3716), sighbot(3704), _iouri_(3690) 2010-10-10T07:00:26 *** BtbN has quit IRC (Quit: Verlassend) 2010-10-10T07:06:17 *** krokokrusa has joined #aichallenge 2010-10-10T07:11:11 *** Sylph has quit IRC (Read error: Connection reset by peer) 2010-10-10T07:11:13 *** Sylph2 has joined #aichallenge 2010-10-10T07:14:13 *** fawek has quit IRC (Quit: No Ping reply in 180 seconds.) 2010-10-10T07:30:44 *** boegel has joined #aichallenge 2010-10-10T07:30:50 yo y'all 2010-10-10T07:31:11 I seem to recall there were some bugs in the Java starter package, were these resolved recently? 2010-10-10T07:31:22 im not sure. 2010-10-10T07:31:30 do you think you can help me though? 2010-10-10T07:31:51 I have the starter package but im not sure how to make my own game environment 2010-10-10T07:32:00 *** BtbN has joined #aichallenge 2010-10-10T07:32:17 shoot 2010-10-10T07:32:36 not sure how long I'll be here, but I may be able to help you 2010-10-10T07:32:38 boegel: there are a few non official starter packages on the forums... they may be a better start 2010-10-10T07:33:00 ahh 2010-10-10T07:33:02 *** fawek has joined #aichallenge 2010-10-10T07:33:35 Aha! Auto-tune spit out that I need to be less restrictive with moving ships to the front. 2010-10-10T07:33:43 delt0r: well, I already have an environment 2010-10-10T07:33:54 in fact, I'm using Haskell, not Java :) 2010-10-10T07:33:59 right 2010-10-10T07:36:56 delt0r, it's 10/10/10 and 101010b =42d 2010-10-10T07:37:11 yea--I just saw a /. post 2010-10-10T07:37:24 *** bartwe_ has quit IRC (Ping timeout: 255 seconds) 2010-10-10T07:37:28 *** bartwe_ has joined #aichallenge 2010-10-10T07:37:56 delt0r: I am using the PlayGame.jar and ShowGame.jar stuff to visualize games though... one of them had a bug in the original starters package, right? 2010-10-10T07:38:19 I think thats been fixed... 2010-10-10T07:38:46 ah, cool 2010-10-10T07:38:48 o 2010-10-10T07:38:50 and there are now a number of game engines avaible that are somewhat better 2010-10-10T07:38:57 maybe I should reload it 2010-10-10T07:39:00 delt0r: where? 2010-10-10T07:39:04 it won't come up for some reason 2010-10-10T07:39:07 when i click it 2010-10-10T07:39:18 on the forums... theres a few 2010-10-10T07:39:47 including my one 2010-10-10T07:39:58 http://www.ai-contest.com/forum/viewtopic.php?f=18&t=751/forum/viewtopic.php?f=18&t=751 2010-10-10T07:40:10 wtf 2010-10-10T07:40:11 sorry 2010-10-10T07:40:27 http://www.ai-contest.com/forum/viewtopic.php?f=18&t=751 2010-10-10T07:42:04 delt0r: is yours the best one out there? :p 2010-10-10T07:42:12 well 2010-10-10T07:42:25 perhaps not.... Its good for me.. quite a few downloads... 2010-10-10T07:42:31 its not polished 2010-10-10T07:42:35 but should be stable 2010-10-10T07:42:44 the source is included 2010-10-10T07:42:53 you can use the command line too 2010-10-10T07:43:17 The reply tab is the reason most seem to use it 2010-10-10T07:44:03 And it has income as well...and you can just drag the turn to where you want 2010-10-10T07:44:22 Human play is fun... but even dumb bots beat me 2010-10-10T07:44:39 And watching tcp games is cool when your winning 2010-10-10T07:45:00 nice set of features 2010-10-10T07:45:05 * boegel will give it a go 2010-10-10T07:45:17 deltor 2010-10-10T07:45:35 bots actions are instantaneous 2010-10-10T07:45:39 ^.^ 2010-10-10T07:48:05 I lost against ragebot ... but consdering I've not trained it 138 moves is impressive :) http://72.44.46.68/canvas?game_id=406630 2010-10-10T07:49:07 whos who 2010-10-10T07:49:26 ah 2010-10-10T07:49:43 delt0r: nice game engine dude, very useful 2010-10-10T07:50:38 *** Meatkat has quit IRC (Ping timeout: 240 seconds) 2010-10-10T07:51:15 *** Meatkat has joined #aichallenge 2010-10-10T07:52:07 *** JensT1 has joined #aichallenge 2010-10-10T07:53:48 *** Frontier has quit IRC (Ping timeout: 276 seconds) 2010-10-10T08:00:23 Top 10 players: rsergio(3840), felixcoto(3825), ofrias(3815), Hazard(3792), MariusCG(3775), cfaftw(3773), EBraun(3750), dmj111(3723), sighbot(3712), _iouri_(3697) 2010-10-10T08:06:00 *** mega1 has joined #aichallenge 2010-10-10T08:08:37 *** Frontier has joined #aichallenge 2010-10-10T08:09:27 I think my laptop overheated :/ 2010-10-10T08:12:17 raise it off whatever you have it on with something small 2010-10-10T08:12:28 allows for air circulation 2010-10-10T08:12:56 *** delt0r___ has joined #aichallenge 2010-10-10T08:14:11 *** delt0r_ has quit IRC (Ping timeout: 240 seconds) 2010-10-10T08:15:24 *** boegel has quit IRC (Quit: *poof!*) 2010-10-10T08:16:46 *** vizier has quit IRC (Quit: Page closed) 2010-10-10T08:24:21 *** area has quit IRC (Ping timeout: 276 seconds) 2010-10-10T09:00:23 Top 10 players: rsergio(3847), ofrias(3828), felixcoto(3828), Hazard(3800), MariusCG(3782), cfaftw(3778), EBraun(3757), dmj111(3728), sighbot(3716), sequoh(3703) 2010-10-10T09:08:29 *** hellman_ has joined #aichallenge 2010-10-10T09:09:50 *** hellman has quit IRC (Ping timeout: 240 seconds) 2010-10-10T09:35:07 *** tty1 has quit IRC (Read error: Connection reset by peer) 2010-10-10T09:35:47 *** tty1 has joined #aichallenge 2010-10-10T09:35:48 *** Migi32 has joined #aichallenge 2010-10-10T09:41:09 *** yasith has quit IRC (Ping timeout: 255 seconds) 2010-10-10T09:51:57 *** davidjliu has joined #aichallenge 2010-10-10T09:52:00 *** yasith has joined #aichallenge 2010-10-10T09:53:57 *** Appleman1234 has quit IRC (Ping timeout: 250 seconds) 2010-10-10T10:00:24 Top 10 players: rsergio(3845), felixcoto(3825), ofrias(3822), Hazard(3799), MariusCG(3781), cfaftw(3774), EBraun(3756), dmj111(3729), sequoh(3710), adumlah(3707) 2010-10-10T10:01:04 *** sigh has quit IRC (Remote host closed the connection) 2010-10-10T10:11:10 *** pdunstan has joined #aichallenge 2010-10-10T10:14:51 i want 4 cores intel and new power suply 2010-10-10T10:15:46 don't we all 2010-10-10T10:20:50 ? 2010-10-10T10:22:25 4 cores -only 4--1 want 16 + 32G of ram and 16T of hdd in a raid 2010-10-10T10:25:34 *** fawek has quit IRC (Remote host closed the connection) 2010-10-10T10:26:06 it make from my room sahara 2010-10-10T10:26:16 very hot plase 2010-10-10T10:26:50 *** pdunstan has left #aichallenge 2010-10-10T10:29:50 Ok, evolutionary algorithms DO pay off :D 2010-10-10T10:30:23 you got one working? 2010-10-10T10:31:06 It may not be what you expect. I call it auto-tune(tm). Look for "frontier" on the TCP server and compare my two entries. 2010-10-10T10:31:54 They both use the same code base. One is running with my guessed values in the formulas, the other is running with auto-tune(tm) values. 2010-10-10T10:32:55 right 2010-10-10T10:34:25 can i see some of these algorithms? 2010-10-10T10:35:55 epona: There is nothing to see ^^. Do you have some constants in your code? 2010-10-10T10:36:31 like planet_value = 3.4 * distance + pow(ships, 0.9) ? 2010-10-10T10:36:43 need 1 more mounth 2010-10-10T10:36:51 well, to be honest I kind of have a problem with the platform to run the bots 2010-10-10T10:36:54 to many worck 2010-10-10T10:37:03 i should probably be focusing on that 2010-10-10T10:38:37 These constants are hard to get right. You end up recompiling and testing for hours. So I made these constants variables that I can tune from outside the bot in a host application. There I automagically narrow down the optimal values for each constant and update the code once I got a good working set. 2010-10-10T10:42:13 Frontier> map releated 8-) 2010-10-10T10:42:22 heh 2010-10-10T10:42:31 http://72.44.46.68/canvas?game_id=408963 <-- wth? 2010-10-10T10:46:10 red mega failed on the first step 2010-10-10T10:46:40 *** perror has joined #aichallenge 2010-10-10T10:46:44 lol 2010-10-10T10:46:56 I was meaning the single planet with no growth factor in the middle :) 2010-10-10T10:47:05 woah... what? 2010-10-10T10:47:26 and then blue doesn't notice it 2010-10-10T10:47:40 massive error on the map maybe? 2010-10-10T10:48:17 Frontier, you're kicking my arse ... but I have a much simpler ga in play 2010-10-10T10:48:20 and the blue is me *hides* 2010-10-10T10:49:02 thank you for the compliment. what is your ga doing? 2010-10-10T10:50:24 Attacking a 0 growth planet with 115 ships really wasn't smart ^^. 2010-10-10T10:51:04 it's a very simple perceptron thing, which is being trained - probably very badly ... but I'm now running two bot to train them 2010-10-10T10:52:18 it isn't very bright yet :) 2010-10-10T10:53:50 http://72.44.46.68/getplayer?player=opless.2 ... very stupid by the looks of it 2010-10-10T10:54:50 *** wh1teside has joined #aichallenge 2010-10-10T11:00:24 Top 10 players: rsergio(3849), felixcoto(3820), ofrias(3814), Hazard(3803), cfaftw(3773), MariusCG(3761), EBraun(3758), dmj111(3743), sequoh(3727), adumlah(3708) 2010-10-10T11:03:15 I see. I have no idea how your bot is learning and modifying its code, but it sounds interesting. 2010-10-10T11:04:42 *** JensT1 has left #aichallenge 2010-10-10T11:05:34 *** boegel has joined #aichallenge 2010-10-10T11:09:33 oPless: i was considering trying out a neural network, but the complexity is rather high 2010-10-10T11:10:03 *** McLeopold has joined #aichallenge 2010-10-10T11:10:27 what software did you use, or are you doing it from scratch 2010-10-10T11:11:15 Frontier: there's a bunch of data exposed by the planetwars, and planet classes that the bot weights, and chooses a victim, then it does a similar calculation again but this time on how many ships to send 2010-10-10T11:11:41 it's really quite simple 2010-10-10T11:18:21 *** McLeopold has quit IRC (Quit: Leaving) 2010-10-10T11:30:50 would my bot be at a big disadvantage if it never attacked 0-growth planets at all? 2010-10-10T11:31:41 sure, 0-growth planets may give some positional advantage but I can't really think of an example where it'd be worth a lot of ships 2010-10-10T11:33:35 Frontier: you totally stole my idea. i was going to do that! 2010-10-10T11:33:48 Migi32: I'd say no, but my bot needs to improve a lot, so maybe also listen to others 2010-10-10T11:34:00 jmcarthur: what idea? 2010-10-10T11:34:14 Migi32: would it still attack if your opponent has ships there? 2010-10-10T11:34:33 boegel: tuning his bot parameters via an external evolutionary algorithm 2010-10-10T11:37:07 oPless: can we see your bot in action? 2010-10-10T11:37:24 jmcarthur: I had that in mind too :) 2010-10-10T11:37:44 jmcarthur: but I need to make it better and parametrize things first 2010-10-10T11:40:05 *** rebelxt_ has joined #aichallenge 2010-10-10T11:40:57 I keep converting constants to auto-tune(tm) parameters. Some don't have a big impact others give the new bot a slight advantage over the previous generation. 2010-10-10T11:41:07 epona: neural networks are usually just ensembles of generalized linear models that use the dependent variables of some glms as independent variables of others 2010-10-10T11:41:45 http://72.44.46.68/canvas?game_id=409936 2010-10-10T11:41:45 what are glms? 2010-10-10T11:41:54 generalized linear models 2010-10-10T11:44:20 *** ichti has joined #aichallenge 2010-10-10T11:44:33 alright, ill start playing around with more math related python scripts to brush up on this 2010-10-10T12:00:24 Top 10 players: rsergio(3852), Hazard(3820), felixcoto(3818), ofrias(3812), cfaftw(3783), dmj111(3768), EBraun(3760), sequoh(3737), adumlah(3711), MariusCG(3710) 2010-10-10T12:00:42 *** mega1 has quit IRC (Ping timeout: 240 seconds) 2010-10-10T12:04:44 *** boegel has quit IRC (Ping timeout: 245 seconds) 2010-10-10T12:06:41 *** Mathnerd314 has joined #aichallenge 2010-10-10T12:07:45 *** mega1 has joined #aichallenge 2010-10-10T12:10:14 what does the oppo. column mean on the tcp rankings? 2010-10-10T12:10:14 *** imaginative_name has joined #aichallenge 2010-10-10T12:14:17 *** deepblue has joined #aichallenge 2010-10-10T12:14:26 hi 2010-10-10T12:14:37 lesson learned today 2010-10-10T12:14:52 long int 32bit < long int 64bit 2010-10-10T12:15:06 local machine 64bit, official serves 32bit 2010-10-10T12:15:08 -_- 2010-10-10T12:16:19 nrub: oppo columns shows average opponent's rating, AFAIK 2010-10-10T12:17:03 *** boegel has joined #aichallenge 2010-10-10T12:17:48 deepblue: good to know. i'm on 64 bit but have not had any problems on the server probably due to using python 2010-10-10T12:18:00 aha 2010-10-10T12:18:23 yeah, I should switch to something with guaranteed data type sizes 2010-10-10T12:18:35 but I need a lot of cpu power so C it is 2010-10-10T12:18:44 * oPless leaves his stuff running in the hope it'll learn rapidly now :) 2010-10-10T12:20:35 deepblue: i'm now thinking of re-writing my bot in c++.. new python bot will likely time out on the server unless i bail out before calculations are finished 2010-10-10T12:21:16 will probably give you a few percent more performance 2010-10-10T12:21:37 i force my bot to be done under 800ms 2010-10-10T12:22:21 deepblue: are you working on new submission? our bots play regularly on the official server since our bots seem to hang out pretty close on the leaderboard 2010-10-10T12:23:23 just submitted a new version I worked on the last days after not working on the bot for a week 2010-10-10T12:23:31 all my submissions on the official server were buggy 2010-10-10T12:23:38 due to the long int issue :/ 2010-10-10T12:24:47 ahh 2010-10-10T12:25:26 lets hope it works this time ;) 2010-10-10T12:25:34 are you using the same algorithm that managed to beat Garry Kasparov? ;-) 2010-10-10T12:26:38 *** optimum has quit IRC (Remote host closed the connection) 2010-10-10T12:26:48 @predict 2010-10-10T12:26:48 I predict that Arthur will win! 2010-10-10T12:28:34 *** hellman_ has quit IRC (Remote host closed the connection) 2010-10-10T12:29:54 predict? 2010-10-10T12:30:25 deepblue: or maybe your nickname is not a reference to IBM's super computer? http://en.wikipedia.org/wiki/Deep_Blue 2010-10-10T12:30:36 well 2010-10-10T12:30:43 I'm just using a lot of cpu power 2010-10-10T12:31:03 and the nickname is from way back when I played counterstrike and starcraft 1 too much :D 2010-10-10T12:31:31 haha cool 2010-10-10T12:31:47 but not really brute forcing 2010-10-10T12:32:34 heuristics for the win 2010-10-10T12:36:08 *** fawek has joined #aichallenge 2010-10-10T12:37:14 *** espes has quit IRC (Ping timeout: 272 seconds) 2010-10-10T12:44:23 deepblue, can I ask what you're using all that cpu power for? 2010-10-10T12:44:35 i'm so happy, it finally does as it is supposed to do.. 2010-10-10T12:44:35 http://ai-contest.com/visualizer.php?game_id=5714285 2010-10-10T12:45:21 lol, my new bot is pretty hilarious http://ai-contest.com/visualizer.php?game_id=5713053 2010-10-10T12:45:25 in a bad way, methinks 2010-10-10T12:45:47 @Migi32: mostly trying to predict the future ;) 2010-10-10T12:45:48 deepblue: An error occured while trying to show the previous error. 2010-10-10T12:46:22 i'm not sure what contestbot is trying to tell me 2010-10-10T12:46:51 lots of action in your replay david 2010-10-10T12:47:45 lol, yeah... something's messed up. As in, everything 2010-10-10T12:48:03 maybe I should just switch back to my actually functional bot... 2010-10-10T12:48:42 this new bot seems to have made it to the top 50, so it looks pretty functional as well 2010-10-10T12:49:11 I mean, I guess it's functional, but it's just... weird 2010-10-10T12:50:01 it's a little aggressive, and is quick to abandon planets in exchange for capturing a planet of the opponent 2010-10-10T12:50:17 or that's at least how it looks to me 2010-10-10T12:50:47 yeah, this is very true. As in, it doesn't reserve any ships ever, unless being directly attacked 2010-10-10T12:57:25 davidjliu: yeah, a funny game ;) 2010-10-10T13:00:25 Top 10 players: rsergio(3861), felixcoto(3827), Hazard(3825), ofrias(3821), cfaftw(3792), EBraun(3778), dmj111(3774), sequoh(3748), MariusCG(3718), adumlah(3718) 2010-10-10T13:15:06 *** wvdschel has joined #aichallenge 2010-10-10T13:16:08 *** rebelxt_ has quit IRC (Quit: Page closed) 2010-10-10T13:21:42 *** perror has quit IRC (Quit: Bye all !) 2010-10-10T13:30:25 *** perror has joined #aichallenge 2010-10-10T13:42:11 !cpu 2010-10-10T13:42:11 Meatkat: I have taken 2065.75 seconds of user time and 66.72 seconds of system time, for a total of 2132.47 seconds of CPU time. My children have taken 0.00 seconds of user time and 0.02 seconds of system time for a total of 0.02 seconds of CPU time. I'm taking up 16296 kB of memory. 2010-10-10T13:52:33 *** rwa has joined #aichallenge 2010-10-10T13:53:37 *** Umsturz has joined #aichallenge 2010-10-10T13:53:57 *** bhasker has quit IRC (Quit: bhasker) 2010-10-10T13:54:40 hey guys, whats the adress of the private server? 2010-10-10T13:55:04 do you mean TCP server? 2010-10-10T13:55:09 yeah 2010-10-10T13:55:19 http://72.44.46.68/ 2010-10-10T13:55:47 cool thanks Naktibalda ! 2010-10-10T13:56:07 *** n9thbit has joined #aichallenge 2010-10-10T13:56:20 i hate it when i write code that is not obvious to understand later. i usually end up effectively rewriting it just to understand it later 2010-10-10T13:59:20 *** deepblue has quit IRC (Ping timeout: 265 seconds) 2010-10-10T14:00:26 Top 10 players: rsergio(3842), ofrias(3822), felixcoto(3815), Hazard(3805), cfaftw(3795), dmj111(3786), EBraun(3767), sequoh(3738), MariusCG(3726), adumlah(3712) 2010-10-10T14:01:12 *** Mathnerd314 has quit IRC (Quit: ChatZilla 0.9.86-rdmsoft [XULRunner 1.9.2.8/20100722155716]) 2010-10-10T14:07:46 so tcp.c should compile without anything else? 2010-10-10T14:08:59 gcc gives me a undefined symbols error... 2010-10-10T14:09:16 *** wvdschel has quit IRC (Quit: Leaving) 2010-10-10T14:18:45 jmcarthur, how about comments? 2010-10-10T14:20:00 Migi32: comments are a poor subsitute for good code and are easy to make out of date 2010-10-10T14:20:12 real programmer doesn't write comments 2010-10-10T14:20:18 i wouldn't say that 2010-10-10T14:20:22 but i do write very few comments 2010-10-10T14:20:31 aside from library documentation 2010-10-10T14:20:42 in C++ I couldn't survive without comments. C++ is great but not very readable 2010-10-10T14:21:01 i don't think C++ is that hard to read 2010-10-10T14:21:04 i dislike it for other reasons 2010-10-10T14:21:17 Real programmers don't comment their code. If it was hard to write, it should be hard to read. 2010-10-10T14:21:22 lol 2010-10-10T14:22:11 jmcarthur, compare "foreach (i in array)" to "for (std::vector::const_iterator iter = array.begin(); iter != array.end(); ++iter)" 2010-10-10T14:22:44 that's not a syntactic issue. that's a semantic issue 2010-10-10T14:22:56 if C++ semantics allowed for higher level loops then that crazy iterator wouldn't have to exist 2010-10-10T14:23:05 they do 2010-10-10T14:23:12 *** aerique has joined #aichallenge 2010-10-10T14:23:20 and even harder to read are the errors gcc gives you when you mess up something involving any std type 2010-10-10T14:23:22 #define FOREACH(T, iter, array) 2010-10-10T14:23:23 :P 2010-10-10T14:23:23 Zannick: An error occured while trying to show the previous error. 2010-10-10T14:23:30 Zannick: ew :P 2010-10-10T14:23:32 oh, go to hell contestbot 2010-10-10T14:23:48 yeah g++ errors suck 2010-10-10T14:24:03 i'm using c, and i love macros 2010-10-10T14:24:15 i have a very nice FOREACH setup for my circular buffers 2010-10-10T14:24:18 i'm using the C preprocessor in my haskell code 2010-10-10T14:24:44 but not for things like that 2010-10-10T14:24:49 just conditional compilation 2010-10-10T14:25:03 test code, specifically 2010-10-10T14:25:48 *** tapwater has joined #aichallenge 2010-10-10T14:27:30 well the code for my bot is still small enough to fit in my brain (which isn't very big) so I don't yet have to worry about readability too much 2010-10-10T14:27:37 *** smellyhippy has quit IRC (Ping timeout: 252 seconds) 2010-10-10T14:28:49 my main problem is just that i'm not creating enough functions 2010-10-10T14:29:29 even a five or six line function in haskell has enough information density to be a little hard to read 2010-10-10T14:38:36 *** Bobng has joined #aichallenge 2010-10-10T14:38:53 *** aerique has quit IRC (Quit: ...) 2010-10-10T14:39:06 *** bhasker_ has joined #aichallenge 2010-10-10T14:41:37 *** Blkt has joined #aichallenge 2010-10-10T14:43:40 mcleopold there? 2010-10-10T14:47:10 *** Bobng has quit IRC (Ping timeout: 265 seconds) 2010-10-10T14:49:41 *** boegel has quit IRC (Quit: *poof!*) 2010-10-10T14:51:53 *** wvdschel has joined #aichallenge 2010-10-10T14:53:55 *** perror has quit IRC (Quit: Bye all !) 2010-10-10T14:56:02 *** amstan has quit IRC (Remote host closed the connection) 2010-10-10T14:57:16 *** Bobng has joined #aichallenge 2010-10-10T14:58:48 *** ShaneHudson has joined #aichallenge 2010-10-10T15:00:27 Top 10 players: rsergio(3834), felixcoto(3819), Hazard(3817), ofrias(3815), cfaftw(3800), dmj111(3773), MariusCG(3759), EBraun(3751), sequoh(3730), adumlah(3697) 2010-10-10T15:00:58 *** wvdschel has quit IRC (Quit: Leaving) 2010-10-10T15:03:29 opless? 2010-10-10T15:05:18 topless? 2010-10-10T15:06:38 jalopyless 2010-10-10T15:07:02 http://72.44.46.68/canvas?game_id=413148 2010-10-10T15:07:04 opless^ 2010-10-10T15:07:08 does nothing bot 2010-10-10T15:15:46 *** Umsturz has quit IRC (Quit: Umsturz) 2010-10-10T15:16:14 *** boegel has joined #aichallenge 2010-10-10T15:20:32 bhasker_, why is your bot reinforcing the middle planet and not the "battlefront"? 2010-10-10T15:20:44 averages 2010-10-10T15:20:50 and i think i have a bug in my staging code 2010-10-10T15:21:07 i stage to the planets at best average distance between mine and his 2010-10-10T15:21:13 so sometimes taging can be behind the front 2010-10-10T15:23:49 *** Acount has quit IRC () 2010-10-10T15:29:00 *** jesionaj` has joined #aichallenge 2010-10-10T15:30:53 *** jesionaj has quit IRC (Ping timeout: 252 seconds) 2010-10-10T15:33:20 *** bhasker_ has quit IRC (Quit: bhasker_) 2010-10-10T15:35:48 *** Acount has joined #aichallenge 2010-10-10T15:38:23 *** Itkovian has joined #aichallenge 2010-10-10T15:38:42 *** optimum has joined #aichallenge 2010-10-10T15:56:47 *** n9thbit has quit IRC (Ping timeout: 265 seconds) 2010-10-10T16:00:27 Top 10 players: rsergio(3827), ofrias(3826), Hazard(3822), felixcoto(3816), cfaftw(3814), dmj111(3795), EBraun(3760), sequoh(3745), MariusCG(3741), adumlah(3698) 2010-10-10T16:03:35 *** Bobng has joined #aichallenge 2010-10-10T16:05:46 *** jesionaj` is now known as jesionaj 2010-10-10T16:09:47 *** Bobng has quit IRC (Quit: Leaving) 2010-10-10T16:11:12 *** Bobng has joined #aichallenge 2010-10-10T16:11:20 *** Appleman1234 has joined #aichallenge 2010-10-10T16:12:30 *** Bobng has quit IRC (Client Quit) 2010-10-10T16:12:54 *** Bobng has joined #aichallenge 2010-10-10T16:12:55 *** delt0r_ has joined #aichallenge 2010-10-10T16:14:28 *** delt0r___ has quit IRC (Ping timeout: 265 seconds) 2010-10-10T16:30:06 *** boegel has quit IRC (Quit: *poof!*) 2010-10-10T16:30:27 *** sigh has joined #aichallenge 2010-10-10T16:31:37 *** HakanD has joined #aichallenge 2010-10-10T16:34:30 *** Bobng has quit IRC (Quit: Leaving) 2010-10-10T16:39:21 *** Bobng has joined #aichallenge 2010-10-10T16:40:25 *** Palmik has quit IRC (Remote host closed the connection) 2010-10-10T16:41:39 *** Bobng has joined #aichallenge 2010-10-10T16:42:58 @seen j3camero 2010-10-10T16:42:58 Bobng: j3camero was last seen in #aichallenge 5 days, 18 hours, 23 minutes, and 12 seconds ago: amstan: yes, but no names, haha. 2010-10-10T16:43:03 D: 2010-10-10T16:43:09 *** Bobng has quit IRC (Client Quit) 2010-10-10T16:46:36 :( 2010-10-10T16:49:44 this guy is submitting every few minutes: http://ai-contest.com/profile.php?user_id=10080 2010-10-10T16:49:56 does he want to DOS a server again? :) 2010-10-10T16:50:46 *** Bobng has joined #aichallenge 2010-10-10T16:51:40 *** Bobng has quit IRC (Client Quit) 2010-10-10T16:52:57 *** bhasker has joined #aichallenge 2010-10-10T16:53:38 on both C++ and python too :) 2010-10-10T16:55:52 *** sigh has quit IRC (Quit: Leaving...) 2010-10-10T16:58:05 I wonder how many turns the longest round had 2010-10-10T17:00:32 Top 10 players: rsergio(3857), Hazard(3854), ofrias(3851), felixcoto(3833), cfaftw(3832), dmj111(3799), sequoh(3783), EBraun(3781), MariusCG(3761), goldman(3729) 2010-10-10T17:01:03 *** amstan has joined #aichallenge 2010-10-10T17:01:03 *** ChanServ sets mode: +o amstan 2010-10-10T17:04:28 wow 2010-10-10T17:05:10 someone from india 2010-10-10T17:06:08 maybe he's testing the available space on the machine :P 2010-10-10T17:06:35 his profile says "I want to learn AI" 2010-10-10T17:06:50 *** yasith has quit IRC (Ping timeout: 276 seconds) 2010-10-10T17:07:03 *** Blkt has quit IRC (Ping timeout: 272 seconds) 2010-10-10T17:07:47 any admins around? http://ai-contest.com/forum/viewtopic.php?f=18&t=911 2010-10-10T17:08:52 *** HakanD has quit IRC (Ping timeout: 240 seconds) 2010-10-10T17:09:14 the server has been really weird of late 2010-10-10T17:09:21 http://ai-contest.com/profile.php?user_id=7530 2010-10-10T17:09:32 this exact code used to be ranked 20th 2010-10-10T17:09:47 now most of the time i "lose" first turn 2010-10-10T17:10:11 its the mutating algorithms 2010-10-10T17:10:22 pjreddie: it says the other playing didn't start.. ? 2010-10-10T17:10:55 pjreddie: do you have a watch on clock time? Maybe we're getting less cycles now 2010-10-10T17:11:00 half the time it says :pjreddie crashed / did not start / timeout 2010-10-10T17:11:09 the other half it says the other person 2010-10-10T17:11:29 from what i can tell it's always the person who is red thta "crashes" 2010-10-10T17:11:39 the error message always says player one 2010-10-10T17:11:41 it's a known bug 2010-10-10T17:11:47 ah 2010-10-10T17:12:01 does it say lose or win on your profile? 2010-10-10T17:12:07 lose 2010-10-10T17:12:15 then it's you :P 2010-10-10T17:13:04 yeah but i donow why, it's literally the exact same code 2010-10-10T17:13:10 maybe we are getting less cycles 2010-10-10T17:13:21 hmm. pretty sure my error is something different. havent been able to run my bot since day 1 2010-10-10T17:13:23 but it would have to be a lot less, my code is pretty simple 2010-10-10T17:16:54 ya... i really don't think i'm timing out. i just did a little experiment, ran it with smaller and smaller time limit, i don't time out until i get below 50 ms for time limit 2010-10-10T17:17:10 what language? 2010-10-10T17:17:12 and server time limit is supposed to be 1000ms right 2010-10-10T17:17:14 c++ 2010-10-10T17:17:45 sorry, no idea 2010-10-10T17:18:17 is there anywhere we can dl new maps? 2010-10-10T17:19:09 this is way to frustrating 2010-10-10T17:20:30 *** yasith has joined #aichallenge 2010-10-10T17:21:39 can someone explain to me how to get my code to run in a map so I can see it? 2010-10-10T17:21:45 in action i mean 2010-10-10T17:22:20 what language? 2010-10-10T17:22:35 python 2010-10-10T17:23:07 does the standard command not work? 2010-10-10T17:23:19 java -r PlayGame.jar etc. 2010-10-10T17:23:25 er, -jar 2010-10-10T17:23:32 *** Bobng has joined #aichallenge 2010-10-10T17:23:47 ahhh, your saying I should boot up linux and do it? 2010-10-10T17:23:48 *** krokokrusa has quit IRC (Quit: Don't panic!) 2010-10-10T17:24:00 should work in windows, too 2010-10-10T17:24:14 everytime I click on it nothing happens 2010-10-10T17:24:17 up to you 2010-10-10T17:24:22 yeah, you need a command line 2010-10-10T17:24:34 I c 2010-10-10T17:24:38 thankyou 2010-10-10T17:25:09 *** Bobng has quit IRC (Read error: Connection reset by peer) 2010-10-10T17:25:32 *** Bobng has joined #aichallenge 2010-10-10T17:25:41 java -jar tools/PlayGame.jar maps/map1.txt 1000 200 log2.txt "python MyBot.py" "java -jar example_bots/BullyBot.jar" | java -jar tools/ShowGame.jar 2010-10-10T17:25:48 for example 2010-10-10T17:26:17 *** Bobng has quit IRC (Client Quit) 2010-10-10T17:32:42 *** Bobng has joined #aichallenge 2010-10-10T17:38:57 *** rebelxt_ has joined #aichallenge 2010-10-10T17:41:53 *** mega1 has quit IRC (Ping timeout: 272 seconds) 2010-10-10T17:42:32 *** Bobng has quit IRC (Quit: Leaving) 2010-10-10T17:43:23 *** Bobng has joined #aichallenge 2010-10-10T17:43:36 *** espes has joined #aichallenge 2010-10-10T17:46:16 *** yasith has quit IRC (Ping timeout: 240 seconds) 2010-10-10T17:56:22 *** BtbN has quit IRC (Quit: Verlassend) 2010-10-10T17:57:39 *** jesionaj` has joined #aichallenge 2010-10-10T18:00:27 Top 10 players: ofrias(3839), Hazard(3836), rsergio(3832), felixcoto(3820), cfaftw(3812), dmj111(3797), sequoh(3777), EBraun(3768), MariusCG(3726), goldman(3715) 2010-10-10T18:01:00 *** jesionaj has quit IRC (Ping timeout: 245 seconds) 2010-10-10T18:04:50 *** yasith has joined #aichallenge 2010-10-10T18:05:36 *** oPless has quit IRC (Quit: This computer has gone to sleep) 2010-10-10T18:08:00 *** sigh has joined #aichallenge 2010-10-10T18:15:47 *** Bobng has left #aichallenge ("Leaving") 2010-10-10T18:16:17 *** Bobng has joined #aichallenge 2010-10-10T18:17:11 *** Bobng has joined #aichallenge 2010-10-10T18:18:24 does anyone have a modified script that will replay a server game and let you visualize it locally? 2010-10-10T18:18:31 like if i want to make changes to my bots and see the game play out 2010-10-10T18:22:04 not sure how that would work once your bot makes a different move than was played in the game 2010-10-10T18:22:12 bhasker: you can get the playback string, save it to a file "replay" or something, then do something like "$ replay | java -jar tools/ShowGame.jar" 2010-10-10T18:22:13 I think 2010-10-10T18:22:36 hmm 2010-10-10T18:23:16 but yeah, that's only for replaying a server game 2010-10-10T18:23:50 there's a script somewhere in the forums, i seem to recall 2010-10-10T18:24:19 but I wouldn't recommend inserting different orders while keeping the enemy's orders the same, that makes no sense 2010-10-10T18:25:13 it's like "ok I know the opponent played scissors, let's see if I can change my strategy to beat that". "Hey, new strategy: I should always play rock!" 2010-10-10T18:25:48 parse_game_state_url.py, my logs say 2010-10-10T18:26:10 that is weird, my bot just lost after 1001 turns twice lol 2010-10-10T18:26:17 yeah, i think the point is to replay the game so you can see what your bot is doing that you can't 2010-10-10T18:26:27 that you can't on the server 2010-10-10T18:27:25 *** Bobng has quit IRC (Ping timeout: 264 seconds) 2010-10-10T18:27:36 ooo I just watched the replay, at one point mine just stopped got like 4000 fleets lol 2010-10-10T18:28:01 ah k 2010-10-10T18:29:25 Migi32: I think it's more like, on turn x, my bot issues a stupid order... let's try to change the code so that it doesn't 2010-10-10T18:29:30 bhasker: if you want to parse and replay the whole game locally, you could use http://github.com/apinkin/planetwars-python-kit/blob/master/parse_game_state_url.py (based on dmj's script) 2010-10-10T18:29:39 if you are only looking at one state, then it makes perfect sense 2010-10-10T18:29:40 ah k thanks 2010-10-10T18:29:48 yea pretty much what i want to do 2010-10-10T18:29:52 @sigh 2010-10-10T18:29:52 i know... 2010-10-10T18:29:57 my bot is doing some stupid stuff 2010-10-10T18:30:24 *** ShaneHudson has quit IRC (Quit: leaving) 2010-10-10T18:30:37 bhasker: or use bookmarklet if you just need to analyze a single move 2010-10-10T18:30:46 hmm k 2010-10-10T18:31:20 lol, @sigh is a special opcode for contestbot 2010-10-10T18:31:31 yup 2010-10-10T18:31:51 bhasker: http://ai-contest.com/forum/viewtopic.php?f=18&t=790&start=0&hilit=bookmarklet - this is great 2010-10-10T18:32:02 ah k thanks 2010-10-10T18:32:27 np man 2010-10-10T18:33:25 *** jaspervdj has quit IRC (Quit: NEVER GONNA GIVE YOU UP NEVER GONNA LET YOU DOWN) 2010-10-10T18:33:59 goddamit jaspervdj 2010-10-10T18:34:54 rebelxt_, thanks for linking that. It is indeed awesome. 2010-10-10T18:38:28 yuw Migi32, it was nice of iouri_ to share his code. best tool for debugging and improving algorithm hands down 2010-10-10T18:40:21 yep 2010-10-10T18:40:38 interesting to see how fast bots are improving.. if i don't touch my bot for a few days, its rank declines significantly within just a few days ;( it's arms race between algorithms ... ;) 2010-10-10T18:41:50 yeah... it's very competitive 2010-10-10T18:42:28 yeah 2010-10-10T18:42:30 if, when i finally finish my bot, it doesn't do well, i might not continue :P 2010-10-10T18:42:37 *** sigh has quit IRC (Remote host closed the connection) 2010-10-10T18:42:47 *** sigh has joined #aichallenge 2010-10-10T18:43:05 *** fawek has quit IRC (Ping timeout: 245 seconds) 2010-10-10T18:43:09 Zannick, if your *first* submission of a bot does better than me, I might give up >_< 2010-10-10T18:43:16 *** yasith has quit IRC (Ping timeout: 260 seconds) 2010-10-10T18:43:23 my *first* submission was copybot :P 2010-10-10T18:43:32 that doesn't count :P 2010-10-10T18:43:33 some 38 days ago 2010-10-10T18:47:10 *** Migi32 has quit IRC (Quit: http://boinc.berkeley.edu) 2010-10-10T18:49:05 it also feels like small never-ending incremental tweaks are not always the way to go.. i've gotten most success from just dropping most of my code (except for utilities) and re-writing main logic from scratch. did any of you guys have the same experience? 2010-10-10T18:49:37 not yet :) 2010-10-10T18:50:00 rebelxt_: all of last week was spent refactoring my entire framework 2010-10-10T18:50:02 so yes :P 2010-10-10T18:50:33 even with all that it only got me into the top 10 for a couple hours, and now I've been booted out again :( 2010-10-10T18:50:49 *** Tankh has joined #aichallenge 2010-10-10T18:52:22 sigh, haha same here! i got 11th spot two days ago after complete re-write of the main logic.. but could never crack top10 though.. now down to 30th.. 2010-10-10T18:52:48 sigh, we seem to fight quite a bit on the tcp server. good games! 2010-10-10T18:52:53 :) 2010-10-10T18:55:06 I just hate it when you think you have a great idea, implement it and test it and end up losing more games than winning against the old version of the bot... 2010-10-10T18:55:27 rebelxt_: hang on, I haven't had my bot on tcp for a while now 2010-10-10T18:55:33 not sure who you are fighting :P 2010-10-10T18:56:15 haha, maybe not recently but i recall some games on both official and tcp servers 2010-10-10T18:56:31 *** yasith has joined #aichallenge 2010-10-10T18:56:36 yeah, I do too :) 2010-10-10T18:56:58 Tankh: happens to all of us 2010-10-10T18:57:17 Tankh: yeah this happens to everyone probably 2010-10-10T18:58:47 tried to improve the supply line to send the ships in small steps instead of the whole way at once 2010-10-10T18:59:18 I found direct works better for me too 2010-10-10T18:59:44 ships get to the front quicker 2010-10-10T18:59:49 i still think its better to go small steps, as long as it's not too much zig zag going on 2010-10-10T19:00:19 but even on a map where all my planets were in a straight line i lost, and that part confused me alot 2010-10-10T19:00:23 I've tried many variant, but couldn't get anything better 2010-10-10T19:00:28 Top 10 players: Hazard(3845), rsergio(3842), ofrias(3833), felixcoto(3826), cfaftw(3817), dmj111(3806), EBraun(3782), sequoh(3778), MariusCG(3721), goldman(3711) 2010-10-10T19:00:36 *variants 2010-10-10T19:00:46 hm 2010-10-10T19:01:05 cause it helps my defend planet algorithm that more than one planet has a constant flow of ships 2010-10-10T19:01:23 but again, it has to work well with your bot's overall strategy 2010-10-10T19:01:37 my redistribution doesn't necessarily go just to one planet 2010-10-10T19:01:46 just to planets at the "front" 2010-10-10T19:01:59 yeah i used to just send to 1 single planet 2010-10-10T19:02:09 the one with most enemy ships nearby 2010-10-10T19:02:52 the new algorithm looked awesome on map 1, cause there are 2 fronts kinda, and they flanked the enemy from both sides :P 2010-10-10T19:03:04 yeah, that's what my bot does 2010-10-10T19:04:01 my bot doesn't have an global redistribution strategy, each planet decides independently who it wants to send its ships 2010-10-10T19:04:29 yeah I took that approach aswell 2010-10-10T19:04:45 and has to request for help 2010-10-10T19:04:53 and then other planets look for planets that need help 2010-10-10T19:05:01 which works most of the time at least 2010-10-10T19:05:17 I just know that planets need help if it is going to be taken over :P 2010-10-10T19:05:27 they don't need to ask :D 2010-10-10T19:06:29 those little buggers just wants a life of their own so i thought i might let them control the help part themselves :P 2010-10-10T19:06:37 but I treat help as part of the attack phase, I "attack" planets that are not owned by me in the far future 2010-10-10T19:07:09 the redistribution is just "I've got left over ships, lets send them somewhere useful" 2010-10-10T19:07:21 its the very last step of the strategy 2010-10-10T19:08:59 yea same here, first all planets check if they need help, then check if someone else needs help. then attack different planets, and lastly if they dont have enough ships to attack or are in range to help planets they just send support 2010-10-10T19:09:22 well thats the idea. sometimes it does stuff i didn't intend :P 2010-10-10T19:09:35 :) 2010-10-10T19:09:37 like sending support to front line when a planet next to them needs help 2010-10-10T19:09:48 sometimes they do sometimes they don't. not sure what's going on there 2010-10-10T19:10:37 *** imaginative_name has quit IRC (Ping timeout: 265 seconds) 2010-10-10T19:15:14 *** yasith has quit IRC (Ping timeout: 265 seconds) 2010-10-10T19:18:44 Tankh: haha same issue here. also experimented with a lot of different "ship supply chain" algorithms and had similar experience to sigh's and yours. so no longer sending ships through intermediate planets.. 2010-10-10T19:19:47 aww damnit 2010-10-10T19:19:51 i thought it looked so cool 2010-10-10T19:20:07 I remember back when just having any supply chain got you into the top 10 :P 2010-10-10T19:20:25 Tankh: however, i have a lot of experience playing Galcon, and it's common practice among top level players to supply ships through intermediate hops.. i.e. no long flights that put you at risk of getting attacked when too many of your ships are in flight 2010-10-10T19:21:28 i dont have much galcon experience.. i had played it like 50 games half a year ago so i knew what the game was about 2010-10-10T19:21:35 I think hopping might be better as well, but I think direct exploits a weakness that exists in many of the current bots (even top ones) 2010-10-10T19:21:41 also did the challenges on www.galcon.com/flash 2010-10-10T19:21:48 just a guess though :P 2010-10-10T19:22:17 imho, the goal should be to send supply ships as far as possible towards closest "front" planet without putting any of your planets at risk 2010-10-10T19:23:15 sigh, you must be right! i'm also keeping hopping algorithm just in case we will need it later when bots get even smarter ! ;-) 2010-10-10T19:23:28 haha 2010-10-10T19:23:50 good thing i did a backup on my old straight line algorithm then 2010-10-10T19:24:04 yeah, most of my variants are just different config options to my bot 2010-10-10T19:24:55 im also wondering about the thing some bots do, where they just sit and wait when they know they have more ships and production than the enemy 2010-10-10T19:25:10 and just act defensively when they get attacked 2010-10-10T19:26:24 I tried to make my bot change strategy based on if it was ahead or behind 2010-10-10T19:26:26 fialed 2010-10-10T19:26:28 *failed 2010-10-10T19:27:15 Tankh: yep. my bot does this as a side effect of the way scoring function works. - i.e. scoring function calculates profit from attacking neutral and enemy planets does not return anything worthy of attack 2010-10-10T19:27:43 right 2010-10-10T19:28:03 as long as any planet has ships enough to capture a planet, it will try 2010-10-10T19:28:34 always aiming at the best valued one, but best doesnt mean good. it just takes the best that is available at the time 2010-10-10T19:29:10 also hard to tweak distance penalty 2010-10-10T19:29:12 here is what human players seem to do in Galcon when they are just a bit ahead on production and ship count - they start putting minimal pressure (small attacks) on enemy to make sure he is not able to expand and get even/ahead 2010-10-10T19:30:10 Tankh: man, I know exactly what you mean about distance penalty 2010-10-10T19:30:11 sounds smart 2010-10-10T19:30:38 I have quite a few hacks so that my bot doesn't attacks things at a stupid distance 2010-10-10T19:30:44 right now i think i add a penalty of twice or three times the distance to their ship count 2010-10-10T19:31:28 sigh. yeah i hate these hacks in my code... looking for a way to avoid one-offs and it's hard 2010-10-10T19:31:30 wonder if it's worth to try something other than a linear increase of penalty according to distance 2010-10-10T19:31:48 Tankh: yep. there is a way to build a scoring functions which takes "everything" into account (cost of conquering, distance, growth, planet's neighbour score, etc) 2010-10-10T19:31:50 rebelxt_: yeah, I agree 2010-10-10T19:32:12 yeah of course you can always make it more complex :P 2010-10-10T19:32:50 in the end you might have improved your bot with 0.1% but its performance varies +/- 50% depending on which map it is 2010-10-10T19:33:05 *** yasith has joined #aichallenge 2010-10-10T19:34:10 I'm having great fun with this contest though. There are some plans on putting up a local server for our school to motivate some programming for new students, which would be awesome 2010-10-10T19:34:18 :) 2010-10-10T19:34:41 I've gone through several iterations of layering little hacks on one after the other, then I realize what I really wanted to do could be put into one fundamental framework, then I remove all the little hacks... 2010-10-10T19:35:13 but can be a bit frustrating some times.. failed two times with the "helping planets" algorithm. then stopped for a week and got it finished yesterday 2010-10-10T19:35:38 I'm only now getting close to my algorithm doing what I thought it was doing... getting my debugging tools in order 2010-10-10T19:36:15 Tankh: nice. i have a private tcp server, and it helps a lot.. running 5-6 of my bot's versions on it 24x7. this makes testing new bots much easier and more reliable than just benchmarking against one previous version.. 2010-10-10T19:36:17 ah yeah debugging.. first i tried putting crazy "issueOrder" commands on various places just to see when/if it would crash :P 2010-10-10T19:36:51 yeah I think one of us has a way to simulate combat on all 100 maps, but haven't got hold of it yet 2010-10-10T19:37:25 haha 2010-10-10T19:37:35 get proper logging :P 2010-10-10T19:37:38 Tankh: i have scripts for all of that ;-) let me get a link 2010-10-10T19:39:10 Tankh: http://github.com/apinkin/planetwars-python-kit (see README and shell scripts). are you using unix by any chance? 2010-10-10T19:39:39 oh that for python? 2010-10-10T19:40:30 using win 7 2010-10-10T19:40:32 Tankh: yep but you can still re-use batch scripts (there is one for running a tournament against example bots on all maps, another script to run against tcp server, a script to parse a server game and replay it locally, etc) 2010-10-10T19:40:50 Tankh: that's fine. you can use Cygwin to run shell scripts on Win7 2010-10-10T19:41:25 ok, thanks for the link man 2010-10-10T19:41:31 yeah, I have python scripts to do pretty much the same things 2010-10-10T19:41:38 hopefully testing bots will be easier now 2010-10-10T19:41:54 sigh: nice! ;-) 2010-10-10T19:42:15 Tankh: cool hope this helps 2010-10-10T19:42:55 my tournament script stores the results in an sqlite db, if I ask it to run a match it's already played it outputs the stored result :) 2010-10-10T19:43:02 well i gotta go now but ill idle in the channel just for funsies. 2010-10-10T19:43:15 c ya 2010-10-10T19:43:18 bye 2010-10-10T19:43:29 *** yasith has quit IRC (Ping timeout: 272 seconds) 2010-10-10T19:43:38 sigh, sweet i did not go that far 2010-10-10T19:44:29 it just means that I can quickly run large round robins and only the games for the new bot will actually be played 2010-10-10T19:44:31 sigh: i did set up a private tcp server - http://ec2-72-44-46-51.compute-1.amazonaws.com/planetwars/getratings - just to have bots running 24x7 lol ;-) 2010-10-10T19:44:46 yeah, I think that goes a bit further :P 2010-10-10T19:45:03 yikes... I'm having all my bots run in a single process a la my bastardized monolithic combination c++ engine/bot1,...botN/logger 2010-10-10T19:45:18 but really with my CPU horsepower I can afford to do nothing less 2010-10-10T19:47:12 wh1teside: 24 cores? ;) someone posted on the forums he bought a 24-core system just for this contest! 2010-10-10T19:47:54 o_O 2010-10-10T19:48:07 D: hahaha I'm fearing all over that 2010-10-10T19:48:33 hahaha 2010-10-10T19:48:45 yeah no just the one core for me 2010-10-10T19:48:46 unless it was a server bought just for this contest that's hilarious 2010-10-10T19:49:18 * sigh suddenly feels inadequate with his 4 cores 2010-10-10T19:49:22 http://ai-contest.com/forum/viewtopic.php?f=17&t=486&hilit=genetic&start=10#p4468 - look at this! 2010-10-10T19:49:36 my laptop is 4 years old and has 2 cores 2010-10-10T19:50:03 my other computer has 4 cores but it's gaming and such and not programming so :P 2010-10-10T19:50:13 Zannick: same as mine 2010-10-10T19:50:24 I ssh to my desktop so that I can work from my laptop 2010-10-10T19:50:35 ah, i don't have sshd on my desktop 2010-10-10T19:50:40 so i'm only on my laptop 2010-10-10T19:50:59 then I use the canvas visualiser and serve the webpage so taht I can view on my laptop -_- 2010-10-10T19:51:12 heh, nice 2010-10-10T19:51:28 i used to have a server before i moved 2010-10-10T19:51:47 :) 2010-10-10T19:51:48 i would totally do that, but eh 2010-10-10T19:52:17 well, i am up to 12 source files now 2010-10-10T19:53:10 i was bored and fully automated bot testing ;-) i.e. just fto or scp bot file to the tcp server, and server automatically detects it and starts playing never-ending tournament against other bots. this might have been an overkill though ;-) 2010-10-10T19:53:27 meant ftp or scp 2010-10-10T19:53:45 I wanted to set something like that up locally, but was lazy 2010-10-10T19:53:50 *** ichti has quit IRC (Ping timeout: 264 seconds) 2010-10-10T19:55:11 sigh: yeah it took half a day to set this up :( 2010-10-10T19:55:26 half a day is not bad 2010-10-10T19:55:43 12 source files? Dang, I'm still in 1. 2010-10-10T19:56:16 well, 7 .h files and 5 .c files 2010-10-10T19:56:38 oh in that case, I've got more than Zannick 2010-10-10T19:56:41 it helps me keep components separate 2010-10-10T19:57:47 yeah, the classes are in different files, and different parts of my strategy are in different files 2010-10-10T19:58:17 Some time in the next five minutes you should receive an email letting you know whether or not we successfully compiled your code. 2010-10-10T19:58:32 why does this text have a line through it on their page? 2010-10-10T19:58:40 were they too lazy to delete or what 2010-10-10T19:58:46 *** yasith has joined #aichallenge 2010-10-10T19:58:50 it means that it might be borken 2010-10-10T19:59:10 so i dun goofed? 2010-10-10T19:59:18 epona, since there is only 25% chance you will get an e-mail ;-) 2010-10-10T19:59:31 ah 2010-10-10T19:59:36 epona: no, server might be broken 2010-10-10T19:59:39 not you :P 2010-10-10T20:00:15 yeah their e-mail server or some other component is still buggy. 2010-10-10T20:00:28 Top 10 players: Hazard(3829), ofrias(3821), rsergio(3817), felixcoto(3816), cfaftw(3800), dmj111(3795), sequoh(3769), EBraun(3767), MariusCG(3720), _iouri_(3707) 2010-10-10T20:01:25 I haven't seen anyone break 3900 2010-10-10T20:01:51 top was above 3900 for a bit last week or so 2010-10-10T20:01:58 rebelxt_: I think your dedicated server is smart. I think top 10 will be decided by small edges, and you'll need big sample sizes to "see" those edges and not just be looking at variance. 2010-10-10T20:02:13 but the top bots keep beating each other and dragged it down 2010-10-10T20:03:17 heeheehee 2010-10-10T20:03:26 defense(budget); 2010-10-10T20:03:34 rwa, agreed it's going to be a very close contest.. dhartmei open sourced his tcp server code, so anyone can set up a private server 2010-10-10T20:03:36 heh 2010-10-10T20:04:22 Zannick: I assume that method handles all the invasions? 2010-10-10T20:04:39 something like that 2010-10-10T20:04:43 i haven't written it yet. 2010-10-10T20:04:47 :P 2010-10-10T20:04:51 but i know what arguments it takes :) 2010-10-10T20:05:07 sigh: precisely; and I was the only one who did the honorable thing and stopped beating the other top bots so as not to uhhh bog the rest of them down 2010-10-10T20:05:43 :) 2010-10-10T20:06:07 whoops 2010-10-10T20:06:08 13 files 2010-10-10T20:06:12 i missed a .h file 2010-10-10T20:07:19 wh1teside: lol 2010-10-10T20:08:08 wh1teside: my bot decided to go on a losing streak too just to help someone get ELO record ;--) 2010-10-10T20:14:55 I wonder what their looking for 2010-10-10T20:15:09 google, i mean 2010-10-10T20:15:31 this isn't google-run 2010-10-10T20:15:41 it just has google's name on it 2010-10-10T20:16:50 ahh 2010-10-10T20:20:47 hmm, just noticed after replying to a post on the forum that network communication isn't specifically disallowed on the rules page 2010-10-10T20:20:54 looks like we need an update 2010-10-10T20:25:03 *** rebelxt_ has quit IRC (Ping timeout: 265 seconds) 2010-10-10T20:28:54 my bot sometimes does this weird thing where it gets leisurely about finishing off an opponent that it has vastly outnumbered 2010-10-10T20:29:08 it's like a cat toying with a half-dead mouse 2010-10-10T20:30:33 thats fine 2010-10-10T20:30:43 kind of classy, actually 2010-10-10T20:31:21 but you could make an if statement shifting its tactics once youve outnumbered them 2010-10-10T20:31:23 probably better than the couple of bots i've seen shooting off fireworks :P 2010-10-10T20:33:05 fireworks are beautiful 2010-10-10T20:35:16 *** evilbotzz has joined #aichallenge 2010-10-10T20:35:22 fireworks would be far more effective at ending the game 2010-10-10T20:35:53 if only they had the MD device in Enders game as like a final weapon 2010-10-10T20:36:09 toying with my opponent: http://ai-contest.com/visualizer.php?game_id=5041300 2010-10-10T20:36:10 ragebot would like that 2010-10-10T20:36:16 boom 2010-10-10T20:36:16 lawl 2010-10-10T20:36:28 end game, homeworld gone 2010-10-10T20:37:08 *** davidjliu has quit IRC (Ping timeout: 265 seconds) 2010-10-10T20:37:43 sigh: that looks real familiar 2010-10-10T20:37:56 what do you mean? 2010-10-10T20:38:02 you probably have a bigger weight on the value of that planet that just keeps accumulating ships 2010-10-10T20:38:43 nah, it's just that version only sends enough ships to just capture a planet, and doesn't understand the concept that an enemy planet can get more reinforcements 2010-10-10T20:39:07 so it think 11 ships is enough, but the other planets send more ships in the mean time 2010-10-10T20:42:05 hmm why are organizations not updated after writing in that thread? ie. Org Name|User 2010-10-10T20:42:53 Amstan does them in batches so the next time he grabs them all it should get updated 2010-10-10T20:43:03 ah, thanks 2010-10-10T20:43:58 maybe watch for a couple after you to get updated as well just in case he misses it somehow and you can bump him about it then 2010-10-10T20:46:00 i checked for some other people as well, seems no one has gotten updated 2010-10-10T20:46:21 but its only been 4 days 2010-10-10T20:46:46 yeah, probably just hasn't done another batch yet then 2010-10-10T20:48:22 *** evilbotzz has quit IRC (Quit: Page closed) 2010-10-10T20:49:52 sigh, what lang are you using 2010-10-10T20:50:18 c++ 2010-10-10T20:50:21 why? 2010-10-10T20:52:31 Im just wondering, Ive never worked with c++ 2010-10-10T20:52:49 I'm pretty much in the same boat :P 2010-10-10T20:53:17 so, do you have programs for tactics? or how are you approaching the game 2010-10-10T20:53:50 what do you mean by programs for tactics 2010-10-10T20:54:25 like does your bots strat change a lot 2010-10-10T20:54:35 not really 2010-10-10T20:54:54 overall I just score all target for attack, and attack the highest scores 2010-10-10T20:55:09 to attack a target, I just grab the closest ships 2010-10-10T20:55:38 score is a function of distance, cost and growth rate 2010-10-10T20:57:16 I have a defence phase at the start to lock ships that can't be used for attack, and a redistribution phase at the end to move ships to better locations 2010-10-10T20:57:33 that's my entire strategy in a nutshell 2010-10-10T20:59:32 ahh 2010-10-10T21:00:29 Top 10 players: Hazard(3826), rsergio(3820), ofrias(3805), cfaftw(3798), felixcoto(3794), dmj111(3793), sequoh(3765), EBraun(3760), _iouri_(3718), MariusCG(3715) 2010-10-10T21:03:48 *** ali has joined #aichallenge 2010-10-10T21:04:22 *** ali has quit IRC (Client Quit) 2010-10-10T21:09:25 java -jar tools/PlayGame.jar maps/map7.txt 1000 1000 log.txt "java -jar example_bots/RandomBot.jar" "java -jar example_bots/DualBot.jar" | java -jar tools/ShowGame.jar 2010-10-10T21:09:33 this command is not working for me 2010-10-10T21:10:09 what is it saying is wrong? 2010-10-10T21:10:27 bash: java: command not found 2010-10-10T21:10:27 bash: java: command not found 2010-10-10T21:10:39 then you need to install java 2010-10-10T21:10:46 you don't have java on your path 2010-10-10T21:11:14 which likely means you have to install it 2010-10-10T21:11:17 or that if you do have it installed 2010-10-10T21:11:17 alright, Ill try and get that into the equation 2010-10-10T21:11:17 *** Azrathud has joined #aichallenge 2010-10-10T21:13:40 *** Meatkat has quit IRC (Ping timeout: 265 seconds) 2010-10-10T21:14:54 I wish I knew of this before today. 2010-10-10T21:18:38 *** blbrown_win3 has joined #aichallenge 2010-10-10T21:20:26 *** davidjliu has joined #aichallenge 2010-10-10T21:23:00 knew of what? 2010-10-10T21:31:26 This AI challenge. The same thing happened to me last year. 2010-10-10T21:32:05 ah yeah, I missed the tron contest 2010-10-10T21:40:40 whats the prize? 2010-10-10T21:42:26 *** Mathnerd314 has joined #aichallenge 2010-10-10T21:42:35 *** dmj111 has joined #aichallenge 2010-10-10T21:45:36 epona: ask a1k0n :) 2010-10-10T21:53:34 *** blbrown_win3_ has joined #aichallenge 2010-10-10T21:54:44 *** Naktibalda has quit IRC (Remote host closed the connection) 2010-10-10T21:56:32 *** blbrown_win3 has quit IRC (Ping timeout: 240 seconds) 2010-10-10T21:58:16 *** Utkarsh has quit IRC (Ping timeout: 240 seconds) 2010-10-10T22:00:29 Top 10 players: Hazard(3802), rsergio(3787), cfaftw(3785), dmj111(3784), ofrias(3782), felixcoto(3759), sequoh(3738), EBraun(3734), goldman(3699), _iouri_(3698) 2010-10-10T22:01:12 jeez... I don' 2010-10-10T22:01:21 t know how many games it takes for the leaderboard to stabilize 2010-10-10T22:01:33 it seems lik even with the same bots, they change positions a lot... 2010-10-10T22:01:37 and I can't spell 2010-10-10T22:03:43 *** Utkarsh has joined #aichallenge 2010-10-10T22:10:46 *** rebelxt3 has joined #aichallenge 2010-10-10T22:17:33 i wonder if it isn't stable because it isn't really valid. given enough games, I am sure cfaftw and felixcoto would pound mine into dust. yet somehow,.... 2010-10-10T22:17:59 (among others, that is.) 2010-10-10T22:18:23 davidjliu: if I recall, our last dozen matches were decidely in your favor :) 2010-10-10T22:18:42 there is simply no dominant strategy 2010-10-10T22:19:14 that may have been what I was reaching for. but you hit it on the head. 2010-10-10T22:19:36 it's easily possible that a bot can come along that beats Hazard all the time but loses to _iouri_ all the time and changes the entire top 10 with one entry 2010-10-10T22:21:46 heh, yeah... I hope for the final competition that after a while they take the top fifty or so and just play like a thousand games for each bot or something to firmly establish placement 2010-10-10T22:21:59 yeah 2010-10-10T22:22:08 I think that stuff is happening all the time I think, just not as obviously 2010-10-10T22:22:20 and it will happen in the final tournament as well 2010-10-10T22:22:31 *** rebelxt3 has quit IRC (Ping timeout: 265 seconds) 2010-10-10T22:23:50 i wouldn't get your hopes up 2010-10-10T22:24:12 people (myself included) bitched and moaned about getting a final top x round robin on maps and it didn't happen 2010-10-10T22:24:23 last time ** 2010-10-10T22:25:35 well, I guess someone just needs to write a bot that _clearly_ beats everyone else. That will settle it :) 2010-10-10T22:25:56 *** deepblue has joined #aichallenge 2010-10-10T22:26:03 i'm already working on it 2010-10-10T22:26:07 as fast as i can :P 2010-10-10T22:26:33 awww, dmj is letting out all the secrets 2010-10-10T22:26:42 Zannick: glad someone is on the case. 2010-10-10T22:26:49 sigh: oops. my bad. 2010-10-10T22:26:58 :) 2010-10-10T22:27:19 the only secret i'll reveal is that i use c 2010-10-10T22:27:23 :P 2010-10-10T22:27:56 oh no you too? 2010-10-10T22:28:04 don't tell everybody! 2010-10-10T22:28:04 haha 2010-10-10T22:28:05 ;) 2010-10-10T22:28:13 oops, sorry 2010-10-10T22:28:15 wow. and sigh was making fun of me for using emacs as my irc client. 2010-10-10T22:28:37 well, at least i haven't written assembly 2010-10-10T22:28:43 for the bot, i mean 2010-10-10T22:29:10 that would atleast had prevent the stupid 32bit/64bit issues I had earlier -_ 2010-10-10T22:29:12 also, reading the output assembly from gcc is not that useful 2010-10-10T22:33:25 *** sigh has quit IRC (Read error: Connection reset by peer) 2010-10-10T22:39:24 i have more assignments than I can poke a stick at this week, then 5 exams 2010-10-10T22:39:32 no idea how i'll find time for ai before that's all done 2010-10-10T22:39:41 but thankfully last exam is like november 9 2010-10-10T22:41:36 antimatroid: what subject are you studying? 2010-10-10T22:42:05 graduate with majors in cs, economics and maths this year 2010-10-10T22:42:10 doing honours year in maths next year 2010-10-10T22:42:40 i basically just try and learn whatever seems fun :) 2010-10-10T22:42:53 pure maths seems to have hooked me mostly though 2010-10-10T22:44:10 sounds good. I have a math degree too. I always loved pure math, but knew I didn't have it in me to be a real mathematiciain. 2010-10-10T22:44:39 The cs course work I did was good for getting jobs :) 2010-10-10T22:44:44 yeah, i don't know if i do either, i want to do a phd in game theory, but no idea which of the 3 areas to do that through 2010-10-10T22:44:49 *** sigh has joined #aichallenge 2010-10-10T22:44:56 applied math 2010-10-10T22:45:08 and from there into game theory 2010-10-10T22:45:24 applied maths is so gross, i hate my current pdes unit (although it's got lots of hand wavy physics explanations) 2010-10-10T22:45:43 applied seems to be a lot of messy grunt work, whereas pure makes you actually think and is really awesome 2010-10-10T22:45:45 I can't do pure math so I'm getting my PhD in theoretical physics next year 2010-10-10T22:46:11 Yeah.... applied math for me was mostly for signal processing... a lot of FFTs. pde's and ode's just didn't work well for me. 2010-10-10T22:46:26 applied actually has something todo with the real world and solving problems. I liked that 2010-10-10T22:46:30 algebra and analysis :) 2010-10-10T22:46:34 mmh 2010-10-10T22:46:35 * sigh feels stupid listening to all this 2010-10-10T22:46:53 I only did game theory stuff in applied math 2010-10-10T22:46:54 and network theory 2010-10-10T22:47:16 i had a crappy eco course on it, the rest is self taught 2010-10-10T22:47:30 so crappy it stuck to 2 player games 2010-10-10T22:47:48 I agree 2010-10-10T22:47:51 would have been fun 2010-10-10T22:47:58 but looking at my code now 2010-10-10T22:48:06 I bought a bunch of Dover books on it. I manage to get through some of them... 2010-10-10T22:48:09 I would need to redoo a few sections and that would suck 2010-10-10T22:48:22 did you get the differential games ones? i term that as multiplayer optimal control theory 2010-10-10T22:48:29 haven't learnt much yet, but looks awesome 2010-10-10T22:48:45 no... I tried to stick to the intro to game theory types... that way I might understand them. 2010-10-10T22:48:46 if you have a maths background, myersons book is my favourite for game theory and still reasonably affordable 2010-10-10T22:49:10 I heard about that 2010-10-10T22:49:17 I should check it out from the library 2010-10-10T22:49:32 it's an actual maths text, fudenberd and tiroles is good, but it's written for economists 2010-10-10T22:49:39 fudenberg* 2010-10-10T22:50:45 what are you guys doing these days? 2010-10-10T22:50:50 i think you're all older than i :| 2010-10-10T22:51:17 although my partner for these comps is at the same point as me (finishing 4 year undergrad degree) and only turned like 20 earlier this year :| 2010-10-10T22:51:56 definitely.... I have been a programmer for 12 years, and all of my positions have been for someone who can deal with math. 2010-10-10T22:52:00 be happy 2010-10-10T22:52:11 my life is way to serious right now 2010-10-10T22:52:31 that was at @antimatroid 2010-10-10T22:53:28 i am happy, i don't have to do anything but learn it's great 2010-10-10T22:53:34 i plan to be a student for a long time yet 2010-10-10T22:54:13 if i went to us for phd, would be another 7 years of uni after this 2010-10-10T22:54:15 hehe 2010-10-10T22:54:30 where are you from? 2010-10-10T22:54:43 australia, the little island at the bottom called tasmania 2010-10-10T22:54:57 that's far away! 2010-10-10T22:55:10 yep, but everyone says american have the best graduate programs 2010-10-10T22:55:27 america has** (my english is terrible, despite it being my only language) 2010-10-10T22:55:31 well, I'm from germany and now doing my PhD research here 2010-10-10T22:55:34 in the US 2010-10-10T22:55:55 which uni? (or too much info?) 2010-10-10T22:56:01 it does have advantages to have an US university on paper 2010-10-10T22:56:15 Northwestern, Chicago 2010-10-10T22:56:15 *** sigh has quit IRC (Read error: Connection reset by peer) 2010-10-10T22:56:21 there's a really cool lab for game theory in the cs department at mit, but not sure i'd be able to get the funding 2010-10-10T22:56:42 MIT is awesome, was there for a conference 2010-10-10T22:56:51 so much fun stuff to toy with in their labs 2010-10-10T22:56:59 I wouldn't get anything done there ;) 2010-10-10T22:57:02 yeah, that's basically my number 1 preference, but doubt i'll get to go there 2010-10-10T22:57:47 if i could somehow win this competition that might help ;) 2010-10-10T22:59:24 the "computing" department at my uni sucks too (extremely maths averse as well), so my cs background is lacking in many places quite a lot 2010-10-10T23:00:03 if you can get a stipend or something 2010-10-10T23:00:13 things like that help a lot 2010-10-10T23:00:30 Top 10 players: Hazard(3799), rsergio(3776), cfaftw(3772), dmj111(3770), ofrias(3758), felixcoto(3746), sequoh(3736), EBraun(3718), goldman(3699), _iouri_(3680) 2010-10-10T23:01:00 yeah, i've been working as a research assistant for a good year and a half as well but for a macroeconomist, so i may be able to hook myself an ra position or something 2010-10-10T23:01:18 i'm a couple of years away from having to worry about that anyway 2010-10-10T23:03:40 http://72.44.46.68/canvas?game_id=419757 2010-10-10T23:03:42 hihi 2010-10-10T23:04:57 i can't for the life of me prevent my bot for going too close to the middle on my side with long journeys without the bot sucking, if that makes any sense 2010-10-10T23:05:07 i did it friday and put it up and the bots down at like 40 2010-10-10T23:06:32 I just uploaded my newest creation 2010-10-10T23:06:35 I hope it gets up to like top30 2010-10-10T23:06:39 it should 2010-10-10T23:06:57 but I had so many bugs that only appeared on the official 2010-10-10T23:07:17 hope that the 32/64bit issues are all sorted out now 2010-10-10T23:07:29 i'm not going to have anything new for a while 2010-10-10T23:07:46 is anyone known to have a half decent approximate minimax approach going yet? 2010-10-10T23:08:33 i'm sure it'll work quite well if someone gets the heuristics ingrained into the payoff function well enough 2010-10-10T23:08:57 the payoff function is the biggest problem 2010-10-10T23:09:05 yes, and which moves to try 2010-10-10T23:09:11 *** sigh has joined #aichallenge 2010-10-10T23:09:15 basically how do you rate a specific game situation 2010-10-10T23:09:38 i'm considering running those terminal nodes for a while, even trying to predict characteristics of their bot based off past moves 2010-10-10T23:09:47 but that'll take quite a while to implement and fine tune 2010-10-10T23:10:20 would be awesome if that works 2010-10-10T23:10:25 with the prediction 2010-10-10T23:10:48 for a first naive implementation of a tree i'll probably determine payoffs based off of ourShips - theirShips + remainingTurns(ourGrowth - theirGrowth) 2010-10-10T23:11:07 maybe thow in something about position strengths as well 2010-10-10T23:11:33 throw*, I'm using position strengths when evaluating planets naively atm 2010-10-10T23:11:59 I'm already curious how it will play out :) 2010-10-10T23:12:01 i have average distance to all other planets, then I also do a sweep each turn for average neutral/enemy/friend planet distance too 2010-10-10T23:12:38 I had a funny issue 2010-10-10T23:12:42 I once told my bot 2010-10-10T23:13:07 antimatroid: I plan to try minimax, I set up some of the framework last week 2010-10-10T23:13:10 to calculate the mass center of mine and the enemies ships and try to get them closer together 2010-10-10T23:13:23 in hope 2010-10-10T23:13:29 it would move more ships to the front 2010-10-10T23:13:35 antimatroid: my current bot is half a step minimax, just a max if you will :P 2010-10-10T23:13:37 huh, that's interesting 2010-10-10T23:14:01 I tried a few centrality measures, that failed 2010-10-10T23:14:05 yes that's what mine is doing atm, and it's still picking the most individually optimal moves it can, rather than optimal combination 2010-10-10T23:14:08 result was it attached the furthest away enemy planet and trie to push the centers closer this way 2010-10-10T23:14:12 and i am not looking forward for expansion at all :\ 2010-10-10T23:14:51 hehe 2010-10-10T23:14:51 *** sigh has quit IRC (Read error: Connection reset by peer) 2010-10-10T23:15:01 I think I solved my expansion issues for now 2010-10-10T23:15:19 hmmm, here would be a possible idea for selecting moves, just try to write yourself a bunch of reasonably different bots that cover specific points well 2010-10-10T23:15:22 now I need to wait for more results from the official server to see what I can improve on nex 2010-10-10T23:15:24 I've had only moderate success with center mass calculations 2010-10-10T23:15:27 then make moves from each at each node in the tree 2010-10-10T23:15:37 it always fails if you get flanked 2010-10-10T23:16:02 yeah 2010-10-10T23:16:13 I've given up on centrality measures or compactness measures 2010-10-10T23:16:27 there are maps 2010-10-10T23:16:32 with small clusters 2010-10-10T23:16:37 it just fails there 2010-10-10T23:16:42 you need to have more localized calculations 2010-10-10T23:16:46 i don't try to make any "macro" decisions, i want the strategy to work on any possible map 2010-10-10T23:17:06 but i will use such measures for "micro" decisions 2010-10-10T23:17:13 i think it 2010-10-10T23:17:27 I think it's good to look for clusters to attack 2010-10-10T23:17:32 oh god, measure theory makes my head want to explode 2010-10-10T23:18:13 *** sigh has joined #aichallenge 2010-10-10T23:18:54 hehe 2010-10-10T23:20:21 aye, someone is above hazard 2010-10-10T23:20:40 !rankings 2010-10-10T23:20:41 antimatroid: Top 10 players: extrapostrophe(3841), Hazard(3790), cfaftw(3770), rsergio(3768), dmj111(3763), ofrias(3749), felixcoto(3738), sequoh(3722), EBraun(3710), goldman(3680) 2010-10-10T23:21:38 probably a fluke though since it's under 20 games and doesn't have a loss yet 2010-10-10T23:21:39 *** sigh has quit IRC (Read error: Connection reset by peer) 2010-10-10T23:22:23 * janzert again wishes confidence interval reporting was in 2010-10-10T23:22:41 possibly 2010-10-10T23:23:24 so.. who's hazard? 2010-10-10T23:23:31 looks like it'll be a top ten bot certainly 2010-10-10T23:23:40 I still have a notion of centrality in my bot. Several times I've thought it was misleading the bot, but then I pull it out and the bot gets worse overall. 2010-10-10T23:23:40 whoa, someone new is first 2010-10-10T23:25:00 janzert: i expect the bots to get a lot better yet 2010-10-10T23:25:20 I agree, I meant in the current top ten 2010-10-10T23:25:24 im so fail at this 2010-10-10T23:25:27 not final 2010-10-10T23:25:37 necessarily :) 2010-10-10T23:25:41 i was getting decent periods in first place on the main server with under a week to go for tron, then we were pretty well set at around 10th in the final tourny till we timed out 2010-10-10T23:26:17 I'm glad time outs have been changed to a loss now 2010-10-10T23:26:24 I'm to the point of considering trying to make some of parameters dynamic to adjust to my opponent... 2010-10-10T23:26:25 same 2010-10-10T23:26:26 *** Mathnerd314 has quit IRC (Ping timeout: 245 seconds) 2010-10-10T23:26:27 I don't think the bots will improve much over the current level 2010-10-10T23:26:53 I think they will 2010-10-10T23:26:56 I'll be surprised if they don't 2010-10-10T23:27:07 deepblue: will be interesting to see. I would not actually be too surprised if somebody came along with something that dominates the current leaders. 2010-10-10T23:27:13 yeah, i expect mine to get a lot better yet 2010-10-10T23:27:26 i think i'm going to stop working on mine for tonight, though 2010-10-10T23:27:40 there are several new guys on top 10 atm 2010-10-10T23:33:18 is there anywhere to dl bigger map packs 2010-10-10T23:34:14 you can download more from http://72.44.46.68/maps/map###.txt 2010-10-10T23:34:28 or just download the map generator and make all you want 2010-10-10T23:34:33 just iterate from 100 up to ~450 2010-10-10T23:35:22 http://code.google.com/p/ai-contest/source/browse/trunk/planet_wars/backend/map_generator.py 2010-10-10T23:35:46 for i in {100..400};do wget "http://72.44.46.68/maps/map$i.txt";done 2010-10-10T23:37:32 @nrub lol was just writing this in python but that's easier, thanks 2010-10-10T23:37:33 pjreddie: I'm sorry Dave, I cannot nrub. 2010-10-10T23:38:04 ok guys 2010-10-10T23:38:07 I'm tired 2010-10-10T23:38:21 and the official server puts out games way to slowly 2010-10-10T23:38:26 I'm heading to bed :) 2010-10-10T23:38:41 *** deepblue has quit IRC (Quit: Page closed) 2010-10-10T23:39:11 nrub: actually i don't have wget and python is easy, just gonna do it in that 2010-10-10T23:41:02 amstan: don't know if you can but I think the cloud servers need a kick again 2010-10-10T23:41:32 spot price is still low enough that I think there should be two of them going 2010-10-10T23:43:01 how many active admins do we have these days? 2010-10-10T23:43:15 i heard danielvf is away for a while, and j3camero never shows up... 2010-10-10T23:43:26 yeah, I'm not really sure 2010-10-10T23:44:26 of course the weekend is probably worst, I've heard that some people do stuff then :) 2010-10-10T23:44:46 *** davidjliu has quit IRC (Quit: Page closed) 2010-10-10T23:45:14 yeah, no 2010-10-10T23:45:32 i can do some stuff 2010-10-10T23:45:48 but i have no idea of the cloud servers, or how they work 2010-10-10T23:45:56 ok 2010-10-10T23:47:44 amstan: i may be asking you later this week to do a pull of the compile script 2010-10-10T23:47:59 i was hoping to have jeff take a look, but... 2010-10-10T23:48:22 * Zannick shrug 2010-10-10T23:54:18 i don't think i can do a pull, lots of ppl edited the contest working dir, and didn't update svn 2010-10-10T23:54:21 so a pull would replace them 2010-10-10T23:54:24 so annoying.. 2010-10-10T23:54:33 oh 2010-10-10T23:58:33 yeah, the server changes really need to be added back into svn 2010-10-10T23:58:52 I'm afraid to work on much more until then 2010-10-10T23:59:00 *** sigh has joined #aichallenge