2011-02-07T00:15:38 aichallenge: Alexandru Stan epsilon * r760f141 / (25 files in 4 dirs): Merge branch 'simpleants' into HEAD - http://bit.ly/g5vRIs 2011-02-07T00:15:39 aichallenge: Alexandru Stan epsilon * r872cd42 / (Makefile engine.py): made it run - http://bit.ly/h4Gkmu 2011-02-07T00:15:42 aichallenge: Alexandru Stan epsilon * r592ffb1 / (26 files in 4 dirs): merged stuff from mcleopold - http://bit.ly/gIDYch 2011-02-07T00:16:21 amstan: did you just check in an ants engine? Or is that the old planet wars engine? 2011-02-07T00:16:46 j3camero: that was in epsilon 2011-02-07T00:16:52 so.. both 2011-02-07T00:16:59 we need to clean epsilon up 2011-02-07T00:17:04 delete all those games that we're not using 2011-02-07T00:17:10 move over the engine from planetwars 2011-02-07T00:17:12 delete that as well 2011-02-07T00:18:09 Yeah there's so much stuff in there that is not used. 2011-02-07T00:19:04 I have been rethinking the epsilon/zeta stuff. It seems like we might have enough folks around that we could race right into zeta. 2011-02-07T00:19:13 Just sprint for the finish. 2011-02-07T00:19:38 If we're doing ants we need to change all the ranking and scheduling stuff anyways. 2011-02-07T00:19:45 Might as well go whole-hog. 2011-02-07T00:20:31 j3camero: we can do that at any time 2011-02-07T00:21:16 contestbot: later tell mcleopold I sent you a pm on github with some notices about your simpleants stuff, please check 2011-02-07T00:21:16 amstan: I come to serve. 2011-02-07T00:21:34 ok.. now i feel organized 2011-02-07T00:21:45 except the stuff from zeta 2011-02-07T00:21:48 epsilon* 2011-02-07T00:21:49 lol 2011-02-07T00:22:25 j3camero: tron is epic coding music 2011-02-07T00:22:41 j3camero: also: https://github.com/aichallenge/aichallenge/wiki 2011-02-07T00:28:01 Yeah I am listening to Tron right now! 2011-02-07T00:28:05 It's soooo good for coding. 2011-02-07T00:29:52 j3camero: there we go: https://github.com/aichallenge/aichallenge/issues/issue/1 2011-02-07T00:32:48 j3camero: it might be better not to rush zeta? 2011-02-07T00:33:13 zeta would take a month to get out 2011-02-07T00:33:31 unless we all start coding like crazy on stuff we're not comfortable with 2011-02-07T00:34:13 You think we can do zeta in a month? 2011-02-07T00:34:21 That seems optimistic. 2011-02-07T00:34:37 if we have at least 4 coders, which we do 2011-02-07T00:34:39 yeah 2011-02-07T00:35:05 Hofstadter's law - "It always takes longer than you expect, even when you take into account Hofstadter's Law." 2011-02-07T00:35:23 *back to lurk mode* 2011-02-07T00:35:31 Hm okay let's do it. 2011-02-07T00:35:42 what's going on? 2011-02-07T00:35:48 I don't think it's going to take a month, but nevertheless we could do it. 2011-02-07T00:35:50 well.. our previous contests, the problem was that it was only a like 2 ppl working at the same time 2011-02-07T00:36:11 Yeah. I burned myself out in a big way last time. 2011-02-07T00:36:21 you didn't have help 2011-02-07T00:36:31 i'm onboard, i see jbroman's interested 2011-02-07T00:36:36 Assigning more programmers to a project running behind schedule will make it even later 2011-02-07T00:36:38 and he's doing a cool job too 2011-02-07T00:36:44 What was amazing was tron. I had mspang helping me out the whole time. 2011-02-07T00:36:50 - Fred Brooks 2011-02-07T00:37:00 janzert: lol 2011-02-07T00:37:04 janzert: we're not behind schedule though, we knew this would happen 2011-02-07T00:37:17 you can't be behind schedule when there isn't one :P 2011-02-07T00:37:30 it's now february, how late are you planning to launch? 2011-02-07T00:37:36 somebody said february though ;) 2011-02-07T00:37:42 Yeah. The AI Contest is like Duke Nukem Forever. When will it be released? When it's done. 2011-02-07T00:37:43 that's why i said when it is :P 2011-02-07T00:37:47 janzert: i had to say something.. lol 2011-02-07T00:38:29 also.. i might be getting ipv6 soon, that would be cool 2011-02-07T00:39:23 whoa, jbroman really did a lot on zeta! 2011-02-07T00:39:35 i know! lol, and good pythonic stuff too 2011-02-07T00:40:07 Orly? 2011-02-07T00:40:14 j3camero: https://github.com/aichallenge/aichallenge/tree/zeta/worker 2011-02-07T00:41:39 ugh, tabs in the source. please consider sticking to pep 8 style 2011-02-07T00:42:18 i do like tabs, but yeah.. 2011-02-07T00:43:32 there is something that needs to be done with respect to when submissions close 2011-02-07T00:43:35 well, i guess i can stop working on my own engine now 2011-02-07T00:44:05 the server should close submissions, then attempt to compile every submission in the queue, it doesn't currently do that 2011-02-07T00:44:58 antimatroid: you're referring to zeta or what was used? 2011-02-07T00:45:08 the last contest did do that 2011-02-07T00:45:21 janzert: i'm just saying at some point that needs to be fixed 2011-02-07T00:45:25 if it's left too late, it wont happen 2011-02-07T00:45:33 antimatroid: which branch? 2011-02-07T00:45:36 Oh wow jbroman and Zannick are awesome. 2011-02-07T00:45:38 what's broken though? 2011-02-07T00:45:48 that is the behaviour that was used in the last contest 2011-02-07T00:45:59 although it all had to be done manually of course 2011-02-07T00:46:14 there was a single submission that didn't get compiled yeah? 2011-02-07T00:46:29 i think that's undesirable, if it's submitted before deadlines are closed off, it should be given a chance to compile 2011-02-07T00:46:38 no, they all were attempted to be compiled 2011-02-07T00:46:52 oh, i thought you said there was one that wasn't, my mistake :) 2011-02-07T00:47:09 I think you're remembering that the final (or penultimate) submission failed in the compilation 2011-02-07T00:47:28 I don't remember for sure which, but it did try to get compiled 2011-02-07T00:48:08 i.e. the submission had a bug that caused it be uncompilable 2011-02-07T00:48:16 j3camero: we need a website design, i can code it, but i need ideas 2011-02-07T00:48:33 aichallenge: Benjamin S Wolf zeta * re8f275b / (worker/compiler.py worker/language.py): Merge branches 'master' and 'zeta' into zeta - http://bit.ly/gZsdpn 2011-02-07T00:49:36 luckily that wasn't terrible, only one commit 2011-02-07T00:49:48 now i should figure out how to get rid of 'master' 2011-02-07T00:49:58 Zannick: rename or delete? 2011-02-07T00:50:12 Zannick: rename is git branch -m source dest, delete is git branch -d todelete 2011-02-07T00:50:20 my last pull created a zeta branch 2011-02-07T00:50:53 amstan: go for it. I think the only piece where the website overlaps with the rest of the project is the website and passing submitted zip files to the backend. 2011-02-07T00:51:01 yes, zeta is the same as master, if your zeta's newer feel free to delete master 2011-02-07T00:51:05 amstan: so make sure to coordinate very closely with people on those details. 2011-02-07T00:51:32 amstan: in particular, make sure you nail down a database schema with people working on the backend. 2011-02-07T00:51:50 j3camero: yeah, but i need some design guidance, visual design 2011-02-07T00:51:57 how will the ants contest website look? 2011-02-07T00:53:04 I've always gone with a standard header/sidebar/content pane. 2011-02-07T00:53:27 colour scheme? some art? 2011-02-07T00:54:27 hm 2011-02-07T00:55:36 if i edit jbroman's files i'm either going to have to replace all the tabs or mix tabs and spaces or change my rc to use hard tabs 2011-02-07T00:55:57 Zannick: i think we should convert to spaces 2011-02-07T00:57:06 the last contest site was fine design wise 2011-02-07T00:59:55 amstan: find Alex Riedler. See if he's interested in doing the design this time. 2011-02-07T00:59:58 He did the last one. 2011-02-07T01:00:08 Before that I was using stuff I found on freewebtemplates.com 2011-02-07T01:00:33 i can do it, i just need some ideas 2011-02-07T01:00:34 It's actually not a bad way to go. You're guaranteed to end with something that looks really polished. 2011-02-07T01:00:40 yeah 2011-02-07T01:00:53 What we don't want to end up with is something that looks homemade. 2011-02-07T01:01:26 this could go well with ants: http://www.4templates.com/view/website-templates/0/AX0461CY 2011-02-07T01:01:33 maybe some more noise 2011-02-07T01:02:07 Yep looks "underground" 2011-02-07T01:02:23 Check out the extreme minimalist concept here: http://www.feedparser.org/ 2011-02-07T01:02:32 I would not mind exploring this. 2011-02-07T01:02:51 Like, even that exact design would work for us. 2011-02-07T01:07:41 In any case, do separate the look and layout from the content. We should be able to slap a new look onto the website in minutes. Take a look at some of the pages from planet wars. You see that there is nothing in them but content. All the look and layout is in the stylesheets and header.php. It would take minutes to download a different template from the web, create a new header.php, and deploy the new look. 2011-02-07T01:08:44 yeah, that's fine, i can design something like that 2011-02-07T01:09:17 in django's easy as well.. you have templates that are used by the code 2011-02-07T01:14:18 Sweet. 2011-02-07T01:20:34 omg you would not believe this 2011-02-07T01:20:39 I am trolling this chick SO HARD. 2011-02-07T01:20:40 j3camero: yes? 2011-02-07T01:20:42 I feel kinda bad. 2011-02-07T01:20:50 hmm 2011-02-07T01:20:52 do tell 2011-02-07T01:21:57 Oh on Galcon. 2011-02-07T01:22:55 I noticed this person on Galcon a few days back in the chat. 2011-02-07T01:23:13 She was always asking "hey is firstname here?" 2011-02-07T01:23:30 "firstname lastname?" 2011-02-07T01:23:43 So I didn't take much notice, happens sometimes. 2011-02-07T01:23:48 *** Mathnerd314 has quit IRC (Remote host closed the connection) 2011-02-07T01:24:07 Anyways today I see the same girl asking around. 2011-02-07T01:24:21 She's like "firstname? is firstname here?" 2011-02-07T01:28:09 Oh my goodness I wish I had this logged. 2011-02-07T01:28:15 j3camero: lol 2011-02-07T01:28:37 oh man.. mcleopold managed to merge *everything* together: https://github.com/McLeopold/aichallenge/tree/ 2011-02-07T01:28:45 zeta+epsilon+simpleants 2011-02-07T01:30:01 *** McLeopold has joined #aichallenge 2011-02-07T01:30:10 McLeopold: what did you do? lol 2011-02-07T01:30:19 McLeopold: i suggest just deleting your fork and reforking 2011-02-07T01:30:41 you merged zeto into epsilon 2011-02-07T01:30:49 zeta* 2011-02-07T01:31:19 did I? 2011-02-07T01:31:28 McLeopold: yep: https://github.com/McLeopold/aichallenge/tree/ 2011-02-07T01:31:31 that's not what they were called when I pulled 2011-02-07T01:31:41 oh man so I figure I'll bite. I said "hey yes I am firstname. who is this" 2011-02-07T01:32:19 McLeopold: yeah, you probably used the fork queue, don't use that unless you know what you're doing 2011-02-07T01:32:37 maybe I will delete, I was doing all that work to get used to github, nothing is special there 2011-02-07T01:33:10 I was only doing command line pulls, committing and then pushing to my own repo 2011-02-07T01:33:12 McLeopold: i have your code all in here: https://github.com/aichallenge/aichallenge/tree/epsilon/ants 2011-02-07T01:33:21 okay 2011-02-07T01:33:23 so she's like "this is girlname". And I say "oh awesome. nice to see you on here girlname" 2011-02-07T01:33:30 But she's suspicious. 2011-02-07T01:33:35 j3camero: lol 2011-02-07T01:33:39 lol 2011-02-07T01:34:03 k, going to bed 2011-02-07T01:34:17 isn't the fork queue for pulling in work others have done? 2011-02-07T01:34:29 does my repo look like crap? 2011-02-07T01:34:31 McLeopold: yes, but fork queue shows all the work, from all branches 2011-02-07T01:34:41 McLeopold: yes 2011-02-07T01:34:58 i just added a couple of suggestions in a Python Style Guide on the github wiki 2011-02-07T01:35:15 I hope you just went with pep8 2011-02-07T01:35:22 i wish we could have notifications for wiki changes and issues 2011-02-07T01:35:33 lol 2011-02-07T01:35:38 i get emailed for issues 2011-02-07T01:35:40 Zannick: please link in the homepage 2011-02-07T01:35:43 as i discovered today 2011-02-07T01:36:24 McLeopold: he went with google 2011-02-07T01:36:38 linked 2011-02-07T01:36:56 the google style is more or less what i do personally 2011-02-07T01:37:05 and at google :P 2011-02-07T01:37:27 Power Features 2011-02-07T01:37:28 ▶ Avoid these features. 2011-02-07T01:37:31 what does that mean? 2011-02-07T01:38:00 +1 for Modules should begin with #!/usr/bin/env python 2011-02-07T01:38:09 they should also be chmodded +x 2011-02-07T01:38:19 okay amstan I'll tell you this story tomorrow. This is just too good to be true. 2011-02-07T01:38:40 amstan: it lists a bunch of features if you click 2011-02-07T01:39:01 like grabbing code from functions, system changes 2011-02-07T01:39:04 really crazy stuff 2011-02-07T01:39:22 j3camero: how far are we on engine work? I had an idea for player perspective switching, but don't want to redo work. 2011-02-07T01:39:44 Zannick: yeah, the stuff to avoid makes sense 2011-02-07T01:40:08 I would also suggest not feeling the need to create 1 class per file, but that's just me. 2011-02-07T01:40:19 It makes it feel like java 2011-02-07T01:40:52 i don't do that 2011-02-07T01:41:04 i don't think that's in the guide anywhere, either 2011-02-07T01:43:58 Also, I tend to avoid creating classes as structs. Is there any known performace hit for that? Am I just paranoid? 2011-02-07T01:46:02 *** sigh has quit IRC (Ping timeout: 255 seconds) 2011-02-07T01:51:11 *** McLeopold has left #aichallenge 2011-02-07T01:52:22 * antimatroid loves structs 2011-02-07T01:53:31 amstan: that template was nice, although i think it's better to have the links along the side 2011-02-07T01:53:37 given the trend with wide screens 2011-02-07T01:54:54 McLeopold: I haven't done much with the engine yet. Just gonna copy the engine from PW and work from there. 2011-02-07T01:54:59 I don't think it will be that uncommon for people to take a 2:1 hit or whatever, particularly with the stupider bots 2011-02-07T01:55:03 j3camero: he left 2011-02-07T01:55:29 j3camero: i have some requests for the engine for whoever writes it 2011-02-07T01:55:57 most important is sequentially allowing each player to make their moves 2011-02-07T01:58:13 damn you mcleopold, i have things to say to you! 2011-02-07T02:00:19 Why sequential? 2011-02-07T02:00:33 And how can we keep people from pondering during other people's turns? 2011-02-07T02:01:06 sigstop and sigcont? 2011-02-07T02:02:09 I will ahve to read about these signals. 2011-02-07T02:02:23 Will the kernel just stop scheduling the program for execution? 2011-02-07T02:02:43 might be something other than sigstop, i don't recall off the top of my head 2011-02-07T02:02:50 j3camero: otherwise each player in the worst case gets timeLimit/noPlayers 2011-02-07T02:02:53 and noPlayers is variable 2011-02-07T02:03:02 i hated it before, but it's even worse with varying number of players 2011-02-07T02:06:06 Why not just give more time when there are more players? 2011-02-07T02:07:12 j3camero: i still would prefer it to be sequential 2011-02-07T02:07:26 even if it was just 2 players, i would want it to be sequential 2011-02-07T02:18:04 *** delt0r_ has joined #aichallenge 2011-02-07T02:19:24 *** delt0r___ has quit IRC (Ping timeout: 246 seconds) 2011-02-07T02:19:36 Why? Is it an issue of getting fair amount of CPU time? 2011-02-07T02:20:59 j3camero: yes 2011-02-07T02:21:05 TIL - asians squint in reverse 2011-02-07T02:21:49 During Tron mspang and I did a bit of investigation and we found it wasnt usually a big deal. 2011-02-07T02:21:58 The CPU time is generally pretty fair 2011-02-07T02:22:47 i'd be interested to know the variation with computationally expensive bots playing each other, particularly if there's like 16 of them 2011-02-07T02:23:13 it really shouldn't change the computation time that much 2011-02-07T02:23:42 if each player is given say 0.5 seconds for their moves, then it's not much different to giving 0.5*n seconds per turn 2011-02-07T02:23:47 and people could infer the number of players from that 2011-02-07T02:24:05 We could try doing some SIGSTOP and SIGCONT magic, but it's going to be harder than it seems. Many bots spawn a few processes. 2011-02-07T02:24:18 isn't that forbidden? 2011-02-07T02:24:30 or are these those green processes someone was talking about? 2011-02-07T02:24:32 No a small number is allowed. 2011-02-07T02:24:33 the jvm does it no matter what 2011-02-07T02:25:01 We tell people not to use fork or threads, but in practice they can get away with a little bit. 2011-02-07T02:25:03 * antimatroid has yet another reason to hate java 2011-02-07T02:25:25 Yeah the JVM is balls. It uses 10 threads just to run Hello World. 2011-02-07T02:25:57 I'd say it'd be easiest/nicest compromise if we can afford it to just run the moves simultaneously with a maximum one bot per core 2011-02-07T02:26:39 but that isn't terribly efficient and pretty much limits the maximum players to 4 with EC2 2011-02-07T02:26:58 you can't vary the time limit for the number of players so each gets ~0.5 seconds and they don't know the number of players 2011-02-07T02:27:11 right 2011-02-07T02:27:19 that presents a problem right there 2011-02-07T02:27:30 i could survive with it like it has been, but how do we get around that? 2011-02-07T02:27:33 * janzert doesn't like a varying timelimit in any case 2011-02-07T02:28:11 even sigstop/sigcnt or whatever will probably give away the number of players 2011-02-07T02:28:14 if you don't vary the overall limit, you vary the individual *limit* 2011-02-07T02:28:22 just look at the amount of realtime since your last turn 2011-02-07T02:28:40 i don't think so, only if everyone used all their time 2011-02-07T02:28:45 which most people don't 2011-02-07T02:29:03 j3camero: we might need to tell people how many players there are 2011-02-07T02:29:25 *** delt0r___ has joined #aichallenge 2011-02-07T02:29:41 i'm wondering if maybe the game would be better without fog of war at all 2011-02-07T02:29:51 you already have the problem of playing against n-1 other players 2011-02-07T02:30:45 *** delt0r_ has quit IRC (Ping timeout: 240 seconds) 2011-02-07T02:38:09 *** sigh has joined #aichallenge 2011-02-07T02:38:38 i love how people claim economics is fail because they can't forecast stuff long in advance 2011-02-07T02:39:19 antimatroid: yeah hiding the number of players will be hard because of the time issues. 2011-02-07T02:40:08 j3camero: what are your thoughts on fog of war? 2011-02-07T02:40:22 originally i was all for it, but with n players i'm wondering if it might be more interesting to have people see what is happening 2011-02-07T02:40:39 it's still not at all trivial as to whether you should try to take someone out or not 2011-02-07T02:41:06 One solution is to give them the same amount of time regardless of the number of players. That way you would not be able to squeeze as much CPU time per turn out during games with more players. 2011-02-07T02:41:33 The issue with not having the fog of war is that it will produce deadlocks. 2011-02-07T02:41:47 deadlocks how? 2011-02-07T02:41:53 In a 3-player game, you end up in a situation where if you fight you insta-lose. 2011-02-07T02:41:58 http://en.wikipedia.org/wiki/Swarm_intelligence i expect people to play around with these ideas 2011-02-07T02:42:24 i wasn't considering prime numbers for number of players 2011-02-07T02:42:36 Do you play much Galcon? Think of how a computer would play a situation with three players left, where no one player has more than 50% of the production. The optimal strategy is to sit and wait for the other two to fight. 2011-02-07T02:42:36 as you basically need to do a 1xn region splitting on the torus 2011-02-07T02:42:53 i don't play galcon much but i know what you mean 2011-02-07T02:43:01 i did bang my head with pw for 3 months :P 2011-02-07T02:43:29 hopefully people will try to coordinate cooperation 2011-02-07T02:43:49 and then we can have people band together before breaking the cooperation 2011-02-07T02:44:04 possibleNoPlayers = [4, 6, 8, 9, 10, 12, 14, 15, 16] 2011-02-07T02:44:08 that's what my current map gen uses 2011-02-07T02:44:18 Whereas if you put a fog of war, then it's never a good idea to camp unless you can somehow see all the places where your enemies border eah other. 2011-02-07T02:44:36 you still don't want to camp until you've secured your region 2011-02-07T02:44:42 and i doubt people will do that perfectly symmetrically 2011-02-07T02:44:56 i dont think we should use symmetric maps. 2011-02-07T02:45:08 i am very agains the game not being symmetrical 2011-02-07T02:45:29 there is such a small variety of maps that are symmetrical with large number of players. 2011-02-07T02:45:38 no there isn't 2011-02-07T02:46:11 i mean, you'll get different-looking maps. but from a strategic point of view they land in a small number of categories right? 2011-02-07T02:47:18 You basically have to put the starting positions on some sort of square lattice or radially symmetric pattern. 2011-02-07T02:47:21 That's it. 2011-02-07T02:48:26 i'll outline how i'm currently thinking of making them 2011-02-07T02:48:29 Whereas we could continuously generate non-symmetric maps and put them into the rotation. We can have a script that regularly assesses the maps to measure their fairness. 2011-02-07T02:49:10 We will end up with an unimaginable variety of very interesting maps that have really interesting strategic elements. 2011-02-07T02:50:25 We could even have the system take into account the average ranking score of the players when measuring fairness of maps. So we would end up with a set of maps for each different skill level. 2011-02-07T02:50:44 *** boegel has joined #aichallenge 2011-02-07T02:50:49 The magical part would be that the set of maps would change and respond as people submit different sorts of bots over time. 2011-02-07T02:51:14 * boegel likes ;agic 2011-02-07T02:51:17 urgh 2011-02-07T02:51:17 If somebody comes up with a completely different strategy, and their strategy spreads, the maps will evolve to accommodate that. 2011-02-07T02:51:20 * boegel likes magic 2011-02-07T02:51:26 I'd like to see non-symmetric maps with tracking of the maps to give bonuses for positions that have weaknesses and thrown out if they get too unbalanced. I also think this would be really hard to get right though. 2011-02-07T02:52:11 I don't see it as being too hard. To measure a map's fairness, just do a t-test on the hypothsis "all starting positions are equally likely to win" 2011-02-07T02:52:57 When an unfair map starts to get some play, it will gradually produce evidence that it's not perfectly fair. As soon as there's enough evidence to conclude that the map is unfair, it gets thrown out. 2011-02-07T02:53:39 http://codepad.org/7b2gzhnb 2011-02-07T02:53:47 that's a rough explanation of how i would make symmetric maps 2011-02-07T02:54:04 Or perhaps we set the map's priority to the score from the t-test. That way unfair maps will naturally drop out of favor, fair maps will stick around for a long time, and new maps will get lots of play at first until their t-test score settles down. 2011-02-07T02:55:58 Seems solid antimatroid. 2011-02-07T02:56:16 I just don't think the maps are going to be interesting enough though. 2011-02-07T02:56:25 it isn't the same path to all of your neighbours, but once you goto n>2, you can't have the same path to ALL players anyway 2011-02-07T02:56:25 They're all going to be square lattices. 2011-02-07T02:57:04 if it's not quite clear whether you have an advantage or not on a map, it would also make it harder to tell how well your bot did on the map 2011-02-07T02:58:53 I believe that if we re-assign map priorities once per hour using a statistical analysis of the maps game histories, the majority of games will take place on maps that are well within people's fairness tolerance. 2011-02-07T03:00:52 Plus it would just be so sexy. 2011-02-07T03:01:33 i personally don't like non-symmetric at all 2011-02-07T03:01:45 i did play galcon before and couldn't stop thinking, this would be a lot better if maps were symmetric 2011-02-07T03:02:05 I think I get the jist of your concern. But don't you think that having all the maps be square lattices is a little boring? 2011-02-07T03:03:05 I think your concern is that everyone should have the same shot at winning. Also, that the maps should provide clear feedback to their creators by maximizing the role of skill and minimizing the role of luck. 2011-02-07T03:03:37 I am asking you to also consider the interestingness of the maps. 2011-02-07T03:04:18 I claim that we can maintain all of the fairness of symmetric maps without creating boring maps. 2011-02-07T03:04:46 If we just use some stats. 2011-02-07T03:05:15 you'd really want something that makes near symmetric maps then? 2011-02-07T03:05:22 not near symmertic 2011-02-07T03:05:26 near fair 2011-02-07T03:05:32 how would one even go about that? 2011-02-07T03:05:43 you don't want to be throwing out 99+% of maps 2011-02-07T03:05:59 and i am a big fan of the final maps being new 2011-02-07T03:07:06 however, depending on how well the stats idea works, i could be alright with that 2011-02-07T03:07:16 without that i would have hated it 2011-02-07T03:07:54 There's nothing wrong with throwing out 99% of the maps. 2011-02-07T03:08:00 I think that's probably what would happen. 2011-02-07T03:08:25 The system might end up discovering one good map per day. That would be okay with me. 2011-02-07T03:08:34 People will notice over time that the maps are improving. 2011-02-07T03:08:47 in order to throw out 99% of maps, you're going to be playing an awful lot of games on unfair maps so you can discover they're unfair 2011-02-07T03:09:18 Yes. We can control that by putting an upper cap on the percentage of games that can be played on new maps. 2011-02-07T03:09:54 how many games do you think will be needed to determine the fairness? 2011-02-07T03:10:25 Quite a few probably. The maps with more players will require more trials. 2011-02-07T03:10:49 However, the grossly unfair ones will disapper quicker. 2011-02-07T03:10:57 what could be done is an attempt to rank people not just by scoring but the level of "fairness" of their position as before, but like janzert said, that soon becomes really complicated 2011-02-07T03:11:10 Say you've got a map where one starting location has 4 wins and the other have 6. That one would disappear really fast. 2011-02-07T03:11:32 I think we should just aim to maximize the fairness of maps. 2011-02-07T03:11:51 hmmm, what about... start with your t-test with a confidence bound, then if games > x and you reject the fairness, throw the map 2011-02-07T03:12:14 i still think it's worth making a map gen that at least tries to make a fair map 2011-02-07T03:12:14 Really gross unfairness will disappear really quickly. Proving that a map is really well balanced will take a lot of trials though. 2011-02-07T03:12:22 like maybe place everyone on a square lattice 2011-02-07T03:13:44 I think it's a good idea. We can make a map gen that tends to produce fair-ish maps. We can also invent a way to do a basic pre-screen to detect maps that are grossly unfair. For example, create a voronoi diagram and count how much "area" each player starts with. 2011-02-07T03:15:33 j3camero: it can even be a rectangle lattice 2011-02-07T03:15:40 heh (not a serious proposal but) separate from running games you could let bots bid for positions on maps and then use that as a points handicap in the game, but you need to give bots permanent memories so they can track how they did in the various maps/positions 2011-02-07T03:15:40 verticle and horizontal heights don't need to be equal 2011-02-07T03:16:15 janzert: that would be really interesting. 2011-02-07T03:17:26 yeah, I think it would add too much complication though 2011-02-07T03:17:50 i think it's going to be expensive to work out fairness of a map 2011-02-07T03:18:09 particularly if you have 16 player maps 2011-02-07T03:18:21 antimatroid: i hate the lattice system. I think people will be really bored with the maps. It really restricts things since player positions will be so predictable. You always have a neighbor to the north, east, west, and south, at the exact same distance. 2011-02-07T03:18:55 j3camero: you might have the same neighbour up and down, or it might just be you there 2011-02-07T03:19:14 you also have to pick which direction to go first to start expanding 2011-02-07T03:19:24 and then respond according to what everyone else did etc. 2011-02-07T03:19:34 if you have fog of war, you wouldn't even know what others are doing 2011-02-07T03:19:54 I think that those problems become more interesting when you start off not knowing the structure of the map. 2011-02-07T03:20:08 Imagine how not-fun Starcraft would have been if all the maps were regular lattices. 2011-02-07T03:20:25 i don't really play games, i've not played starcraft 2011-02-07T03:20:38 It would have completed ruined the game. 2011-02-07T03:21:17 when i do play games, i do prefer them to be symmetric, i don't see how the maps are less interesting, with a fog of war, you don't even know the map :P 2011-02-07T03:21:43 players don't know the region layouts or the srow,scol dimensions, so they also need to pattern match to try and work that out 2011-02-07T03:21:54 they could also use that to deduce the number of players 2011-02-07T03:23:10 I just think it's so boring. 2011-02-07T03:23:15 No fun to watch. 2011-02-07T03:23:21 Well, less fun. 2011-02-07T03:23:53 i find it more fun to see how people go knowing it was completely fair, and i don't really see how you're going to get enough games to throw out lots of maps 2011-02-07T03:23:58 I have high hopes that we will be able to find a great set of fair maps that are really interesting. So we get all the win associated with interesting maps without losing the fairness. 2011-02-07T03:24:43 i'll make one, how does this sound... 2011-02-07T03:25:20 normal placement of players in a rectangle lattice, then for each food/wall square, i'll randomly place one in each players "region" for every one added 2011-02-07T03:26:12 then if food is placed, one is placed in every players "starting" region 2011-02-07T03:26:19 should food placement be static for a given map somehow? 2011-02-07T03:26:33 Say we can play 10 games per minute. That's 14,000 games per day. Say we use 20% of these games to trial the newest maps. That's almost 3000 games per day just for testing new maps. I think that's easily enough. 2011-02-07T03:26:40 it would be hard to debug if playing a game on the same map with a deterministic bot gave differing results 2011-02-07T03:27:15 but it also needs to be unpredictable by the bot, hmmm 2011-02-07T03:28:36 What I like is that your bot would not know what is around them at all when they start. 2011-02-07T03:29:15 i like that too, particularly not knowing the number of players, it's like colony wars in the real world, fighting to expand your colony without knowing how many other colonies exist 2011-02-07T03:29:41 Right. It presents risks that are really difficult to measure and control. 2011-02-07T03:30:08 You have no idea where the other players are, how far away they are, or what's right around the next corner. 2011-02-07T03:30:19 *** Palmik has joined #aichallenge 2011-02-07T03:30:23 Acquiring knowledge of the world will be a whole new end in and of itself. 2011-02-07T03:31:01 i like visibility through walls, because then you might see something attractive but there may not actually be an easy path to it 2011-02-07T03:31:13 it can essentially fool you if you assume unseen squares aren't walls 2011-02-07T03:32:12 I'd say let food be completely random and have starter packs record a rng seed with the game that it will also accept to replay the game the same way 2011-02-07T03:32:39 Good idea. 2011-02-07T03:32:57 rng seed? 2011-02-07T03:33:08 What would be really sexy is we could go whole-hog with the map auto-balancing system and just use a genetic algorithm. 2011-02-07T03:33:19 rng = random number generator 2011-02-07T03:34:03 So the map gen takes a bunch of parameters and generates a map. A small change in the parameters should produce a small change in the map. Then we do a big genetic algorithm to evolve good maps. 2011-02-07T03:34:05 aren't we meant to be keeping this simple for beginners? 2011-02-07T03:34:30 None of this complicates the game for beginners. 2011-02-07T03:35:00 j3camero: what constitutes a good map? 2011-02-07T03:35:18 Fairness. 2011-02-07T03:35:21 which language and libraries are you using for the genetic algorithms? 2011-02-07T03:35:36 are they guaranteed to be symmetric or no? 2011-02-07T03:35:41 they = the maps 2011-02-07T03:35:56 Where fairness is the probability that his statement is false: "all starting locations are equally likely to win". We measure this probability with a simple t-test on the game data. 2011-02-07T03:36:25 I don't think symmetric maps are a good idea. 2011-02-07T03:37:33 I am not sure if I would use a library for this GA. 2011-02-07T03:37:52 kk. Which language do you like to use? 2011-02-07T03:38:05 j3camero: http://codepad.org/9svfZ6s9 look at the map at the bottom 2011-02-07T03:38:05 Well different parts would be in different languages. 2011-02-07T03:38:13 that is symmetric, players aren't going to know what to do at the start 2011-02-07T03:38:46 player 1 is on row 16 2011-02-07T03:39:04 Well you still know that you've got neighbors up down left and right, at equal distances. 2011-02-07T03:39:14 You know you're not in a corner. 2011-02-07T03:39:19 your up neighbour is your down neighbour 2011-02-07T03:39:29 and the distance is different to your left and right neighbour distance 2011-02-07T03:39:35 j3camero: hmm interesting. Which languages and why mix? Just to use the strengths of each one? 2011-02-07T03:39:45 and there isn't a clear path straight to them, you don't know a path there and they wont be there when you get there 2011-02-07T03:40:24 it's entirely possible you can't get from a to b without going left 2011-02-07T03:40:27 and off the screen 2011-02-07T03:42:34 You're right that the maps can still be complex and have interesting features. 2011-02-07T03:42:39 i think walls should block fights, but that shouldn't be hard, just block them when scanning outwards from players to make the "fighting" graph 2011-02-07T03:43:04 j3camero: I'm just starting to learn about GAs so that's why I'm asking. I know/use Python, Clojure, and Go and wonder whether the language features you think are useful for GAs can be found in these languages, and which ones 2011-02-07T03:43:22 elmisteve: it was a passing comment about using GA's 2011-02-07T03:43:28 I think these days all languages have really great GA libraries. 2011-02-07T03:43:35 from what i understand, they're only best when used as a last resort 2011-02-07T03:43:36 Python has pygene though I have never used it before. 2011-02-07T03:43:40