2011-12-06T00:04:13 http://aichallenge.org/visualizer.php?game=166717&user=12139 2011-12-06T00:04:17 916 ants in hill 2011-12-06T00:04:22 my bot is fail lol 2011-12-06T00:04:58 red looks like one of my earliest bots 2011-12-06T00:05:24 maybe it's Andrew L. Jackson 2011-12-06T00:06:37 *** yoden has quit IRC (Quit: Leaving.) 2011-12-06T00:08:17 at some point I will find the time to write my combat code haha 2011-12-06T00:08:18 basically my ants just move toward enemy when they outnumber them 2011-12-06T00:10:37 anyone remember a map with narrow gaps offhand? 2011-12-06T00:11:34 *** delt0r_ has joined #aichallenge 2011-12-06T00:12:15 *** pairofdice has joined #aichallenge 2011-12-06T00:12:53 *** Ashoka has joined #aichallenge 2011-12-06T00:15:39 MAP=maps/random_walk/random_walk_05p_01.map 2011-12-06T00:15:41 that one 2011-12-06T00:18:32 I like the cell_maze maps 2011-12-06T00:18:44 my favorite map generator 2011-12-06T00:20:07 *** treeform has joined #aichallenge 2011-12-06T00:21:08 *** Kaji has quit IRC (Ping timeout: 252 seconds) 2011-12-06T00:29:23 *** kost_ has joined #aichallenge 2011-12-06T00:30:10 * kost_ Greets everybody 2011-12-06T00:30:39 *** cyphase has quit IRC (Ping timeout: 252 seconds) 2011-12-06T00:36:52 a rare loss for xathis. (this game also changed the number 1 spot) http://aichallenge.org/visualizer.php?game=165956&user=66 2011-12-06T00:54:29 *** TheLinker has joined #aichallenge 2011-12-06T00:59:16 *** Kaji has joined #aichallenge 2011-12-06T01:05:30 *** u_ has quit IRC (Quit: u_) 2011-12-06T01:13:59 *** ltriant has quit IRC (Quit: Computer has gone to sleep) 2011-12-06T01:17:52 *** Kaji has quit IRC (Ping timeout: 240 seconds) 2011-12-06T01:17:58 *** Kaji has joined #aichallenge 2011-12-06T01:24:40 *** TheLinker has quit IRC (Remote host closed the connection) 2011-12-06T01:34:29 *** Palmik has joined #aichallenge 2011-12-06T01:48:24 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has quit IRC (Quit: Lost terminal) 2011-12-06T01:49:37 in this map, all hills are the same distance away http://aichallenge.org/map.php?map=cell_maze/cell_maze_p03_15.map 2011-12-06T01:51:18 actually, that isn't that noteworthy... 2011-12-06T01:51:41 :) 2011-12-06T01:54:05 *** GeorgeJ has quit IRC (Ping timeout: 252 seconds) 2011-12-06T01:54:19 I was thinking of some 5 players where that was true 2011-12-06T01:54:27 which is slightly more noteworthy 2011-12-06T01:54:48 so, janzert, how are we doing? 2011-12-06T01:55:38 I'm doing fine :) 2011-12-06T01:56:17 like this http://aichallenge.org/map.php?map=cell_maze/cell_maze_p05_19.map 2011-12-06T01:56:50 yeah, that's a little more interesting case :) 2011-12-06T01:57:33 I produce a list of symmetric points, then get the distances, the find the set with the smallest stdev 2011-12-06T01:57:51 it ends up picking very fair hill placements 2011-12-06T01:58:00 but just for the first hill 2011-12-06T01:59:09 just so the placement doesn't get too predictable 2011-12-06T01:59:26 only from the perspective if you know the map 2011-12-06T01:59:54 in cell maze, you aren't going to be next to water, but that's just the map type 2011-12-06T02:00:24 plus, sometimes the map doesn't have very even points, so I just pick the best one 2011-12-06T02:00:30 *** Kaji has quit IRC (Read error: Operation timed out) 2011-12-06T02:00:40 janzert: for matchmaking, I keep thinking of ways to improve it 2011-12-06T02:00:52 save 'em ;) 2011-12-06T02:00:53 I'm wondering if some type of ticket system would be better 2011-12-06T02:01:01 not for this contest :) 2011-12-06T02:01:18 and write 'em down or if it's anything like me you'll forget them 2011-12-06T02:01:22 every hour, an active bot gets one ticket 2011-12-06T02:01:34 *** cyphase has joined #aichallenge 2011-12-06T02:01:35 then it is a lottery for the next game 2011-12-06T02:01:41 something like that 2011-12-06T02:02:05 maybe the answer is more games and it all goes away 2011-12-06T02:02:30 yeah, more games/less resource intensive game would solve most of it 2011-12-06T02:02:33 it's amazing how hard the problem is 2011-12-06T02:04:48 *** Kaji has joined #aichallenge 2011-12-06T02:08:56 *** treeform has quit IRC (Remote host closed the connection) 2011-12-06T02:10:07 *** roflmao has quit IRC (Quit: Leaving.) 2011-12-06T02:10:38 *** amstan has quit IRC (Ping timeout: 240 seconds) 2011-12-06T02:12:40 *** McLeopold has quit IRC (Quit: Leaving.) 2011-12-06T02:34:49 *** jerez_z has joined #aichallenge 2011-12-06T02:35:41 hey, could an organizer help me recover code I have uploaded? I had a crash and lost everything 2011-12-06T02:49:39 *** epicmonkey has joined #aichallenge 2011-12-06T02:51:21 Yeah, me too, I'm pguillory btw 2011-12-06T02:58:04 I believed that for a second 2011-12-06T02:58:11 then I first disbelieved that he'd lose his code 2011-12-06T02:58:15 THEN I realized it was a joke 2011-12-06T02:58:51 :> 2011-12-06T02:59:53 *** delt0r_ has quit IRC (Remote host closed the connection) 2011-12-06T03:01:31 *** delt0r_ has joined #aichallenge 2011-12-06T03:03:24 *** mj41 has joined #aichallenge 2011-12-06T03:10:27 *** epicmonkey has quit IRC (Ping timeout: 240 seconds) 2011-12-06T03:16:46 *** ikaros has joined #aichallenge 2011-12-06T03:20:21 *** g0llum has joined #aichallenge 2011-12-06T03:20:28 *** sigh has joined #aichallenge 2011-12-06T03:33:08 *** mj41 has quit IRC (Ping timeout: 244 seconds) 2011-12-06T03:33:52 *** ikaros has quit IRC (Quit: Ex-Chat) 2011-12-06T03:40:02 *** McLeopold has joined #aichallenge 2011-12-06T03:44:40 *** Blkt has joined #aichallenge 2011-12-06T03:44:52 *** kire has joined #aichallenge 2011-12-06T03:45:49 good morning everyone 2011-12-06T03:50:34 *** mj41 has joined #aichallenge 2011-12-06T03:50:55 *** TwistedLogic has quit IRC (Quit: Page closed) 2011-12-06T03:51:28 *** AntDroid has joined #aichallenge 2011-12-06T03:53:23 *** g0llum has quit IRC (Read error: Connection reset by peer) 2011-12-06T03:55:37 anyone want to see my bot? 2011-12-06T03:56:33 future move calculations http://paste.aichallenge.org/wp9C4/ 2011-12-06T03:57:19 wake up antimatroid 2011-12-06T03:57:31 here 2011-12-06T03:57:56 check out my replay, with the visualizations on 2011-12-06T03:58:38 lol 2011-12-06T03:58:46 i watched the entire thing without the second visualisations 2011-12-06T03:58:49 and was like wtf is this shit :P 2011-12-06T03:58:50 ohhh 2011-12-06T03:58:56 oh, it is 2011-12-06T03:59:04 but I've got 1 cool feature 2011-12-06T03:59:07 that's neat 2011-12-06T03:59:35 Traveling ant 2011-12-06T03:59:56 ant predicted spawn ant 2011-12-06T04:00:00 and 2011-12-06T04:00:17 i gave up on predicting spawning 2011-12-06T04:00:23 how come? 2011-12-06T04:00:24 it's too messy with the random tiebreak :P 2011-12-06T04:00:41 oh, I just assume every hill gets a spawn 2011-12-06T04:01:05 i'd prefer not to rely on a ant ant coming to exist near a hill 2011-12-06T04:01:16 least i try and use that for defense and it never shows 2011-12-06T04:01:34 Assume a hill gets a spawn with a 1/hills probability! 2011-12-06T04:01:39 turn 45 is nice 2011-12-06T04:01:46 I only sent 1 ant south for 2 food 2011-12-06T04:02:16 Yeah, I was working on something similar 2011-12-06T04:03:49 although, it went to the wrong side of the food... 2011-12-06T04:14:52 http://ants.fluxid.pl/replay.15479 interesting game, score at the end it looks like cowbandit razed X32's hill and other bots did nothing 2011-12-06T04:15:40 *** willem has joined #aichallenge 2011-12-06T04:23:39 *** ikaros has joined #aichallenge 2011-12-06T04:27:20 nice! http://ants.fluxid.pl/replay.15490 look at difference in points between turn 330 and 360 - by razing each other hills we both got +1 points 2011-12-06T04:29:08 has anyone had problems with the remaining time? I set my bot to quit when theres 150 ms left but it seems to time out. 2011-12-06T04:29:24 willem: on tcp? 2011-12-06T04:29:54 thestinger: nah when submitting 2011-12-06T04:30:06 oh, no then 2011-12-06T04:30:10 you might have loops that never finish 2011-12-06T04:30:20 or algorithms that have really bad worse cases 2011-12-06T04:31:31 Strange because on my pc, it always exits at most 10ms after the limit I set. 2011-12-06T04:31:45 on a huge map with lots of ants? 2011-12-06T04:31:51 10ms might be ~40ms on the server 2011-12-06T04:31:59 and sometimes it might be a lot more than that 2011-12-06T04:32:31 Well it is dependent on the number of ants so it might be that. 2011-12-06T04:33:42 It is quite hard to guess a good limit though :) I'm just setting it to some number and seeing from the results if it does a time-out or not. 2011-12-06T04:40:12 *** willem has left #aichallenge 2011-12-06T04:41:45 *** antimatroid has quit IRC (Quit: Leaving.) 2011-12-06T04:41:49 *** antimatroid1 has joined #aichallenge 2011-12-06T04:44:20 *** retybok has quit IRC (Quit: leaving) 2011-12-06T04:47:56 *** mviel_ has joined #aichallenge 2011-12-06T04:51:27 *** mviel__ has quit IRC (Ping timeout: 240 seconds) 2011-12-06T04:52:17 http://ants.fluxid.pl/replay.15505 close call, but panic hill defence seems to be working 2011-12-06T05:04:28 *** aerique has joined #aichallenge 2011-12-06T05:10:04 *** praveen has joined #aichallenge 2011-12-06T05:18:57 *** zyberkiddy has joined #aichallenge 2011-12-06T05:24:28 *** zyberkiddy has quit IRC (Ping timeout: 255 seconds) 2011-12-06T05:25:07 *** Conorach has joined #aichallenge 2011-12-06T05:31:27 *** dai-ra has quit IRC (Ping timeout: 244 seconds) 2011-12-06T05:38:46 *** zyberkiddy has joined #aichallenge 2011-12-06T05:43:26 *** ikaros has quit IRC (Ping timeout: 252 seconds) 2011-12-06T05:44:07 *** zyberkiddy has quit IRC (Ping timeout: 252 seconds) 2011-12-06T05:46:24 *** g0llum has joined #aichallenge 2011-12-06T06:07:58 *** mviel__ has joined #aichallenge 2011-12-06T06:11:27 *** mviel_ has quit IRC (Ping timeout: 240 seconds) 2011-12-06T06:11:28 *** NoirSoldats has quit IRC (Quit: Page closed) 2011-12-06T06:12:23 *** mviel__ has quit IRC (Ping timeout: 252 seconds) 2011-12-06T06:16:12 http://ants.fluxid.pl/replay.15612 silly ants >.< 2011-12-06T06:16:43 *** kire has quit IRC (Remote host closed the connection) 2011-12-06T06:17:34 uh, you got attacked from three sides 2011-12-06T06:19:10 same problem as mine. a lot of ants defending on the other side while leaving the hill vulnerable on the other side. 2011-12-06T06:19:23 the line they formed against BenJackson (pink) was a defensive one 2011-12-06T06:19:46 meanwhile yellow is surrounding and destroying the few ants at the hill 2011-12-06T06:21:15 oh 2011-12-06T06:21:21 I actually never saw yellow's hill 2011-12-06T06:26:08 http://ants.fluxid.pl/replay.15586 blue (mybot X32) is busy defending against the red in the right while it was being attacked on the other side. needs some weights needs to be adjusted 2011-12-06T06:27:44 *** dr- has quit IRC (Ping timeout: 248 seconds) 2011-12-06T06:29:25 *** ikaros has joined #aichallenge 2011-12-06T06:29:50 my new attempt at defense seems to work pretty well 2011-12-06T06:30:00 I should just expand it to cover the whole map for territory control :P 2011-12-06T06:32:57 *** jl1990 has joined #aichallenge 2011-12-06T06:39:03 *** Blkt has quit IRC (Remote host closed the connection) 2011-12-06T06:40:16 *** antimatroid1 has quit IRC (Ping timeout: 240 seconds) 2011-12-06T06:42:17 *** Blkt has joined #aichallenge 2011-12-06T06:42:46 *** antimatroid has joined #aichallenge 2011-12-06T06:48:50 *** oddmunds has joined #aichallenge 2011-12-06T06:50:05 *** Anilm3 has joined #aichallenge 2011-12-06T06:50:12 Hi guys 2011-12-06T06:52:28 *** Ashoka has quit IRC (Quit: Page closed) 2011-12-06T07:06:51 *** Kingpin13 has quit IRC (Quit: quit) 2011-12-06T07:07:05 *** TheLinker has joined #aichallenge 2011-12-06T07:11:36 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has joined #aichallenge 2011-12-06T07:15:32 *** Kingpin13 has joined #aichallenge 2011-12-06T07:29:41 *** jacob_strauss has joined #aichallenge 2011-12-06T07:33:30 *** Jak_o_Shadows has quit IRC (Read error: Connection reset by peer) 2011-12-06T07:37:09 *** u_ has joined #aichallenge 2011-12-06T07:37:28 *** u__ has joined #aichallenge 2011-12-06T07:41:27 *** u_ has quit IRC (Ping timeout: 240 seconds) 2011-12-06T07:41:27 *** u__ is now known as u_ 2011-12-06T07:41:28 *** goffrie_ has joined #aichallenge 2011-12-06T07:41:28 *** goffrie has quit IRC (Disconnected by services) 2011-12-06T07:41:29 *** goffrie_ is now known as goffrie 2011-12-06T07:42:24 have people decided on one of the next stanford courses yet? 2011-12-06T07:42:42 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has quit IRC (Quit: Lost terminal) 2011-12-06T07:43:05 machine learning and pgm sound cool 2011-12-06T07:43:13 i think i know all the game theory stuff 2011-12-06T07:50:33 *** olexs has joined #aichallenge 2011-12-06T07:51:57 antimatroid: I'm doing ml right now and it's great 2011-12-06T07:52:11 i thought that started in january? 2011-12-06T07:52:14 or are you at stanford? :P 2011-12-06T07:52:18 nope 2011-12-06T07:52:33 ml started on september/october 2011-12-06T07:52:39 oh, maybe they're just doing it again 2011-12-06T07:52:46 exactly 2011-12-06T07:52:57 Yeah 2011-12-06T07:53:02 i was worried i missed it for a second :P 2011-12-06T07:53:08 :P 2011-12-06T07:53:33 pgm sounds great also 2011-12-06T07:53:36 ml course kinda meh imo ;p (Can't fit it into my brain) 2011-12-06T07:53:39 a1k0n is the only person i've seen succeed while doing cool things that didn't involve game trees 2011-12-06T07:53:44 so i think i need to hop on the ml train :P 2011-12-06T07:54:05 pairofdice: i'm hoping to start a phd in maths next year, i'm hoping i can handle it 2011-12-06T07:54:15 *** rajanaresh has quit IRC (Ping timeout: 268 seconds) 2011-12-06T07:54:19 too many hopings 2011-12-06T07:54:24 You probably can 2011-12-06T07:54:31 antimatroid: free-ml is not hard 2011-12-06T07:54:47 the math level required is quite low 2011-12-06T07:54:49 Anilm3: can you do the assignments in any language? or more specifically can i use c++? 2011-12-06T07:54:52 hmm that sucks 2011-12-06T07:54:55 octave 2011-12-06T07:54:57 but understandable i guess 2011-12-06T07:55:04 The material is provided in octave 2011-12-06T07:55:44 antimatroid: octave is great for this stuff, I think you wont miss C++ 2011-12-06T07:56:17 i think the daddy longlegs are getting too comfortable in my room 2011-12-06T07:56:25 i have seen like 3 in the last hour walking around 2011-12-06T07:56:41 antimatroid: I think you can still register to this semester class 2011-12-06T07:56:51 i'm quite busy atm anyway 2011-12-06T07:56:55 january sounds good 2011-12-06T07:57:09 Just for the purpose of taking a look on the material www.ml-class.org 2011-12-06T07:57:11 maybe pgm would have more maths? 2011-12-06T07:57:25 I hope so 2011-12-06T07:58:21 antimatroid: at first I was a bit pissed with the low math level, but since I'm learning a lot, it doesn't matter anymore 2011-12-06T07:58:44 antimatroid: you can still follow the original ml class, videos and materials are open 2011-12-06T07:58:56 http://see.stanford.edu/see/lecturelist.aspx?coll=348ca38a-3a6d-4052-937d-cb017338d7b1 2011-12-06T07:58:59 is it just that with assignments? 2011-12-06T07:59:27 *** delt0r_ has quit IRC (Ping timeout: 240 seconds) 2011-12-06T07:59:51 antimatroid: I think it's covers less topics 2011-12-06T07:59:54 it* 2011-12-06T08:00:18 Well, I don't think, I know.. 2011-12-06T08:00:37 there's no good game theory things online if you want above a beginner level which sucks 2011-12-06T08:00:41 I do like the ai class a lot 2011-12-06T08:00:58 there's some notes on mit for grad courses but until you know the material from other places they don't really explain anything 2011-12-06T08:01:12 i'm fairly good with game theory now though 2011-12-06T08:05:08 the pgm teacher seems to be a very successful researcher 2011-12-06T08:05:29 *** patrisk has joined #aichallenge 2011-12-06T08:07:10 lol, she published a paper called "game theory" 2011-12-06T08:08:10 she has done a lot 2011-12-06T08:08:51 *** mviel_ has joined #aichallenge 2011-12-06T08:11:49 *** delt0r_ has joined #aichallenge 2011-12-06T08:12:23 a1k0n: have you redone your ascii Yahoo thing for Google yet? 2011-12-06T08:14:07 wow I didn't know that animation was his! 2011-12-06T08:14:43 my friend pointed it out to me at uni earlier in the yea 2011-12-06T08:17:36 *** Nihil688 has joined #aichallenge 2011-12-06T08:18:19 *** jacob_strauss has quit IRC (Quit: Page closed) 2011-12-06T08:18:45 *** Nihil688 has quit IRC (Client Quit) 2011-12-06T08:24:53 *** patrisk has quit IRC (Quit: Leaving) 2011-12-06T08:29:08 *** carbonix has joined #aichallenge 2011-12-06T08:30:26 anyone willing to upload their bot binary so that anybody could download for local testing? 2011-12-06T08:33:40 sure 2011-12-06T08:34:29 it's 64 bit linux of course, but should be suitably mediocre to be useless as a submission 2011-12-06T08:36:20 AntDroid: you can use some of the example bots or old versions of your own bot for local testing; running binaries someone else gives you can be dangerous and i definitely would not recommend it. 2011-12-06T08:37:33 Minthos: currently testing on win7 64bit only 2011-12-06T08:39:14 bqf: example bots are not that good, old versions of my bot just keeps on drawing the new bot due to similarity, much older versions are no good also 2011-12-06T08:39:36 testing on tcp server is somewhat slow 2011-12-06T08:40:09 *** sigh has quit IRC (Remote host closed the connection) 2011-12-06T08:43:38 *** zoli1 has joined #aichallenge 2011-12-06T08:43:57 *** zoli1 has left #aichallenge 2011-12-06T08:48:08 *** LouisMartin has quit IRC (Ping timeout: 265 seconds) 2011-12-06T08:49:20 *** yoden has joined #aichallenge 2011-12-06T08:51:47 *** words has joined #aichallenge 2011-12-06T08:52:18 *** olexs has quit IRC (Quit: Leaving.) 2011-12-06T08:53:24 AntDroid: testing on tcp servers being slow should not be a reason to expose your system to possibly hostile code. but it is your computer, so you are of course free to do whatever you like with it ;) 2011-12-06T08:53:25 *** antimatroid has quit IRC (Ping timeout: 248 seconds) 2011-12-06T09:00:06 *** amstan has joined #aichallenge 2011-12-06T09:00:06 *** ChanServ sets mode: +o amstan 2011-12-06T09:09:22 *** antimatroid has joined #aichallenge 2011-12-06T09:11:12 *** avdg has joined #aichallenge 2011-12-06T09:11:24 bqf: you have a point. i may be just too trusting. i'll just make do with older versions of my bot for now. 2011-12-06T09:11:36 *** Seeker` has quit IRC (Remote host closed the connection) 2011-12-06T09:11:44 Most code you run is possibly hostile :) 2011-12-06T09:11:57 Just a matter of who do you trust 2011-12-06T09:13:51 Anyway, I have a great bot for you to try 2011-12-06T09:16:05 are you in fluxid or in tcpants? 2011-12-06T09:16:38 *** amstan has quit IRC (Ping timeout: 240 seconds) 2011-12-06T09:16:49 (it was a joke, I don't have a great bot) 2011-12-06T09:16:53 *** Areks has quit IRC (Ping timeout: 268 seconds) 2011-12-06T09:17:33 lol. it might be a great bot in my own bot's perspective 2011-12-06T09:22:34 Google should compete with paypal not facebook 2011-12-06T09:24:27 nevermind, didn't know about checkout 2011-12-06T09:24:32 antimatroid: ever heard of google checkout? ;) 2011-12-06T09:24:37 :) 2011-12-06T09:24:37 :p 2011-12-06T09:25:02 They are competing with everyone 2011-12-06T09:26:30 *** Mystret has joined #aichallenge 2011-12-06T09:27:52 hi 2011-12-06T09:29:19 *** Kingpin13 has quit IRC (*.net *.split) 2011-12-06T09:29:19 *** Conorach has quit IRC (*.net *.split) 2011-12-06T09:29:19 *** mj41 has quit IRC (*.net *.split) 2011-12-06T09:31:39 *** Kingpin13 has joined #aichallenge 2011-12-06T09:31:40 *** Conorach has joined #aichallenge 2011-12-06T09:31:40 *** mj41 has joined #aichallenge 2011-12-06T09:34:57 thestinger: http://ants.fluxid.pl/replay.15822 you were pushin pushing pushing and razed my hill :( 2011-12-06T09:59:19 very cool 2011-12-06T09:59:35 *** amstan has joined #aichallenge 2011-12-06T09:59:36 *** ChanServ sets mode: +o amstan 2011-12-06T10:16:00 *** ikaros has quit IRC (Quit: Ex-Chat) 2011-12-06T10:23:27 *** g0llum has quit IRC (Ping timeout: 240 seconds) 2011-12-06T10:38:20 *** praveen has quit IRC (Ping timeout: 265 seconds) 2011-12-06T10:44:11 *** NotABug has joined #aichallenge 2011-12-06T10:45:51 *** NotABug has quit IRC (Client Quit) 2011-12-06T10:49:05 *** rexou has quit IRC (Ping timeout: 260 seconds) 2011-12-06T10:53:11 *** amstan has quit IRC (Ping timeout: 252 seconds) 2011-12-06T10:57:01 *** AntDroid has quit IRC (Quit: Page closed) 2011-12-06T10:58:41 *** tobym has joined #aichallenge 2011-12-06T11:07:46 *** ikaros has joined #aichallenge 2011-12-06T11:08:07 *** dom7b5 has quit IRC (Ping timeout: 240 seconds) 2011-12-06T11:10:01 *** dom7b5 has joined #aichallenge 2011-12-06T11:12:37 *** JorgeB has joined #aichallenge 2011-12-06T11:12:45 *** olexs has joined #aichallenge 2011-12-06T11:18:44 *** aerique has quit IRC (Quit: ...) 2011-12-06T11:19:28 *** McLeopold has left #aichallenge 2011-12-06T11:19:56 *** UncleVasya has joined #aichallenge 2011-12-06T11:27:51 *** treeform has joined #aichallenge 2011-12-06T11:39:30 *** mviel__ has joined #aichallenge 2011-12-06T11:42:47 *** mviel_ has quit IRC (Ping timeout: 240 seconds) 2011-12-06T11:43:33 Servers down? 2011-12-06T11:44:24 *** McLeopold has joined #aichallenge 2011-12-06T11:50:58 *** amstan has joined #aichallenge 2011-12-06T11:50:58 *** ChanServ sets mode: +o amstan 2011-12-06T11:57:12 *** carbonix has quit IRC (Quit: Leaving.) 2011-12-06T11:57:30 *** amstan has quit IRC (Read error: Connection reset by peer) 2011-12-06T11:57:54 *** amstan has joined #aichallenge 2011-12-06T11:57:54 *** ChanServ sets mode: +o amstan 2011-12-06T11:59:02 *** kire has joined #aichallenge 2011-12-06T12:03:08 *** amstan has quit IRC (Ping timeout: 240 seconds) 2011-12-06T12:07:17 Mystret: servers are dying 2011-12-06T12:07:50 Mystret: servers are dying 2011-12-06T12:07:59 wups 2011-12-06T12:09:04 *** amstan has joined #aichallenge 2011-12-06T12:09:04 *** ChanServ sets mode: +o amstan 2011-12-06T12:11:12 *** dr- has joined #aichallenge 2011-12-06T12:14:20 *** amstan has quit IRC (Ping timeout: 244 seconds) 2011-12-06T12:14:35 @ranking 2011-12-06T12:14:36 Anilm3: An error occured while trying to show the previous error. 2011-12-06T12:14:40 @rank 2011-12-06T12:14:40 Anilm3: Error: The command "rank" is available in the ChannelStats and Factoids plugins. Please specify the plugin whose command you wish to call by using its name as a command before "rank". 2011-12-06T12:15:20 @rankings 2011-12-06T12:15:21 Anilm3: Top 10 players: pguillory(91.7), xathis(91.1), ChrisH(89.7), delineate(87.6), RVeerdonk(86.6), ThisIsNotABug(86.5), lazarant(86.2), BenJackson(86.2), Migi32(85.9), Memetix(85.7) 2011-12-06T12:22:21 *** Blkt has quit IRC (Read error: Connection reset by peer) 2011-12-06T12:23:11 *** JorgeB has quit IRC (Ping timeout: 240 seconds) 2011-12-06T12:23:28 *** Antimony has joined #aichallenge 2011-12-06T12:24:38 *** mj41 has quit IRC (Ping timeout: 248 seconds) 2011-12-06T12:26:01 servers are waking up again 2011-12-06T12:26:09 *** lethlawire has joined #aichallenge 2011-12-06T12:28:23 *** treeform has quit IRC (Remote host closed the connection) 2011-12-06T12:38:56 *** kurti has quit IRC (Changing host) 2011-12-06T12:38:57 *** kurti has joined #aichallenge 2011-12-06T12:40:41 holy rap my bot is getting slaughtered on fluxid 2011-12-06T12:42:52 haha 2011-12-06T12:43:45 *** jacob_strauss has joined #aichallenge 2011-12-06T12:43:57 @later tell janzert what do you think about a hill density requirement? http://forums.aichallenge.org/viewtopic.php?f=21&t=2056 2011-12-06T12:43:57 McLeopold: Aye, aye, sir 2011-12-06T12:44:46 *** kilae has joined #aichallenge 2011-12-06T12:45:37 i added a smarter move generator and of course it makes everything worse 2011-12-06T12:45:57 soo that means you're not done? :) 2011-12-06T12:45:59 a1k0n, keep in mind that some bots probably use the wrong time limit on fluxid 2011-12-06T12:46:16 i'm removing code, not adding!@! 2011-12-06T12:46:30 this was an earlier experiment 2011-12-06T12:46:51 I've gotten this : " Compiling: compiling and running tests" Since last night. Should the tests have started already? 2011-12-06T12:47:09 yeah, that should last seconds 2011-12-06T12:47:46 In Game: Playing in a game right now. 2011-12-06T12:47:51 first time I have seen that heh 2011-12-06T12:48:32 a1k0n, do you try to estimate number of enemy ants? 2011-12-06T12:49:18 no, i just count the ones i can see 2011-12-06T12:49:21 Darn. I've had problems since I've started. Last night I kept getting the "Unpack error:" and aamstan fixed this. But now I just get the "Compiling" result. And all I'm doing is reuploading the package I downloaded for the 5 minute tutorial! 2011-12-06T12:49:45 So you decide how much losses are worth based on enemies you can see only? 2011-12-06T12:50:03 well, yes. 2011-12-06T12:50:15 I've been thinking about trying to estimate enemy ants, since I'm worried that will fail miesrably in a defender situation 2011-12-06T12:50:17 my bot is pretty stupid; it deserves to be slaughtered 2011-12-06T12:50:24 if you can't see most of the map the bot gets overoptimistic 2011-12-06T12:50:25 lol 2011-12-06T12:50:26 oh 2011-12-06T12:50:34 well your bot is a lot higher ranked than me 2011-12-06T12:50:39 wow my bot is getting cooler on his own! 2011-12-06T12:50:46 not anymore 2011-12-06T12:50:57 that's only because you resubmitted 2011-12-06T12:51:05 it will be up there again in a couple games 2011-12-06T12:51:14 *** lethlawire has quit IRC (Quit: Page closed) 2011-12-06T12:51:26 a pity I'm not developing anymore :( 2011-12-06T12:51:30 By bot has never even broken mu=85, and yours was at 91 several versions ago 2011-12-06T12:51:42 I'm down to 8th but mainly because the average skill in the top 10 is going up quite a bit 2011-12-06T12:52:03 more top bots are floating up the average 2011-12-06T12:52:28 I really need to figure out combat soon 2011-12-06T12:52:41 used to be a ton of bots in the 83.xx and it's evening out as we all play games 2011-12-06T12:53:13 the only reason I win any games on the official server is by numbers of ants 2011-12-06T12:53:17 i shoudln't have resubmitted since i think my new bot is actually worse 2011-12-06T12:53:50 *** JorgeB has joined #aichallenge 2011-12-06T12:54:26 re-resubmit 2011-12-06T12:54:36 then it'll just be worse again 2011-12-06T12:55:06 then resumbit your old version.. 2011-12-06T12:55:11 sumbit* 2011-12-06T12:55:32 but it could be worse than what's up there, i can't actually tell 2011-12-06T12:55:57 At my level is much easier.. 2011-12-06T12:57:31 I usually evaluate my bots against each other 2011-12-06T12:58:13 Anilm3: when I do that, my v2 bot wins half the time 2011-12-06T12:58:21 same with me 2011-12-06T12:58:25 it's the current main counter to my current bot 2011-12-06T12:58:32 In my case, my last bot is the best one 2011-12-06T12:58:38 since it pretty much just suicides on your hill 2011-12-06T12:58:49 once you've got a decent bot, testing it against old versions isn't useful anymore 2011-12-06T12:59:09 so my bot's not decent then.. 2011-12-06T12:59:43 I'm not even near the top 100 2011-12-06T13:00:09 getting into the top 100 is hard nowadays 2011-12-06T13:00:17 mine is barely 90 2011-12-06T13:00:37 I desperately need combat skills 2011-12-06T13:01:54 I need skills 2011-12-06T13:02:04 :( 2011-12-06T13:02:09 and time 2011-12-06T13:02:10 Antimony: that's my issue 2011-12-06T13:03:26 *** g0llum has joined #aichallenge 2011-12-06T13:04:24 how come xathis gets three fights in a row and I have to wait hours... 2011-12-06T13:05:02 because so many bots want a piece of him 2011-12-06T13:05:17 haha 2011-12-06T13:05:35 crap, i do have to resubmit 2011-12-06T13:06:09 a1k0n: how can I make multiple ants go to a same goal without doing pathfinding on each ant? 2011-12-06T13:06:55 a1k0n or whoever has an answer and is willing to share ^^ 2011-12-06T13:07:14 pathfind from the goal? 2011-12-06T13:07:23 to each ant? 2011-12-06T13:07:33 to the whole map, via a bfs search 2011-12-06T13:07:48 that's also costly isn't it? 2011-12-06T13:07:50 then each ant checks each way it can go and goes the closest way 2011-12-06T13:07:55 how many goals do you have? 2011-12-06T13:08:02 Mainly hills 2011-12-06T13:08:09 for food I just do A* 2011-12-06T13:08:21 but I want multiple ants to go and raze hills 2011-12-06T13:08:23 so seed a bfs search with all the hills, and then have each ant go towards the closest hill 2011-12-06T13:08:41 interesting, thanks 2011-12-06T13:09:55 *** bazinga has joined #aichallenge 2011-12-06T13:10:43 *** Antimony has quit IRC (Ping timeout: 252 seconds) 2011-12-06T13:11:44 hello i would like to know if there is a way to detect walls without go in the wall 2011-12-06T13:13:12 *** olexs1 has joined #aichallenge 2011-12-06T13:13:13 before going on to the wall check if it's a wall (water) 2011-12-06T13:14:26 *** olexs has quit IRC (Ping timeout: 252 seconds) 2011-12-06T13:14:46 this guy makes a valid point: http://forums.aichallenge.org/viewtopic.php?f=24&t=1954 2011-12-06T13:14:58 ha ok i thought this function worked only on water! thanks 2011-12-06T13:14:58 not just about xathis btw 2011-12-06T13:15:12 bazinga: there are really no walls, just water 2011-12-06T13:15:49 what point is valid? 2011-12-06T13:16:26 beta gives a very big advantage 2011-12-06T13:16:59 *** Parsley has joined #aichallenge 2011-12-06T13:17:07 the only advantage you get is time 2011-12-06T13:17:17 which is already unfairly distributed 2011-12-06T13:17:20 ok thanks anilm3 2011-12-06T13:17:33 Garf: don't you think time is a ver big advantage? 2011-12-06T13:17:48 very* 2011-12-06T13:17:57 *** treeform has joined #aichallenge 2011-12-06T13:21:42 *** Parsley has quit IRC (Ping timeout: 265 seconds) 2011-12-06T13:35:30 *** rmmh has quit IRC (Ping timeout: 258 seconds) 2011-12-06T13:35:55 *** AlliedEnvy has quit IRC (Read error: Connection reset by peer) 2011-12-06T13:36:14 *** rmmh has joined #aichallenge 2011-12-06T13:38:49 *** AlliedEnvy has joined #aichallenge 2011-12-06T13:39:27 *** kire has quit IRC (Remote host closed the connection) 2011-12-06T13:44:00 *** jerez_z has quit IRC (Quit: Page closed) 2011-12-06T13:44:28 Anilm3: It is. It's just unfair, regardless of the existence of a beta period. 2011-12-06T13:44:49 *** UncleVasya has quit IRC (Ping timeout: 252 seconds) 2011-12-06T13:49:39 wh 2011-12-06T13:49:46 asdasdasds/exit 2011-12-06T13:49:47 *** Anilm3 has quit IRC (Quit: Lost terminal) 2011-12-06T13:50:13 *** Anilm3 has joined #aichallenge 2011-12-06T13:50:18 Garf: what do you mean? 2011-12-06T13:52:17 Beta period means that people who check early have more time for programming. But different people in the competition already have different time available for programming. 2011-12-06T13:53:59 *** amstan has joined #aichallenge 2011-12-06T13:53:59 *** ChanServ sets mode: +o amstan 2011-12-06T13:55:02 I agree with that. 2011-12-06T13:55:13 also, as explained in the thread: beta was actually open for everybody. 2011-12-06T13:55:22 so if you knew about the contest in June you could have joined by then. 2011-12-06T13:55:25 for what it's worth I started on about Oct 23 2011-12-06T13:55:35 other people might be sad they join now, instead of at the end of October. 2011-12-06T13:55:58 Other people are without a life and work on their bot 60 hours a week, while others have to scrap together a few precious hours to be able to work on it. 2011-12-06T13:56:45 some people also have a more relevant programming background 2011-12-06T13:56:50 some people know more suitable programming languages 2011-12-06T13:57:06 Ah good point too Ben :) 2011-12-06T13:57:10 some people have friends to bounce non-public ideas off of 2011-12-06T13:58:13 the last winner was a professional AI developer, with a good knowledge of one of the fastest dynamic languages, and even the maintainer of the compiler so he could work around known bugs on the contest servers 2011-12-06T13:58:33 that's unfair :-P 2011-12-06T13:58:59 well, I am not sure that is totally unfair. 2011-12-06T13:59:05 That's being prepared! 2011-12-06T13:59:09 *** roflmao has joined #aichallenge 2011-12-06T13:59:10 exactly. 2011-12-06T13:59:13 you could study AI too :) 2011-12-06T13:59:21 I entered the ICFP many years ago and lost to a team of grad students and their advisors who had developed a C langauge extension ideal for the challenge 2011-12-06T13:59:24 or join an open source project to write compilers :P 2011-12-06T13:59:32 (and the big team was more significant because it was a 3-day contest) 2011-12-06T13:59:33 Yeah, I was only talking about the time factor 2011-12-06T13:59:37 but I still had fun and learned a lot 2011-12-06T13:59:46 even in this contest I learned a lot 2011-12-06T13:59:51 =] that's the whole point right? :) Enjoy and learn. 2011-12-06T13:59:59 I waver about whether I can improve my rank 2011-12-06T14:00:01 or in whatever order you prefer :) 2011-12-06T14:00:04 depending on the day :) 2011-12-06T14:02:17 amstan: as a control you guys should submit a "last place" bot. it would find one of the initial food and then return to suicide with the new ant as fast as possible 2011-12-06T14:02:36 if that doesn't come in dead last then thre's probably a bug in the contest :) 2011-12-06T14:02:44 BenJackson: timeout on turn 1? 2011-12-06T14:02:53 BenJackson: huh? 2011-12-06T14:02:56 ah wait, then you can still kill 1 attacking ant 2011-12-06T14:02:58 why? 2011-12-06T14:04:00 no points are awarded for killing ants 2011-12-06T14:04:28 Minthos: no, but it can prevent an enemy capping your hill 2011-12-06T14:06:23 good point 2011-12-06T14:09:04 *** kost_ has quit IRC (Ping timeout: 265 seconds) 2011-12-06T14:10:19 Next game should be within 932 minutes. 2011-12-06T14:11:36 *** pairofdice has quit IRC (Quit: leaving) 2011-12-06T14:13:27 it shouldn't 2011-12-06T14:14:52 *** bazinga has left #aichallenge 2011-12-06T14:14:55 bots that get 0 points frequently tie other bots, so it might be hard to make it dead last 2011-12-06T14:18:33 *** Accoun has quit IRC () 2011-12-06T14:21:26 amstan: I was just thinking about what kind of bot a starter kit could beat 2011-12-06T14:21:36 and I figured it would require actively being worse than doing nothing 2011-12-06T14:21:39 BenJackson: why? 2011-12-06T14:21:48 BenJackson, if a player suicides he doesnt loose points doesnt he? 2011-12-06T14:21:56 LeftyBot is good 2011-12-06T14:21:57 well you can't compete to win, you can compete to lose 2011-12-06T14:21:59 so if the others are starter bots which dont raze 2011-12-06T14:22:00 it can explore the map 2011-12-06T14:22:06 it is still draw :) 2011-12-06T14:22:21 amstan: plus from a practical (ok, not that practical) consideration it *should* come in last place 2011-12-06T14:22:29 so no one else has to 2011-12-06T14:22:37 and if it doesn't that's probably a bad sign! 2011-12-06T14:23:20 it wouldnt be worse than the starter bots 2011-12-06T14:23:43 *** Hexren has joined #aichallenge 2011-12-06T14:23:53 *** Hexren has left #aichallenge 2011-12-06T14:23:56 perhaps on some small maps where those accidently find the hill 2011-12-06T14:24:34 in 1v1 the starter would get a bonus 2011-12-06T14:25:25 ok in 1on1 :) 2011-12-06T14:25:30 *** arscan has quit IRC (Ping timeout: 265 seconds) 2011-12-06T14:26:05 BenJackson: if you look at the end of rankings it pretty much is like that 2011-12-06T14:26:44 *** treeform has quit IRC (Remote host closed the connection) 2011-12-06T14:27:17 *** treeform has joined #aichallenge 2011-12-06T14:29:28 wow this bot times out every game and has skill 31.23 2011-12-06T14:32:42 it shouldn't 2011-12-06T14:34:34 it's scary how very little change in weights in my bot causes it to win game 150 turns earlier on one map and 200 turns later on other map 2011-12-06T14:35:21 i finally fixed that food-collection-stalemate bug i had where my ants would get stuck between food sometimes 2011-12-06T14:35:43 now that i have real work to do i can't resist procrastinating by working on my bot. argh. 2011-12-06T14:36:36 story of my life 2011-12-06T14:37:05 ↑ this 2011-12-06T14:37:16 *** Accoun has joined #aichallenge 2011-12-06T14:41:57 mine too... 2011-12-06T14:42:09 it's nice to see I'm not alone lol 2011-12-06T14:42:56 whole workday i waited for it to end so i can go back home and fix bot 2011-12-06T14:44:49 a1k0n, i have a question regarding you forum post. when you simulate moves.. say you have 3 ants at positions (a_0,b_0,c_0).. you move them randomly to new positions (a_1,b_1,c_1).. in the next iteration you continue to move them from *_0 or _1 2011-12-06T14:45:01 i hope i explained that right :) 2011-12-06T14:46:00 ikaros: I would think it would be systematically rather than randomly 2011-12-06T14:46:35 or in other words.. in every iteration of the simulation, each ant has max dist of 1 to its origin at iteration 0 2011-12-06T14:46:51 rwest, that's not what i meant. 2011-12-06T14:47:24 yes you choose the best of 5 directions so its not random.. used the wrong words :) 2011-12-06T14:48:37 *** amstan has quit IRC (Ping timeout: 252 seconds) 2011-12-06T14:55:19 *** iglo has joined #aichallenge 2011-12-06T14:57:52 *** JorgeB has quit IRC (Ping timeout: 240 seconds) 2011-12-06T15:11:50 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has joined #aichallenge 2011-12-06T15:23:33 *** besh has joined #aichallenge 2011-12-06T15:29:18 *** Anilm3 has quit IRC (Quit: Lost terminal) 2011-12-06T15:34:42 Trying to understand a1kon's combat method but failing so far. Can anyone give me a short description. 2011-12-06T15:35:09 *** Anilm3 has joined #aichallenge 2011-12-06T15:36:11 what i don't understand: a) is there evaluation ? b) is there a tree ? 2011-12-06T15:37:25 a) yes b) no 2011-12-06T15:37:39 you move the ants one after the other in all possible directions 2011-12-06T15:37:47 and for each move you evaluate 2011-12-06T15:38:23 ok. how do you evaluate? ant count ? 2011-12-06T15:38:34 well evaluation is black magic :) 2011-12-06T15:38:42 you can do several things 2011-12-06T15:38:53 from ant counting to distance to specific targets 2011-12-06T15:39:19 ok. so is it a one ply monte-carlo sampling ? 2011-12-06T15:39:27 or some kind of unseen tiles bonus 2011-12-06T15:39:41 this is up to everyone i guess .. to find a good evaluation 2011-12-06T15:39:57 its not really one ply as far is i understand it 2011-12-06T15:40:10 because you move your ants repeatedly until time runs out 2011-12-06T15:40:18 can it find tactics like cornering an ant in a 2 vs 1 situation which could require 3 steps lookahead f.i. 2011-12-06T15:41:01 yes you move your ants repeatedly but all of them at the root as far as I understood. 2011-12-06T15:41:04 if i understand it right.. you could.. accidentally or by having the evaluation function "detect corners" 2011-12-06T15:41:12 but i think that would be kinda hard 2011-12-06T15:41:30 besh yea that was one thing i wasnt sure myself so far 2011-12-06T15:41:53 but i agree.. i think you move them at a distance of max 1 to their origin always 2011-12-06T15:42:31 *** kapoc has joined #aichallenge 2011-12-06T15:43:40 but you could always include something like attack and defense vectors in your eval function 2011-12-06T15:43:42 e.g. 2011-12-06T15:43:50 i have no idea if this would be performant.. 2011-12-06T15:45:33 i think it can be helpful to mix different strategies if you have a lot. When tactics is involved (face to face with enemy ants), that dominates. 2011-12-06T15:46:51 i suspect xathis does at least a 2-ply lookahead. Some of the 2 vs 1 situations I observed are impossible to guess with simple rules. 2011-12-06T15:48:10 *** kapoc has left #aichallenge 2011-12-06T15:48:26 *** bysin has quit IRC (Quit: Leaving) 2011-12-06T15:49:19 possible.. will be interesting to read the post-mortems :) 2011-12-06T15:50:06 *** epicmonkey has joined #aichallenge 2011-12-06T15:50:22 Right now I consider enemy ants to be static... I considered a monte-carlo thingy like a1kon's i.e assuming different positions for enemy ants. 2011-12-06T15:50:48 you can also do it with ants moving more than distance 1 2011-12-06T15:50:55 But i probably run out of time in 3-4 iterations though. 2011-12-06T15:51:16 3-4 iterations of your ant list? 2011-12-06T15:51:30 *** Paradoxial has joined #aichallenge 2011-12-06T15:51:47 yes if i do it for all of them including those not involved in battle. 2011-12-06T15:52:05 *** ltriant has joined #aichallenge 2011-12-06T15:52:18 I spend 100ms on average 2011-12-06T15:52:44 well you have 500ms :) but still that sounds slow 2011-12-06T15:52:53 is your eval function very complex? 2011-12-06T15:53:29 if not id try to benchmark making moves only without eval and try to optimize the data structure you use 2011-12-06T15:54:25 i don't have eval. I just do move orders in pre-determined order like in the tutorials. 2011-12-06T15:54:47 language? 2011-12-06T15:54:54 Java 2011-12-06T15:55:05 hmm that shouldnt be a show stopper :) 2011-12-06T15:55:37 no it isn't. But anyway i want to spend the rest of my time (400ms) in the best way possible ;) 2011-12-06T15:55:57 well if you use monte carlo only to move you dont need anything else 2011-12-06T15:56:01 *** Wraithan is now known as _ 2011-12-06T15:56:31 *** _ is now known as Guest18875 2011-12-06T15:56:47 *** Guest18875 is now known as ^ 2011-12-06T15:57:11 You also need to implement that complicated attack rule to know which ants go off, after making moves temporarly. 2011-12-06T15:57:17 *** ^ is now known as Guest23932 2011-12-06T15:57:24 *** Guest23932 is now known as Wraithan 2011-12-06T15:59:43 *** g0llum has quit IRC (Read error: Connection reset by peer) 2011-12-06T15:59:47 *** delt0r_ has quit IRC (Ping timeout: 240 seconds) 2011-12-06T16:01:16 it's just one-ply 2011-12-06T16:01:23 you don't move your ants more than one square away from where they started 2011-12-06T16:01:28 i guess i didn't explain that part very well 2011-12-06T16:01:31 ok thx 2011-12-06T16:01:46 based on that i have another question :) 2011-12-06T16:01:56 a1kon: thanks. How does it perform regarding tactics ? 2011-12-06T16:02:18 it isn't smart enough to know it's going to chase ants into a wall, but it will do it anyway 2011-12-06T16:02:23 like in my example above.. you move ants (a_0,b_0,c_0) to (a_1,b_1,c_1) 2011-12-06T16:02:32 then on the next iteration 2011-12-06T16:02:54 a moves from a_0 again but the others are still at _1, right? 2011-12-06T16:03:19 a moves from a_1 to some other place, possibly a_0 again 2011-12-06T16:03:45 hmm but if it moves from the target it could move 2 tiles away from the origin 2011-12-06T16:03:55 oh 2011-12-06T16:04:01 oh, right. it moves from a_0 again, yes. 2011-12-06T16:04:06 ok 2011-12-06T16:04:20 a1kon: I understand that it is useful for mixing different strategies but I thought it was mainly for resolving combat guessing from the tile of the trhead. 2011-12-06T16:04:22 but you cant really be sure there are no collisions in the end?! 2011-12-06T16:04:24 *** amstan has joined #aichallenge 2011-12-06T16:04:25 *** ChanServ sets mode: +o amstan 2011-12-06T16:04:44 ikaros: well you disallow any moves that cause collisions 2011-12-06T16:05:22 so if a moves right and b moves to where a was already, then a can't later decide to not move 2011-12-06T16:05:31 (unless b moves away again) 2011-12-06T16:05:51 but i check collisions only in the current state? so there is no guarentee my highest sample will lead to an unoccupied tile in the end 2011-12-06T16:05:58 or im getting something really wrong 2011-12-06T16:05:58 besh: it's kinda for both. without an opponent, you don't need to do any of the fancy stuff 2011-12-06T16:06:02 (probably) 2011-12-06T16:06:28 well, whenever you move an ant provisionally, mark the tile they're going to be 2011-12-06T16:06:33 yea 2011-12-06T16:06:36 and whenever you geenrate moves, check that no ant is going there 2011-12-06T16:07:22 i dont know how to explain my problem.. probably i should think about it first :)) 2011-12-06T16:08:05 a1k0n: wow you have confused the hell out of people 2011-12-06T16:08:11 I approve! 2011-12-06T16:08:13 sigh 2011-12-06T16:08:24 anyways it's certainly not *the* solution for combat, it's just better than nothing 2011-12-06T16:08:25 so if you make your moves in the end(sending them to stdout) you dont check the target tile again? 2011-12-06T16:08:41 what's to check once you've moved? you're done 2011-12-06T16:08:51 hmm 2011-12-06T16:09:03 *** analyst74 has joined #aichallenge 2011-12-06T16:09:05 oh you mean if you send one ant's moves, then re-check for the rest of them? 2011-12-06T16:09:11 i thought the moves in the simulation only influence the dirichlet 2011-12-06T16:09:21 right 2011-12-06T16:09:24 and in the end you choose the best one 2011-12-06T16:09:40 yeah, but the best one can be obscured by an ant being in the way from previous sampling 2011-12-06T16:09:40 but couldnt it be two ants choose the same target then? 2011-12-06T16:09:51 a1k0n: the latest post on the forum is thikning you are simulating multiple moves *into the future* 2011-12-06T16:09:52 well, one of them has to choose somewhere else 2011-12-06T16:10:16 ok but this is what you check again when you send the moves 2011-12-06T16:10:29 well.. i think i got it :) 2011-12-06T16:10:34 haha. well, you could do it for multiple turns 2011-12-06T16:10:38 but it'd be.. had 2011-12-06T16:10:39 hard 2011-12-06T16:12:23 *** delt0r_ has joined #aichallenge 2011-12-06T16:12:39 Well to for resolving tactics it needs to expand the the promising moves 2-plies or more as in UCT. 2011-12-06T16:13:04 ideally, yes. i'm not sure how to implement UCT in this case 2011-12-06T16:13:48 I would think the evaluation has to guess the best opponent response as there are no terminal positions here. 2011-12-06T16:14:02 right. 2011-12-06T16:14:16 well, unless you count "enough ants died" as terminal 2011-12-06T16:14:36 but yes, it'd be impossible to simulate the game all the way through to all hills being taken 2011-12-06T16:14:57 yes there could be a threshold that flags a position as practically won for one side or the other. 2011-12-06T16:15:37 couldnt you do a monte carlo tree search until a max number of plys and then evaluate 2011-12-06T16:15:41 Before I knew you can monte-carlo to the end of the game, I used something similar for 9-9 go. 2011-12-06T16:16:29 go solvers can see the end of a 9x9 go game from the start?? 2011-12-06T16:16:44 go ai simulates whole games yes 2011-12-06T16:16:53 the monte carlo approaches at least 2011-12-06T16:16:59 yes if you do random moves that is. 2011-12-06T16:17:58 i never got to try multiple-move lookahead with ants. it's a little tricky to implement the state, but probably worth doing 2011-12-06T16:18:07 in the pousse contest years ago I had no idea how to play at all but I needed a starting move 2011-12-06T16:18:17 so I ran my minmax on the largest game I could see clear to the end 2011-12-06T16:18:25 and extrapolated that move to larger boards 2011-12-06T16:18:44 but it was only 4x4 or something 2011-12-06T16:19:39 a1k0n: if I looked ahead more moves I'd somehow have to trade that cost off against looking at the 1 move 2011-12-06T16:19:44 *** ikaros_ has joined #aichallenge 2011-12-06T16:19:48 which would be tricky 2011-12-06T16:19:59 yeah 2011-12-06T16:22:33 *** iglo has quit IRC (Quit: .) 2011-12-06T16:23:05 *** freddykrueger has joined #aichallenge 2011-12-06T16:23:09 what really helped the winners in that one was avoiding recalculating from known board positions 2011-12-06T16:23:29 I pondered that for this game but it would be much more complex 2011-12-06T16:25:56 a1kon: one reason why I shouldn't have started with starter package. If I had enough time to start by myself, first thing I would implement those trick combat rules. 2011-12-06T16:27:25 *** Antimony has joined #aichallenge 2011-12-06T16:29:37 besh: I don't follow that 2011-12-06T16:29:43 *** Sir_Ragnarok_ has joined #aichallenge 2011-12-06T16:30:23 *** Sir_Ragnarok has quit IRC (Ping timeout: 252 seconds) 2011-12-06T16:31:12 The starter packages do not allow to temporarliy move ants for lookaheads. a1kon's and similar methods require that. 2011-12-06T16:31:40 oh 2011-12-06T16:31:58 I think the data structures for combat are key to the implementation 2011-12-06T16:32:05 I don't think any reasonable starter could reflect that 2011-12-06T16:32:19 mine look nothing like a grid of ants... 2011-12-06T16:33:02 yes. I would have to rewrite a lot of stuff to get that working. But I didn't have high ambitions when I started :) 2011-12-06T16:34:11 all i did was add some stuff to Square.h 2011-12-06T16:34:14 it's not that hard 2011-12-06T16:36:12 *** Palmik has quit IRC (Remote host closed the connection) 2011-12-06T16:37:10 The java one has list for ants and hills a. And also determining which ants get captured is not easy. 2011-12-06T16:37:45 All moves should be made to determine which ants get captured. 2011-12-06T16:38:14 besh: you don't want to do combat calculations on all ants 2011-12-06T16:38:21 well I wouldn't 2011-12-06T16:38:29 *** Antimony has quit IRC (Ping timeout: 252 seconds) 2011-12-06T16:39:00 so yes, you need to maintain the set of enemy ants which have entered the combat radius of each ant 2011-12-06T16:40:01 yes exactly. You implemented it so you know better :) 2011-12-06T16:40:45 the rule that says "our ant should be less distracted than the ant we are attacking" is difficult. 2011-12-06T16:40:56 it is 2011-12-06T16:45:15 ahh shit 2011-12-06T16:45:19 I just found a huge bug 2011-12-06T16:45:23 or possibly a huge feature 2011-12-06T16:45:57 speaking of "random heuristics" 2011-12-06T16:46:41 move_score = rand()? 2011-12-06T16:46:43 if I go from seeing enemy hills to not seeing any enemy hills I have stale hill proximity data until I see another hill 2011-12-06T16:47:06 I fixed this incidentally but then "failed" my regression check against a test game 2011-12-06T16:47:19 I implemented the code from the spec page exactly 2011-12-06T16:48:13 this requires a per-ant gathering of neighbours in attack range 2011-12-06T16:48:26 *** sigh has joined #aichallenge 2011-12-06T16:48:43 *** kilae has quit IRC (Quit: ChatZilla 0.9.87 [Firefox 8.0/20111104165243]) 2011-12-06T16:50:59 *** DrBrask has joined #aichallenge 2011-12-06T16:52:10 *** sigh has quit IRC (Remote host closed the connection) 2011-12-06T16:53:29 Garf: yes that is the way. Do also color differentiate in multi player games ? 2011-12-06T16:53:35 Do you * 2011-12-06T16:53:40 *** olexs1 has quit IRC (Quit: Leaving.) 2011-12-06T16:54:34 *** jerez_z has joined #aichallenge 2011-12-06T16:55:12 there is the one-on-one-on-one rule in the specs f.i. 2011-12-06T16:57:00 I am gonna have to change so much in this java starter package to make my combat code work 2011-12-06T16:58:41 rwest: I am too lazy to do it. Please be kind enough to strip off your magic code and update starter package. 2011-12-06T16:58:49 :) 2011-12-06T17:00:00 *** amstan has quit IRC (Ping timeout: 268 seconds) 2011-12-06T17:03:59 hm, I was wrong about my bug. still different results, but back to no explanation 2011-12-06T17:10:30 *** besh has quit IRC (Quit: Page closed) 2011-12-06T17:10:30 *** Conorach has quit IRC (Ping timeout: 248 seconds) 2011-12-06T17:13:46 *** retybok_ has joined #aichallenge 2011-12-06T17:15:41 ok, two last things on my TODO list, possibly hard 2011-12-06T17:16:06 ahhh I think my problem is that a "meaningless" change in my new implementation is causing a slightly different sequence of rand() to happen 2011-12-06T17:16:10 *** AntDroid has joined #aichallenge 2011-12-06T17:16:11 which is permuting my random sampling 2011-12-06T17:18:19 *** tobel has joined #aichallenge 2011-12-06T17:20:57 *** AntDroid has quit IRC (Ping timeout: 265 seconds) 2011-12-06T17:23:26 Why do I tend to get 3 games in a row, then nothing for several hours? Isn't that a bit wasteful? 2011-12-06T17:25:28 *** DrBrask has quit IRC (Quit: Page closed) 2011-12-06T17:26:22 you are the seed in one and players of nearby rank pull you into their games around the same time 2011-12-06T17:26:40 *** LouisMartin has joined #aichallenge 2011-12-06T17:27:03 BenJackson: it doesn't seem to happen at random 2011-12-06T17:27:13 BenJackson: I've noticed it several times 2011-12-06T17:29:45 antimatroid, are you there? 2011-12-06T17:30:17 *** freddykrueger has quit IRC (Quit: Page closed) 2011-12-06T17:31:10 *** Antimony has joined #aichallenge 2011-12-06T17:32:05 besh: no, it will be wrong with >2 fighting colors 2011-12-06T17:33:30 *** raemde has quit IRC (Read error: Connection reset by peer) 2011-12-06T17:34:19 *** AntDroid has joined #aichallenge 2011-12-06T17:34:53 *** jerez_z has quit IRC (Quit: Page closed) 2011-12-06T17:35:15 retybok_: right, the algorithm for picking opponents is not random 2011-12-06T17:35:20 *** amstan has joined #aichallenge 2011-12-06T17:35:20 *** ChanServ sets mode: +o amstan 2011-12-06T17:35:46 Regarding http://forums.aichallenge.org/viewtopic.php?f=26&t=1872&start=10 and the possible modification of https://github.com/aichallenge/aichallenge/blob/epsilon/worker/compiler.py, can anyone help/proof me as to how to proceed? 2011-12-06T17:35:55 *** raemde has joined #aichallenge 2011-12-06T17:37:29 I think the line 318 would become: "C#" : [["gmcs", "-warn:0", "--optimize=all", "-out:%s.exe" % BOT]], 2011-12-06T17:38:01 to add the optimization flag to the C# compilation. 2011-12-06T17:40:13 now I need to create a pull request with that change, but I'm not used to Git. I create a new branch, make the change and then do a pull request on this branch? not even sure it makes sense 2011-12-06T17:40:48 amstan? 2011-12-06T17:40:53 aichallenge: Alexandru Stan epsilon * r9008cf8 / worker/compiler.py : added optimization to C# as per request from http://forums.aichallenge.org/viewtopic.php?f=26&t=1872 - http://git.io/Da2Jog 2011-12-06T17:40:58 LouisMartin: how's that? 2011-12-06T17:41:21 you don't have to make a new branch 2011-12-06T17:41:44 just pull, clone your repo, make your change, push to your repo, submit pull request 2011-12-06T17:42:15 LouisMartin: anyway, is that change correct for sure? 2011-12-06T17:42:37 *** ikaros_ has quit IRC (Quit: Ex-Chat) 2011-12-06T17:42:50 amstan: okay, but what about CIA-53 just wrote about? .. I'm confused or have you already done the change? 2011-12-06T17:43:03 i already did the change, click it 2011-12-06T17:43:25 amstan: per mono documentation, it is. 2011-12-06T17:44:11 contestbot: later tell janzert can we have new worker instances so c# optimization kicks in? 2011-12-06T17:44:11 amstan: Ready to serve. 2011-12-06T17:44:26 but now that you ask, I'm getting nervous :P will triple check right now 2011-12-06T17:44:28 only 3 workers online? weird 2011-12-06T17:45:18 amstan: will all C# players have to resubmit? 2011-12-06T17:45:53 BenJackson: no, each worker recompiles on their own space, new workers means more compiling 2011-12-06T17:46:07 "--optimize=MODE , -O=MODE. MODE is a comma separated list of optimizations. They also allow optimizations to be turned off by prefixing the optimization name with a minus sign." 2011-12-06T17:46:35 unless you think you're in a better position to play a game, in that case resubmit so you get new games 2011-12-06T17:46:38 so, --optimize=all or -O=all .. yep. 2011-12-06T17:47:26 Thanks amstan 2011-12-06T17:49:50 *** LouisMartin has quit IRC (Quit: Page closed) 2011-12-06T17:51:39 that means I'll have even more time to fill up with meaningful stuff that doesn't exist yet D: 2011-12-06T17:51:48 amstan: amazon made sure you're getting your new workers ;) 2011-12-06T17:52:10 janzert: i noticed 2011-12-06T17:52:14 you did update the worker tarball right? 2011-12-06T17:52:14 lol 2011-12-06T17:52:23 how? 2011-12-06T17:52:32 i've never done it 2011-12-06T17:52:48 from the aichallenge directory run setup/create_worker_setup.py website/ 2011-12-06T17:52:55 I can do it 2011-12-06T17:53:04 i already have a shell 2011-12-06T17:53:06 but means the workers are still running the old code 2011-12-06T17:53:42 I'll wait a bit to manually restart them since the spot price has been jumping around anyway 2011-12-06T17:53:45 i need to change worker_ssh -r to ping in reverse orde so i can kill it 2011-12-06T17:53:52 :) 2011-12-06T17:54:05 or only ping workers that have had a game in the last week or something 2011-12-06T17:54:32 janzert: nope, worker 510 is back 2011-12-06T17:54:48 host key verification failed though 2011-12-06T17:55:07 it's another worker with the same ip but different os 2011-12-06T17:55:16 i.e. someone elses server 2011-12-06T17:55:24 nope 2011-12-06T17:55:26 it's ours 2011-12-06T17:55:30 i can login to it 2011-12-06T17:55:44 it's just that i've logged in to another server we had on the same ip 2011-12-06T17:55:51 ahh, yes 2011-12-06T17:56:03 shouldn't it have a different key in the db? 2011-12-06T17:56:09 nope 2011-12-06T17:56:23 key is by ip 2011-12-06T17:56:50 worker_ssh would have found 510 if only querying servers within past week though 2011-12-06T17:57:02 it already did 2011-12-06T17:57:07 but it's actually a new server 2011-12-06T17:57:10 but it has an old id 2011-12-06T17:57:14 that's fine 2011-12-06T17:57:15 because of same ip stuff 2011-12-06T17:57:31 so why not restart now? what does spot prices have to do with restarting? 2011-12-06T17:57:54 amazon will probably be restarting anyway and if we restart we pay for the remainder of the hour 2011-12-06T17:58:09 i see 2011-12-06T17:58:11 ok 2011-12-06T17:58:13 I normally time restarts so they fall at the end of the hour charge 2011-12-06T17:58:27 not that it's a big cost :) 2011-12-06T18:02:07 it's too bad time controls are baked into every bot 2011-12-06T18:02:14 it would have been worth reworking the engine to use cpu time 2011-12-06T18:04:32 *** Zed_ has joined #aichallenge 2011-12-06T18:09:55 *** Apophis has quit IRC (Ping timeout: 244 seconds) 2011-12-06T18:10:01 *** Jak_o_Shadows has joined #aichallenge 2011-12-06T18:13:46 Is there a way to get the leaderboard in json format? 2011-12-06T18:13:59 (or anything easily parsable) 2011-12-06T18:14:21 @botrepo 2011-12-06T18:14:22 amstan: botrepo could be https://github.com/aichallenge/contestbot. 2011-12-06T18:14:29 retybok_: check the rankings.py file i think 2011-12-06T18:14:34 it's used to give this: 2011-12-06T18:14:36 @rankings 2011-12-06T18:14:37 amstan: Top 10 players: pguillory(92.1), xathis(91.3), ChrisH(89.3), delineate(87.6), lazarant(87.1), ThisIsNotABug(87.1), RVeerdonk(86.7), GreenTea(86.4), Migi32(85.9), BenJackson(85.7) 2011-12-06T18:14:41 uses json 2011-12-06T18:15:35 Thanks! I'll have a look 2011-12-06T18:16:19 better resubmit before I'm out of the top 10 altogether :) 2011-12-06T18:16:29 *** Apophis has joined #aichallenge 2011-12-06T18:17:23 BenJackson: you could go ahead and change the engine 2011-12-06T18:17:31 it won't be a waste to fix this for next contest 2011-12-06T18:18:53 amstan: would it be possible to extend the deadline of the contest? 2011-12-06T18:18:58 I'll work on it after this contest 2011-12-06T18:19:07 Anilm3: we never extended any past contest 2011-12-06T18:19:14 Anilm3: do you have a good reason? 2011-12-06T18:19:36 Not good enough 2011-12-06T18:19:56 *** Antimony has quit IRC (Ping timeout: 240 seconds) 2011-12-06T18:20:09 I thought it would be interesting to put up a vote and see what people think about it 2011-12-06T18:20:41 hehe my reason would be that i have more time after dec 25. :P but i guess that isnt enough :)) 2011-12-06T18:23:19 don't you love it when you spend hours on a new feature and it performs way worse than before 2011-12-06T18:25:20 *** Antimony has joined #aichallenge 2011-12-06T18:26:00 Friblurks: at least you've learned somethin :) 2011-12-06T18:26:11 true 2011-12-06T18:26:42 if we start the next contest backend in django, I'd probably like to see the beta period use ants 2011-12-06T18:27:05 at least until we get a new game 2011-12-06T18:27:09 are these contests once a year? 2011-12-06T18:27:16 meaning you could still write and submit bots and we could keep the top 100 going 2011-12-06T18:27:49 amstan: I'm thinking switch to django, then design a new game, to shorten the period that everyone works on the next game 2011-12-06T18:28:02 Any idea what the next game will be? 2011-12-06T18:28:14 retybok_: depends on the suggestions 2011-12-06T18:28:19 but no 2011-12-06T18:31:31 *** words has quit IRC (Ping timeout: 265 seconds) 2011-12-06T18:32:08 *** antimatroid has quit IRC (Ping timeout: 244 seconds) 2011-12-06T18:33:07 *** bluegaspode has joined #aichallenge 2011-12-06T18:35:16 McLeopold: how do you feel about the worker code and the manager code? 2011-12-06T18:35:26 how much do you think can be reused for zeta? 2011-12-06T18:35:39 also, what do you think about the current zeta worker code made by jbroman? 2011-12-06T18:36:04 *** jacob_strauss has quit IRC (Quit: Page closed) 2011-12-06T18:36:20 *** antimatroid has joined #aichallenge 2011-12-06T18:37:34 *** luizribe` has quit IRC (Remote host closed the connection) 2011-12-06T18:39:19 *** ikaros has quit IRC (Quit: Ex-Chat) 2011-12-06T18:39:47 http://fluxid.pl/misc/2011-12-07-003911_1282x206_scrot.png looks promising... i hope it continues like that. fluxid_6 is current submission 2011-12-06T18:40:03 what about the next contest/future of aichallenge then? i saw a post about zeta on the forums (the roadmap), but that one dated back to january 2011-12-06T18:40:10 is there an updated roadmap available somewhere? 2011-12-06T18:41:19 oh boy, pguillory pwned xathis 2011-12-06T18:43:25 my bot has 0 score on over 80% of games on tcp o.o 2011-12-06T18:44:19 they say on tcp bots are hard 2011-12-06T18:44:22 mine probably isn't far off 2011-12-06T18:46:18 mine tends to win or get 0 2011-12-06T18:46:21 trying to fix the latter : 2011-12-06T18:47:01 Fluxid: they are 2011-12-06T18:47:02 but still 2011-12-06T18:48:12 amstan: I think the worker/manager code is good, and can be used with django 2011-12-06T18:48:27 I think jbroman and I both worked on the same thing in different ways 2011-12-06T18:48:33 we'll have to choose one 2011-12-06T18:49:04 Some help with the bots needed: can I have a data file in my bots submission? 2011-12-06T18:49:20 bluegaspode: yes 2011-12-06T18:49:51 where do I need to put it? 2011-12-06T18:49:57 oh boy, few changes and fixes and my current version totally pwns older one... i'll test it on tcp later 2011-12-06T18:50:50 (for a Java bot) 2011-12-06T18:51:05 bluegaspode: it depends on your code 2011-12-06T18:51:30 http://tcpants.com/replay.5182 my bot is starting to behave like a top bot. beating bugnuts here. :) 2011-12-06T18:52:09 Initially I tried to load via classpath (as the file was just contained in the package) - but this didn't work - as I guess the Java files are all put into a jar - put not my *.txt file 2011-12-06T18:52:23 h 2011-12-06T18:52:34 hm, loads slowly for me 2011-12-06T18:53:18 everybody picks on me - I have a very friendly bot. you should leave me alone. 2011-12-06T18:53:40 AntDroid: you pause when your weakness == enemy weakness? 2011-12-06T18:53:52 your jar is going to be in the same folder as the stuff that you provide, so if you have the jar path somewhere, use that to get to your data file 2011-12-06T18:53:58 bugnuts: man., i know! 2011-12-06T18:54:19 every time i play capture the flag, people keep shooting me, i'm not violent though, all i'm trying to do is carry this red flag 2011-12-06T18:54:53 :))) 2011-12-06T18:54:57 So I guess I need to load my data file via absolut path - but I'd need to now how the directory structure will look like on the server. Is the zip just unzipped and the compiled bot.jar added? 2011-12-06T18:55:21 bugnuts: your bot still beat mine most of the time, just happy with that rare game where i won. :) 2011-12-06T18:56:17 Fluxid: yes. i'm still tweaking it though. my problem now is that a lot of ants get stuck defending in just one direction 2011-12-06T18:56:19 I think the version I am running there is actually sort of irrationally aggressive, which means I lose a lot but high variance 2011-12-06T18:57:35 it is still very strong though. it tends to dominate the game by sheer number. 2011-12-06T18:58:04 i've added a little aggressivness today 2011-12-06T18:58:22 really small and dumb aggressive neuron 2011-12-06T18:59:17 http://fluxid.pl/misc/2011-12-07-005909_285x370_scrot.png holy crap, i can't believe it 2011-12-06T18:59:34 Fluxid: i'm wanting to limit the defense to 2 layers of ants. right now they just all piled up defending layer over layer 2011-12-06T18:59:37 maybe i'll just stop tsting with old version and just use main tcp server 2011-12-06T19:00:23 2 layers? good, i'll be able to make advantage if this tomorrow (it sits on my TODO list for a while) ;) 2011-12-06T19:00:35 take* of this* 2011-12-06T19:03:09 *** ikaros has joined #aichallenge 2011-12-06T19:03:32 i can't figure out yet how to limit it to 2 layers though. 2011-12-06T19:03:41 *** jab_bott has joined #aichallenge 2011-12-06T19:04:02 um.. were a load of maps deleted? I just merged in epsilon and a load have gone! 2011-12-06T19:04:18 yep a bunch were deleted and new ones generated 2011-12-06T19:04:30 but that was a while ago (maybe it happened again?) 2011-12-06T19:04:46 why are the old ones gone though? why not just add more? 2011-12-06T19:05:09 *** tobel has quit IRC (Quit: ERC Version 5.3 (IRC client for Emacs)) 2011-12-06T19:05:35 and why not just re-use the map names and parameters (such as size)? 2011-12-06T19:06:32 for example if the map filename was still present that I am using, but it's been regenerated, then no problem.. but the parameters could stay the same 2011-12-06T19:06:35 the old maps did not meet the specs 2011-12-06T19:07:27 *** grwip has joined #aichallenge 2011-12-06T19:07:33 this is so annoying - I was using a 196x196 map in the cell maze range (p04_01) and it's disappeared and there's no 196x196 map to replace it 2011-12-06T19:10:02 if a map is not up to spec, why not just regenerate a new one over the top but with the same name and size and whatever other parameters may be deemed important? 2011-12-06T19:10:06 That's because 196x196 is too big 2011-12-06T19:10:12 the max size was lowered 2011-12-06T19:10:32 oh, I thought it was 200x200 - that's why I was looking for a nice big map to test on 2011-12-06T19:10:53 the maximum area is now 25000 2011-12-06T19:11:34 So you can have 200x125 or 158x158 2011-12-06T19:13:47 which is the biggest map now? 2011-12-06T19:15:49 is cell_maze_p08_04 still valid? if so I think thats it 2011-12-06T19:17:25 jab_bott: you can generate your own large maps if you want 2011-12-06T19:21:26 *** treeform_ has joined #aichallenge 2011-12-06T19:21:40 I'm grepping for area in the map files and for example cell_maze_p06_13.map says area 24663, then later on it says 23760 - what's going on there? 2011-12-06T19:22:21 water vs full map? 2011-12-06T19:22:35 uuugh! I was hoping to get this vis update done tonight and go to bed, but now I'm gonna have to look into maps a bit so I understand them 2011-12-06T19:23:43 *** Mystret has quit IRC (Ping timeout: 265 seconds) 2011-12-06T19:23:47 https://gist.github.com/c3f873470c3d5acc4bb7 is maps sorted by size 2011-12-06T19:24:22 *** treeform has quit IRC (Ping timeout: 248 seconds) 2011-12-06T19:24:26 bluegaspode: absulute path is a bad idea 2011-12-06T19:25:05 bluegaspode: as i said before, the jar file is in the same folder as the rest of your files 2011-12-06T19:25:18 so all you have to do is load your file from the current directory 2011-12-06T19:27:07 McLeopold: I wouldn't mind say 1/3 or 1/4 of minimum player area as minimum hill area 2011-12-06T19:27:15 ok cheers - I guess I'll use the cell_maze_p08_04 for testing as that looks to be the biggest one < 25k 2011-12-06T19:27:21 I think I just struck the chord 2011-12-06T19:28:34 *** bluegaspode has quit IRC (Ping timeout: 265 seconds) 2011-12-06T19:28:43 janzert: can you expand on that? 2011-12-06T19:28:54 I was thinking in terms of hills per land area 2011-12-06T19:29:21 so, 1/3 would be 300 land per hill? 2011-12-06T19:29:38 *** replore has joined #aichallenge 2011-12-06T19:29:43 don't remember what the actual number we decided for a player actually :) 2011-12-06T19:29:53 900-5000 area per player 2011-12-06T19:29:58 including land and water 2011-12-06T19:30:15 hmm 2011-12-06T19:31:00 hrm.. in playgame.py I can do --fill and it says it fills extra hills with the last bot - how do I get it so that if there's 4 bots and 8 hills then each player has 2 hills? 2011-12-06T19:31:35 jab_bott: you have to specify all 8 bots 2011-12-06T19:32:03 *** amstan has quit IRC (Ping timeout: 252 seconds) 2011-12-06T19:32:10 300 minimum for a hill might be low still? 2011-12-06T19:32:20 yeah 2011-12-06T19:32:32 McLeopold: really? I'm sure I've seen multiple hills per player on games on the server 2011-12-06T19:32:35 maybe 500 2011-12-06T19:33:01 jab_bott: you don't get to specify bots for hills, the multi-hills are already set for the number of players 2011-12-06T19:33:29 McLeopold: ahh - so it's a special map setting, like a 4p map but with 8 hills? 2011-12-06T19:33:43 yeah, or some maps have 2 hills per player 2011-12-06T19:33:47 it depends on the map 2011-12-06T19:34:10 janzert: at 500, any maps with 900 area per player are garrenteed to have 1 hill 2011-12-06T19:34:16 I see, ta 2011-12-06T19:34:28 yes 2011-12-06T19:35:03 what's the density of my exampmle map... 2011-12-06T19:35:18 304 :) 2011-12-06T19:35:25 so 300 is probably too low 2011-12-06T19:35:31 just was figuring it out 2011-12-06T19:36:03 304 2011-12-06T19:36:11 :) 2011-12-06T19:36:22 yeah, 500 sounds good 2011-12-06T19:36:40 that give an upper bound of 10 hills per player :) 2011-12-06T19:37:05 that's more than I want to see anyway I think ;) 2011-12-06T19:37:29 I wouldn't mind an upper bound of ~5 but not real strong on that 2011-12-06T19:37:32 the probability that I generate another hill is 50% 2011-12-06T19:37:41 so, it's like flipping 10 heads 2011-12-06T19:37:47 ok 2011-12-06T19:37:58 I think that is pareto, isn't it? 2011-12-06T19:38:16 *** epicmonkey has quit IRC (Ping timeout: 244 seconds) 2011-12-06T19:38:27 with a specific parameter 2011-12-06T19:38:29 flip coins until you get a tail, your random number is the number of flips 2011-12-06T19:38:38 I think anyway 2011-12-06T19:44:06 exponential distribution 2011-12-06T19:45:44 um, I think it's a converging geometric progression - I forget all that maths stuff! 2011-12-06T19:45:45 P(hills) = 2^-hills. that's even normalized. 2011-12-06T19:46:10 yes, the sum from 1 to infinity of 2^-x is 1 2011-12-06T19:46:29 (by the geometric series) 2011-12-06T19:46:49 a1k0n: so, not pareto? 2011-12-06T19:47:43 hmm. it might be a trivial pareto distribution 2011-12-06T19:49:21 *** xathis has joined #aichallenge 2011-12-06T19:49:38 http://en.wikipedia.org/wiki/Pareto_distribution#Relation_to_the_exponential_distribution 2011-12-06T19:50:07 *** Antimony has quit IRC (Ping timeout: 240 seconds) 2011-12-06T19:50:46 i guess the minimum of 1 hill makes it technically not an exponential distribution 2011-12-06T19:50:57 but, whatever. it still is. 2011-12-06T19:51:37 it's the probability of extra hills then :) 2011-12-06T19:51:39 hey, the rankings ordered by mu instead of skill now? 2011-12-06T19:51:59 @rankings 2011-12-06T19:52:00 McLeopold: Top 10 players: pguillory(92.1), xathis(91.3), ChrisH(89.3), delineate(87.6), lazarant(87.1), ThisIsNotABug(87.1), RVeerdonk(86.7), GreenTea(86.4), Migi32(85.9), BenJackson(85.7) 2011-12-06T19:52:09 oh my mistake, sorry 2011-12-06T19:53:00 xathis: so, how far can I get without combat logic? 2011-12-06T19:54:01 probably top 100 2011-12-06T19:54:15 btw, are the final rankings going to be by mu? 2011-12-06T19:55:45 will all bots be run in the contest, or will it pick the top at a certain date and play just those? 2011-12-06T19:55:57 *** grwip has quit IRC (Quit: Leaving) 2011-12-06T19:56:17 *** analyst74 has quit IRC (Ping timeout: 252 seconds) 2011-12-06T19:56:28 janzert: how do you feel about the winner needing to have a min skill larger than the next players max skill? 2011-12-06T19:56:35 *** xathis has quit IRC (Ping timeout: 265 seconds) 2011-12-06T19:56:57 I'm not going to guarantee it 2011-12-06T19:57:14 be nice and also not surprising though 2011-12-06T19:57:27 look for the highest mu, the find all other max skills in that range, everybody wins 2011-12-06T19:58:06 I'd rather keep the conservative ranking just like we have 2011-12-06T19:58:12 and we need a single winner 2011-12-06T19:58:23 conservative ranking = skill? 2011-12-06T19:58:31 or, take that set of players and run games until sigma drops 2011-12-06T19:58:33 then you are very dependant on correct matching algorithm 2011-12-06T19:58:49 because the winner could have lower mu but out-sigma'ed the 2nd place 2011-12-06T19:59:18 McLeopold: that would be even worse 2011-12-06T19:59:33 since you just drastically changed the distribution of opponents 2011-12-06T19:59:50 that's like the cutoff you did last year 2011-12-06T20:00:47 but way way more severe 2011-12-06T20:01:21 janzert: also, for finals, I thought the best method would be to lock submissions, then "resubmit" every users final bot as a "final submission" 2011-12-06T20:01:38 it automatically resets all sigma/mu values for everyone, without messing with the database 2011-12-06T20:01:40 one of the limits on how far the cutoff can be taken is so it doesn't effect the opponents the top bots are playing 2011-12-06T20:01:52 in that case, being the first one submitted should be an advantage 2011-12-06T20:01:57 you'll get more matches, so lower sigma 2011-12-06T20:02:42 garf they are are essentially all submitted at the same time 2011-12-06T20:02:52 *** jab_bott has quit IRC (Ping timeout: 265 seconds) 2011-12-06T20:02:57 ok 2011-12-06T20:03:07 McLeopold: doesn't really matter to me, the skill reset is already implemented and tested though 2011-12-06T20:03:10 still, what advantage has a final ranking based on skill instead of mu? 2011-12-06T20:03:33 if you don't count sigma you basically lose your variance measure 2011-12-06T20:03:43 and that matters how? 2011-12-06T20:03:48 the winner should be the strongest bot 2011-12-06T20:03:54 tells you how consistently good the bot is 2011-12-06T20:03:58 which is only defined by mu 2011-12-06T20:04:11 but may be bad more often 2011-12-06T20:04:15 no 2011-12-06T20:04:22 not really. if the top two bots are 0.1 mu different but both have sigma=2, then it's pretty hard to tell who's actually better 2011-12-06T20:04:25 on average it must be better 2011-12-06T20:04:34 Garf: http://pastebin.com/EwKG2ptb 2011-12-06T20:04:53 a1k0n: that's true, but the one with biggest mu still has the highest likelihood 2011-12-06T20:04:57 *** Antimony has joined #aichallenge 2011-12-06T20:05:07 at some point, there is no longer a correlation 2011-12-06T20:05:14 McLeopold: your point is that all sigma are very close? 2011-12-06T20:05:17 we will obviously run enough games so it is not a factor 2011-12-06T20:05:21 yes, but they could switch places with a game or two 2011-12-06T20:05:34 no, my point is that sigma and game_count are not correlated 2011-12-06T20:05:39 *** ikaros has quit IRC (Quit: Ex-Chat) 2011-12-06T20:05:41 sigma is ordered, game count varies 2011-12-06T20:05:59 well, they're definitely correlated towards game_count < 10 2011-12-06T20:06:02 uh, if they're not correlated, you have a HUGE bug 2011-12-06T20:06:12 it's just noisy 2011-12-06T20:06:16 right 2011-12-06T20:07:09 Garf: I don't think you understand trueskill 2011-12-06T20:07:15 there is no sound reason to declar highest skill = winner. For example, the sigma*x factor (which is 3 IIRC) is very arbitrary there 2011-12-06T20:07:32 I understand it defines strenght as a mean and a variance 2011-12-06T20:07:51 variance decreases with more information and increases with unexpected results 2011-12-06T20:08:48 You're proposing to order based on confidence margins. I'm pointing out that regardless of confidence margins, the highest mean (of the sampled matches) is most likely to have the highest mean (for an infinite amount of games). 2011-12-06T20:09:23 *** lisson has joined #aichallenge 2011-12-06T20:09:38 TrueSkill uses lower estimates to avoid players with only a few games appearing on top of rating lists. 2011-12-06T20:09:44 *** lisson has quit IRC (Client Quit) 2011-12-06T20:09:54 This is a human psychology consideration, not statistics. 2011-12-06T20:10:15 TrueSkill provides 2 values, using mu-sigma*3 provides conservitive rankings 2011-12-06T20:10:34 I agree that mu, without sigma, should declare the winner 2011-12-06T20:10:42 we need to convince janzert 2011-12-06T20:10:48 ok, good :P 2011-12-06T20:11:05 I'l also like to prove somehow that the "winners" skill range is well above the others 2011-12-06T20:11:23 so, the winner must have the highest max skill as well 2011-12-06T20:11:38 That depends on the submissions. 2011-12-06T20:11:38 *** Anilm3 has quit IRC (Quit: Lost terminal) 2011-12-06T20:11:39 McLeopold: there's no way we can guarantee we'll have enough games to do that 2011-12-06T20:11:41 and if we declare multiple winners, I'm okay with that 2011-12-06T20:12:02 *** Anilm3 has joined #aichallenge 2011-12-06T20:12:14 anyone who's upper bound is inside the top entries lower bound wins? 2011-12-06T20:12:37 maybe something like that, idk 2011-12-06T20:12:44 (I think you can apply the same criticism to that as I just did for skill vs mu) 2011-12-06T20:13:06 but if there isn't 1 person with both the high mu and high mu+sigma*3, I don't think we can pick 1 person 2011-12-06T20:13:41 ideally, all sigmas are very close 2011-12-06T20:13:56 you can compute likelihood of dominance 2011-12-06T20:14:13 ranking for the whole contest has been based on skill, I really don't want to change the for the finals 2011-12-06T20:14:35 the finals are different because the bots don't change any more 2011-12-06T20:14:46 bots don't change now 2011-12-06T20:14:53 having a single winner is major part of the competition, even if I gave in I think Jeff would have a huge problem with it 2011-12-06T20:14:53 *** ltriant has quit IRC (Read error: Connection reset by peer) 2011-12-06T20:14:58 janzert: I don't agree, we have obvious mu and skill values 2011-12-06T20:15:29 *** ltriant has joined #aichallenge 2011-12-06T20:15:51 we only use conservitive so that you naturally climb the ranks rather than bounce around 2011-12-06T20:16:08 exactly 2011-12-06T20:16:20 brb 2011-12-06T20:16:22 that's not the primary reason to use a conservative skill 2011-12-06T20:16:43 I don't really care at all if new entries bounced around 2011-12-06T20:16:52 conservative is used because otherwhise, for example in Halo, you could get lucky in the first few games, get a high mu, and then be effectively be discouraged from playing more, because its likely to drop your skill 2011-12-06T20:17:06 If anyone is interested I put all the new hill data in a google fusion table https://www.google.com/fusiontables/DataSource?dsrcid=2359966 2011-12-06T20:17:19 the conservative formula ensures that your skill slowly grows upwards 2011-12-06T20:17:37 and you don't "accidentally" get ranked too high 2011-12-06T20:17:37 it also says, we're pretty confident you're at least this good 2011-12-06T20:17:39 <_flag> How come elo seemed to be so definitive and trueskill varies so much? Is it just because of the volitility of the game itself? 2011-12-06T20:17:53 this isn't a relevant consideration if we're at the point the matches are automatic and the bots don't change 2011-12-06T20:17:57 elo gives a false pretense of confidence :) 2011-12-06T20:18:05 at least the way displayed 2011-12-06T20:18:07 yeah, elo internally had mu and variance 2011-12-06T20:18:13 if you'd only look at mu, elo would be very similar 2011-12-06T20:18:54 with elo, you effectively choose your own sigma for each match, based on the match type 2011-12-06T20:19:00 aka, the K value 2011-12-06T20:19:19 glicko is better, but still only does 2 players 2011-12-06T20:19:39 glicko for 2 players is extremely similar to trueskill 2011-12-06T20:19:55 you can use glicko for multiplayer games btw, but I think the statistical foundation is weaker 2011-12-06T20:20:08 well tron was using bayeselo by the end 2011-12-06T20:20:10 because you must assume strength adds linearly (in the log domain that ELO operates in) 2011-12-06T20:20:39 trueskill makes the same assumption 2011-12-06T20:21:33 I admit not knowing where the fine differences lie then 2011-12-06T20:22:02 *** Anilm3 has quit IRC (Quit: leaving) 2011-12-06T20:22:10 *** treeform_ has quit IRC (Remote host closed the connection) 2011-12-06T20:22:21 *** Anilm3 has joined #aichallenge 2011-12-06T20:22:28 trueskill uses the sum-product algorithm to compute new values, making certain assumptions for each game where if you rank higher than someone than your skill+performance should be >theirs+epsilon 2011-12-06T20:22:29 trueskill supports more than 2 players/teams 2011-12-06T20:23:09 115 winners by my definition :( http://pastebin.com/AJM3DLJd 2011-12-06T20:23:26 *** retybok_ has quit IRC (Ping timeout: 252 seconds) 2011-12-06T20:23:40 all submissions with mu+sigma*3 within range of the leaders mu-sigma*3 2011-12-06T20:23:55 can you filter bots with < 20 or so games? 2011-12-06T20:24:05 or whatever is realistic to play in the final? 2011-12-06T20:24:58 really, you need to compute likelihood of dominance of the #1 guy but i don't think mysql has erf 2011-12-06T20:31:53 *** Antimony has quit IRC (Ping timeout: 244 seconds) 2011-12-06T20:32:15 *** Antimony has joined #aichallenge 2011-12-06T20:36:58 *** Anilm3 has quit IRC (Quit: Lost terminal) 2011-12-06T20:58:37 *** b0rder has joined #aichallenge 2011-12-06T20:59:25 *** AntDroid_ has joined #aichallenge 2011-12-06T21:07:19 *** Apophis_ has joined #aichallenge 2011-12-06T21:07:19 *** Apophis has quit IRC (Read error: Connection reset by peer) 2011-12-06T21:10:09 *** treeform has joined #aichallenge 2011-12-06T21:12:37 *** Nicka_ has joined #aichallenge 2011-12-06T21:12:46 *** ace1010 has joined #aichallenge 2011-12-06T21:16:46 hi there 2011-12-06T21:17:14 Is anyone seeing a compile error while the server compiles your C# submission? I'm using the same submission that worked last night and the server is saying the following. Error CS2007: Unrecognized command-line option: `--optimize=all' 2011-12-06T21:17:18 I'm currently getting the following error when submitting my c# bot: error CS2007: Unrecognized command-line option: `--optimize=all' 2011-12-06T21:18:01 yeah, me too Nicka 2011-12-06T21:18:13 I think there's a problem with their server 2011-12-06T21:18:35 *** Jak_o_Shadows has quit IRC (Remote host closed the connection) 2011-12-06T21:18:48 yeah, sounds like they broke the compiler 2011-12-06T21:18:55 *** Jak_o_Shadows has joined #aichallenge 2011-12-06T21:20:46 ace1010, agreed 2011-12-06T21:22:51 SO nobody knows anything about this? 2011-12-06T21:24:29 i think this is the culprit: https://github.com/aichallenge/aichallenge/commit/9008cf88cce220530b54f39a024df736300ecb87 2011-12-06T21:25:52 interesting. good find ace 2011-12-06T21:26:01 capitalization error? 2011-12-06T21:29:36 is there a way to make gmail log out all your past sessions or whatever liek with facebook? 2011-12-06T21:29:50 pretty sure i left my email logged in on a professors computer at the uni 2011-12-06T21:29:54 not a huge issue, but still.. 2011-12-06T21:31:25 under account activity (lower right corner) there is a signout all other sessions (before your prof browses your midget horse porn) 2011-12-06T21:32:02 cheers :) 2011-12-06T21:32:28 ace1010 im about to make a forum post of the CS2007 issue 2011-12-06T21:33:22 *** jab_bott has joined #aichallenge 2011-12-06T21:33:42 \o/ another visualizer feature implemented 2011-12-06T21:34:44 *** ace1010 has quit IRC (Quit: Page closed) 2011-12-06T21:35:06 jab_bott you might want to switch your default branch on github to be vis_overlay 2011-12-06T21:35:34 bugnuts: ok - I didn't know you could do that! 2011-12-06T21:36:01 yeah it's under admin for a given repo 2011-12-06T21:36:10 btw vis stuff is awesome. 2011-12-06T21:36:16 bugnuts: re. the fog - I think something in the visualizer gui to change the fog layer is more appropriate than something in the bot 2011-12-06T21:37:36 ok default branch changed 2011-12-06T21:38:26 glad to be learning new stuff - I'd never done any python or javascript before looking at this stuff, or github.. learned so much - life feels better when you're learning 2011-12-06T21:41:38 right, I'm off to bed... enjoy the update :) 2011-12-06T21:42:02 *** jab_bott has left #aichallenge 2011-12-06T21:46:07 *** Nicka_ has quit IRC (Quit: Page closed) 2011-12-06T21:46:55 *** Antimony has quit IRC (Ping timeout: 244 seconds) 2011-12-06T21:53:27 *** suid has quit IRC (Ping timeout: 258 seconds) 2011-12-06T21:54:03 *** suid has joined #aichallenge 2011-12-06T22:01:47 *** Paradoxial has quit IRC (Read error: Connection reset by peer) 2011-12-06T22:04:43 amstan: ^^ 2011-12-06T22:04:54 oops was scrolled back 2011-12-06T22:05:08 amstan: looks like C# compilation is broken now 2011-12-06T22:10:53 *** Jak_o_Shadows has quit IRC (Read error: Connection reset by peer) 2011-12-06T22:12:24 *** treeform_ has joined #aichallenge 2011-12-06T22:15:08 *** treeform has quit IRC (Ping timeout: 240 seconds) 2011-12-06T22:18:37 For finals cutoff, don't you just deactivate everyone with mu+sigma*3 < mu_top-sigma_top*3, then continue running games for the remainder *drawing opponents from the full pool* until sigma drops? 2011-12-06T22:19:42 I like that 2011-12-06T22:21:58 what does that do? 2011-12-06T22:22:04 * antimatroid likes that everyone is reset to the same thing 2011-12-06T22:22:17 i would be quite opposed to doing anything that didn't have everyone starting exactly the same 2011-12-06T22:22:30 well, it would naturally limit the active players, without us picking an arbitrary number 2011-12-06T22:22:40 McLeopold: we need to change "rank stabilised", that's not correct 2011-12-06T22:22:46 oh, assume we reset everyone at once 2011-12-06T22:22:54 it's more like "no participating bot can increase in ranking" 2011-12-06T22:22:59 antimatroid: I'm not going to change it 2011-12-06T22:23:00 antimatroid: It avoids janzert's concern that a straight cut-off would drastically alter the set of opponents 2011-12-06T22:23:14 McLeopold: it's not right though 2011-12-06T22:23:15 right, cause you can still draw from the full set 2011-12-06T22:23:26 antimatroid: it's word smithing, too late for that 2011-12-06T22:23:45 Though presumably the match-making will mean that by that time, those bots are going to be pretty much only playing each other anyway. 2011-12-06T22:24:30 the matching making still picks a random number for the rank distance 2011-12-06T22:24:47 rank in match quality, not mu or skill 2011-12-06T22:28:53 i don't get this mu thing? players end up equally ranked right? 2011-12-06T22:29:27 mu-3sigma is a number you're almost certainly better than 2011-12-06T22:29:33 mu+3sigma is a number you're almost certainly less than 2011-12-06T22:29:54 so using those as the cutoffs means you keep any bot that is remotely possible of being as good at its best as the best are at their worst 2011-12-06T22:36:19 but aren't you still starting bots with different rankings? 2011-12-06T22:51:27 @later tell jab_bott the latest visualizer broke a previous reply and doesn't show lines :( 2011-12-06T22:51:27 McLeopold: Ready to serve. 2011-12-06T23:04:14 *** u_ has quit IRC (Quit: u_) 2011-12-06T23:06:11 *** roflmao has quit IRC (Read error: Connection reset by peer) 2011-12-06T23:15:24 *** skot84 has joined #aichallenge 2011-12-06T23:20:05 *** yoden has quit IRC (Quit: Leaving.) 2011-12-06T23:21:57 *** yoden has joined #aichallenge 2011-12-06T23:23:01 *** antimatroid1 has joined #aichallenge 2011-12-06T23:23:07 *** antimatroid has quit IRC (Read error: No route to host) 2011-12-06T23:25:44 *** TheLinker has quit IRC (Quit: Bye) 2011-12-06T23:26:00 *** yoden has quit IRC (Client Quit) 2011-12-06T23:44:53 who *are* you, bruno5? 2011-12-06T23:47:49 *** avdg has quit IRC (Quit: Leaving.) 2011-12-06T23:47:57 *** raemde_ has joined #aichallenge 2011-12-06T23:48:51 *** JorgeB has joined #aichallenge 2011-12-06T23:51:07 *** raemde has quit IRC (Ping timeout: 240 seconds) 2011-12-06T23:51:18 *** ademar has joined #aichallenge 2011-12-06T23:52:32 *** amstan has joined #aichallenge 2011-12-06T23:52:32 *** ChanServ sets mode: +o amstan 2011-12-06T23:53:17 janzert: we have a problem: http://forums.aichallenge.org/viewtopic.php?f=4&t=2062 2011-12-06T23:53:52 yeap .. something weird is happening on my profile as well .. 2011-12-06T23:54:49 ademar: related to c#? 2011-12-06T23:55:11 yes my bot is C# as well 2011-12-06T23:55:18 amstan: I would guess it's related to "added optimization to C# as per request" suppose the author of that should be able to fix it ;) 2011-12-06T23:55:28 ademar: do you have linux? 2011-12-06T23:55:37 yes i do 2011-12-06T23:55:39 *** Antimony has joined #aichallenge 2011-12-06T23:55:41 janzert: yeah, the flag was wrong 2011-12-06T23:55:47 ademar: what's the right flag for optimizations? 2011-12-06T23:56:11 apparently --optimize=all is not it 2011-12-06T23:56:12 lemme look it up .. never used then before 2011-12-06T23:56:48 -optimize 2011-12-06T23:56:51 or 2011-12-06T23:56:52 -o 2011-12-06T23:57:28 ok it is 2011-12-06T23:57:32 *** delt0r_ has quit IRC (Read error: Operation timed out) 2011-12-06T23:57:38 -o+ 2011-12-06T23:58:07 mcs Ants.cs Bot.cs GameState.cs Location.cs MyBot.cs Tile.cs Log.cs -out:MyBot.exe -o+ 2011-12-06T23:58:38 what's the diff between -o+ and -o? 2011-12-06T23:58:52 nvm 2011-12-06T23:58:54 it's the same 2011-12-06T23:58:55 k 2011-12-06T23:59:06 *** teh0ne has joined #aichallenge 2011-12-06T23:59:19 the correct is -o+ 2011-12-06T23:59:31 the first one failed here 2011-12-06T23:59:50 from the help file 2011-12-06T23:59:52 -optimize[+|-] Enables advanced compiler optimizations (short: -o)