2011-12-24T00:13:16 *** avdg has quit IRC (Quit: Leaving.) 2011-12-24T00:19:14 http://tcpants.com/ranking lets play 2011-12-24T00:23:18 *** roflmao has joined #aichallenge 2011-12-24T00:36:26 Accoun: Do you support playing over network socket and ssh ? 2011-12-24T00:37:26 *** thestinger has joined #aichallenge 2011-12-24T00:37:31 hi. 2011-12-24T00:39:27 *** rebelxt has joined #aichallenge 2011-12-24T00:43:47 *** TheLinker has quit IRC (Quit: Bye) 2011-12-24T00:44:01 Ahhh the unencrypted TCP client that you provide is probably OK, I dunno I just have this mental block against unencrypted TCP seccions 2011-12-24T00:44:32 Oh no the ant traffic might be snooped upon? 2011-12-24T00:45:47 Yeah it's irrational but years of IT experience has scarred me with massive paranoia 2011-12-24T00:46:11 Don't feel bad, lots of people comment on it :-) 2011-12-24T00:46:36 Accoun: how do I register for a username and password? 2011-12-24T00:47:33 for tcp? 2011-12-24T00:47:38 Yeah 2011-12-24T00:48:03 python2 tcpclient.py tcpants.com 2081 ./MyBot.jar red YOUR_PASSWORD 500 2011-12-24T00:48:04 Or is it finished now? 2011-12-24T00:48:07 just like that 2011-12-24T00:48:12 well, people are still playing on tcpants.com 2011-12-24T00:48:20 kinda pointless imo :) 2011-12-24T00:48:26 unless the game rules are changed up a bit 2011-12-24T00:48:47 Well I guess the prize is about the same as the original comp? 2011-12-24T00:50:27 Thestinger congrats on beating me 2011-12-24T00:50:35 Antimatroid_: :D 2011-12-24T00:50:51 I was gone for a while so opening up the page was scary 2011-12-24T00:51:08 You missed green tea taking the lead 2011-12-24T00:51:14 That was a bit crazy 2011-12-24T00:51:23 very close up there 2011-12-24T00:51:41 It swapped for a little bit 2011-12-24T00:52:34 *** cw_ has joined #aichallenge 2011-12-24T00:52:39 heh, and runevision gets 4th 2011-12-24T00:52:43 he came out of nowhere 2011-12-24T00:52:45 *** cw_ has quit IRC (Client Quit) 2011-12-24T00:53:04 *** FranPeruzzi has joined #aichallenge 2011-12-24T00:53:58 so, is it decided what next year's contest will be yet? 2011-12-24T00:55:49 there are a bunch of great ideas on the forums 2011-12-24T00:57:30 I thought tron was awesome, and right now I'm thinking ants will be a hard act to follow. I hope the contest can keep the same inclusiveness it has always had, even if there has to be some kind of immediate starterbot cutoff like what has been suggested in the forums. 2011-12-24T00:57:55 http://forums.aichallenge.org/viewtopic.php?f=21&t=2184 this is kinda like tron 2011-12-24T00:58:54 I thought so too. 2011-12-24T01:00:05 I've been wracking my brain for things that are complex enough to give challenge at the top of the lists, while being easy enough to maintain a low barrier to entry at the bottom. 2011-12-24T01:00:54 (where I'll probably be) 2011-12-24T01:02:54 something with simultaneous turns would be nice. It keeps the possible random element of a mixed strategy in place. 2011-12-24T01:04:06 I like the idea of a strategy game (real-time or turn based, gridded or continuous) with unit options randomly generated at the start of each game 2011-12-24T01:04:25 Re: "next year's comp", I'd like something with economics in it... in honour of the banking undustry 2011-12-24T01:05:37 Kettling: hrmm, so like monopoly meets wall street, your bot has to decide where to build houses, when it's worth it to forclose instead of restructure, etc? =-P 2011-12-24T01:05:43 Errr I meant industry, although the typo might have been divinely influenced 2011-12-24T01:06:01 I can't even understand lemonade stand vs weather economics 2011-12-24T01:06:37 *** QGazQ has quit IRC (Ping timeout: 252 seconds) 2011-12-24T01:06:58 Extrarius: How to place key people in the treasury department mostly... but anyhow we can polish down the details 2011-12-24T01:07:20 some kind of economy might be interesting. 2011-12-24T01:08:07 makes me wonder how the game engine would work though, and if open sourcing it would make the game moot. 2011-12-24T01:08:39 I'd almost like to see a slight modification of ants for the next competition - call it 'bees' and instead of hills, you have queen bees you move around that workers spawn at, and you have to fly food to the queen to make a new worker 2011-12-24T01:09:18 I loved ants, but I think a modification of ants would allow people to take too much advantage of the knowledge that has been put out there by this year's leaders. 2011-12-24T01:09:35 *** rebelxt has quit IRC (Quit: Page closed) 2011-12-24T01:11:20 I think a risk/diplomacy style game would be awesome 2011-12-24T01:11:38 FranPeruzzi: It would raise the bar for entry, but personally I'd love to see another iteration of ants with people taking advantage of what everybody published. Changing a few rules slightly would make room for more innovation (ie movable hills, transporting food, maybe a few other things), and it'd be neat to see what comes next 2011-12-24T01:11:44 while reading post-mortems, one of the things that stood out was the fact that my understanding of the problem and associated strategies just was not as clear as that of the leaders. I think that having something similar to ant would take the advantage away from those who are able to think more strategically 2011-12-24T01:12:00 I would love something like Risk 2011-12-24T01:12:19 also, that kind of competition could be run very soon since most of the work is done already =-) who wants to wait months for another challenge to be developed? =-) 2011-12-24T01:12:29 that's true 2011-12-24T01:12:49 although I think the organizers probably need about a month's worth of sleep right now 2011-12-24T01:13:04 Yeah risk/diplomacy lends itself to economics as well... each territory provides certain resources and you need to agree to trade with neighbouring territories for good results (of course owning BOTH territories is also good) 2011-12-24T01:13:31 e.g. one territory has iron, the other has coal, the next has oil 2011-12-24T01:13:36 Another iteration of ants opens itself to more of the same complaints that some people had about beta vs. non-beta participants 2011-12-24T01:14:03 Not that I've been reading Adam Tooze or anything :-) 2011-12-24T01:14:18 KIettling: are you playing a different Risk than me? I haven't played for ~20y, so it's entirely possible. 2011-12-24T01:15:12 *** mleise has joined #aichallenge 2011-12-24T01:15:17 FranPeruzzi: standard risk just gives a fixed resource per territory and all resources are interchangable 2011-12-24T01:15:50 hrmm, how about settlers of catan meets risk? 2011-12-24T01:15:51 Kettling: but... you get at least 3 units and a holding whole continents gives boni 2011-12-24T01:16:20 FranPeruzzi: but if all resources are interchangable then economics becomes pointless 2011-12-24T01:16:47 I've seen Settlers in the bookstore, but don't know enough about it 2011-12-24T01:16:48 also you collect cards when you conquered/freed an enemy country that round. These cards can be changes depending on their combination into 4, 6, 8 or 10 units 2011-12-24T01:17:02 actually, I'd like to see somebody work out how to do AI for a game like 'magic the gathering' 2011-12-24T01:17:22 The risk game genre lends itself to many extensions... it's well trodden path. 2011-12-24T01:17:25 You can decide to wait until you have a max of 5 cards to have better options when changing cards (always have to change three cards for units) 2011-12-24T01:17:44 I loved MtG when I was in high school (15-20y ago, I probably have cards that are worth $$ now) 2011-12-24T01:18:02 cards have countries on them. if you own the country on the card at time of change, you get 2 extra units to be placed on that country 2011-12-24T01:18:27 mleise: I'm remembering more now. Mostly I remember my dad running around with a makeshift crown after kicking ass against a bunch of kids. 2011-12-24T01:18:54 Extrarius: Demonic hordes anyone? 2011-12-24T01:19:39 :) anyway all these board games tend to be mean for all but one player or the two that allied. risk, monopoly, settlers, ... 2011-12-24T01:20:09 I put up a suggestion of stratego on the forum, although I don't know how it works with computational complexity -- I imagine we need a certain level of it in order to make things worthwhile 2011-12-24T01:20:14 That's when I turned to computer games that offer more depth in the game logic 2011-12-24T01:20:38 Demonic hordes? 2011-12-24T01:20:51 old MtG card. Destroy land on tap 2011-12-24T01:21:34 haven't played Magic The Gathering in more than 15y 2011-12-24T01:21:40 never played these, i don't know if they were popular in germany. i wasn't into fantasy games# 2011-12-24T01:22:51 http://en.wikipedia.org/wiki/Labyrinth_(board_game) was as far as I went 2011-12-24T01:22:52 alliances could be an interesting aspect. I saw something either on here or the forums about sending strings back and forth. Although that may be unfair for lone wolves, like I tend to be. 2011-12-24T01:23:11 Prolly want avoid any obvious trademark infringement 2011-12-24T01:23:19 actually we should go back to 2 player games :) 2011-12-24T01:23:20 not that I have no friends, but they aren't programmers 2011-12-24T01:23:39 Kettling: I imagine that would be a definite consideration 2011-12-24T01:24:04 You mean real alliances like on some browser games? 2011-12-24T01:24:20 not just ad-hoc alliances during the matches? 2011-12-24T01:24:22 If the geniuses around here can figure out full-history raking, multi is cool with me. Even with TrueSkill multi was fun,. 2011-12-24T01:24:54 I guess alliances would be part of the AI. Determining if it is beneficial. 2011-12-24T01:24:55 FranPeruzzi: Well, I prefered the even matches in PlanetWars 2011-12-24T01:25:29 I missed PlanetWars, with a small kid, but I liked the even matchups in Tron. 2011-12-24T01:26:01 The games become so unstable. A weak player can drive a strong player into ruin, if a third player also attacks the strong player 2011-12-24T01:26:04 I think I did take advantage of the luck inherent in the multi-player aspect of ants though. 259 was probably a bit high for me -- thanks RNG 2011-12-24T01:26:30 I wasn't a small kid, I had one 2011-12-24T01:26:52 *** Westicles has quit IRC (Quit: Page closed) 2011-12-24T01:26:55 I agree that there is less uncertainty in 1v1 games than we had in ants 2011-12-24T01:27:13 mleise: I lean towards ad-hoc alliances during matches, because after all, it's an AI competition so the AI needs to make the decision on the ground 2011-12-24T01:27:39 Kettling: but they are mean in risk already :( 2011-12-24T01:27:49 Getting an AI to make sensible decisions about coordination with another entity is pretty darn difficult 2011-12-24T01:27:58 Whenever I play with my mother and sister, they ally against me 2011-12-24T01:28:24 people who base alliances only on a person being their friend are putting themselves at risk 2011-12-24T01:28:53 I wonder how it would affect the current rules about changing your game based on who you play 2011-12-24T01:28:53 mleise: Yeah well make sure the AI has no knowledge about WHO it is up against... needs to work from first principles (like that Turing test thingy) 2011-12-24T01:29:18 Kettling: I like that. Then your alliance decisions are based solely on utility. 2011-12-24T01:29:31 but a group of people may always code their bot in a way that it emits signals for their group to ally 2011-12-24T01:29:57 FranPeruzzi:exactly. No way should it be allowed to know the name of who else in on the field (nor any equivalent token) 2011-12-24T01:30:26 mleise: so you don't let people send arbitrary strings, that wasn't really a good idea 2011-12-24T01:30:32 mleise: the game would have to include a defined list of messages that could be sent between bots 2011-12-24T01:30:40 you just have a way to propose an alliance, and then the player has to accept/reject 2011-12-24T01:30:42 mleise: That would be half the fun, if those secret handshakes are exposed to interpretation by other players. 2011-12-24T01:30:42 hmm 2011-12-24T01:30:53 and it would all be visible to every player, so you could figure out which players could be trusted 2011-12-24T01:31:07 thestinger: interesting 2011-12-24T01:31:10 ofc you have no idea at the start of a game 2011-12-24T01:31:16 communication in the open 2011-12-24T01:31:28 well, the proposals don't have to be in the open 2011-12-24T01:31:34 I think that would bias for people with larger organzations though 2011-12-24T01:31:42 but the existing alliances and breaking alliances should be in the open 2011-12-24T01:31:54 for instance, I work for Xerox. If I get everyone to play, we could organize a protocool 2011-12-24T01:31:56 FranPeruzzi: well, the same "no changing behavior based on the identity of other bots" thing would apply 2011-12-24T01:32:11 not much you can do if you're limited to 1 alliance proposal a turn 2011-12-24T01:32:15 Yeah, I like comms in the open, and also straightjacked comms (like poker or bridge) 2011-12-24T01:32:21 thestinger: right 2011-12-24T01:32:29 difficult to enforce though? 2011-12-24T01:33:09 well, they can see all our source code 2011-12-24T01:33:23 *** Zaphus has joined #aichallenge 2011-12-24T01:33:26 I think multi-player with alliances would be awesome if it can be solved for fairness. 2011-12-24T01:33:28 I don't think people would try to break the rule if they planned on actually winning - they'd be caught 2011-12-24T01:33:32 Actually, some sort of bidding system might be a good combo along with diplomacy 2011-12-24T01:33:41 thestinger: true. source is visible to all organizers 2011-12-24T01:33:44 I'm waving hands here... 2011-12-24T01:34:17 I don't know diplomacy. I opened the wiki link a while ago, but the conversation is moving fast... 2011-12-24T01:34:58 sorry, I was thinking labyrnth 2011-12-24T01:35:41 thestinger: noone looks through 5000 bot source codes, there must be something really suspicious happening 2011-12-24T01:35:49 mleise: yeah, ofc 2011-12-24T01:36:09 but I don't think people would actually try to cheat by building a communication protocol over whatever logic the game gives you 2011-12-24T01:36:28 but if there's something suspiciuous at the top it would be reported 2011-12-24T01:36:38 I'm getting bashed on tcpants as well, oh well *sigh* at least it isn't using much CPU :-) 2011-12-24T01:36:45 is the top the only interesting spot? 2011-12-24T01:36:52 I was getting my ass kicked yesterday 2011-12-24T01:37:17 *** analyst74 has quit IRC (Ping timeout: 240 seconds) 2011-12-24T01:37:22 For a lot of participants the middle ranks are important and they may lose the fun if they have to fear that others are cheating 2011-12-24T01:37:26 I definitely don't think the top is the only interesting spot, I would love to see this competition keep its inclusiveness 2011-12-24T01:37:51 It doesn't have to be true, but it is like with online polls. People may change their IP and vote twice 2011-12-24T01:38:31 mleise: there are so many people playing, that imo it wouldn't affect the middle ranks 2011-12-24T01:38:32 And the respect for the whole thing is gone when the results are not reliable 2011-12-24T01:38:58 you would need like 100 people in a cheater cabal, and it would definitely get leaked out 2011-12-24T01:39:34 Also random shuffle in the games would reduce the value of secret handshakes 2011-12-24T01:39:35 Ok, you can propose alliances for the next game of course, you don't convince me though 2011-12-24T01:39:36 I think if there is a protocol for alliances, it would be in everybody's interest to code something that was utility based instead of user based. Even a 1-point difference in skill was enough in ants to make it so you never see different people in the finals 2011-12-24T01:39:50 mleise: well, it's not necessarily a good idea :P 2011-12-24T01:39:53 just brainstorming 2011-12-24T01:40:06 I agree that n-player games add too much noise 2011-12-24T01:40:20 my bot was jumping around from rank 25 to rank 50 today 2011-12-24T01:40:28 I was lucky and it closed with my rank @ 30 2011-12-24T01:40:35 Noise is good... I like to win a few you know! 2011-12-24T01:40:37 nice 2011-12-24T01:40:42 so unless you're friends with everyone it would be pointless. And if you're friends with everyone what's the point anyways if you're not being fair 2011-12-24T01:41:09 I was anywhere in about a 50-spot range today too. It was nerve wracking 2011-12-24T01:41:16 FranPeruzzi: Another diplomacy game might start *OUTSIDE* the official competition 2011-12-24T01:42:02 Kettling: true, need to keep it as fair as possible 2011-12-24T01:42:19 FranPeruzzi: All's fair in love and war 2011-12-24T01:42:51 this is a *sport* 2011-12-24T01:43:10 not a bycicle sport though 2011-12-24T01:43:38 yeah, but I think it needs to stay a programming competition 2011-12-24T01:44:22 something like stratego or risk is looking really good right now, but diplomacy is looking grim without a strict framework 2011-12-24T01:45:16 Let's make a cool space game with space goblins conquering planets with mother ships and landing troops. And they would build factories and the A.I. decides what modules to combine to a ship to alter its level of speed, defenses and load capacity. 2011-12-24T01:46:28 hmm, sounds like a Battletech variant after they added more large-scale strategy 2011-12-24T01:46:52 *** grc48 has joined #aichallenge 2011-12-24T01:47:04 something like old-school Traveller would be cool minus role-playing aspects 2011-12-24T01:47:31 hey grc48, talking about ideas for next game 2011-12-24T01:47:58 http://tcpants.com/ranking lets play 2011-12-24T01:48:12 Accoun - putting up some fodder for you 2011-12-24T01:49:03 you were killing me yesterday 2011-12-24T01:49:37 Opening family game cupboard... 2011-12-24T01:49:38 Egats, Traveller, I remember when that was a new idea 2011-12-24T01:50:11 hmm, monopoly, parchisi, othello\ 2011-12-24T01:51:02 never got a chance to actually play Traveller myself. My dad had the boxed set (sorry if that makes you feel old). My friends and I played AD&D2, Rifts, Star Wars (d6) 2011-12-24T01:52:09 Kettling: looks like I beat you on tcpants 2011-12-24T01:52:20 *** Zaphus has quit IRC (Quit: Page closed) 2011-12-24T01:52:27 I think that's the first I won 2011-12-24T01:53:42 even bh was nasty on TCP. Even though it was supposed to be a 'lower-level' tcp site 2011-12-24T01:53:55 Battletech is like a game I invented after playing some simpler form on a sheet of grid paper with friends at school 2011-12-24T01:54:11 I used to design mechs for hours 2011-12-24T01:54:34 but really... all this management of damage and fuel is why I was lucky about the invention of computers :) 2011-12-24T01:54:47 really 2011-12-24T01:55:04 *happy i meant 2011-12-24T01:55:20 Does anyone know Battle Isle? 2011-12-24T01:55:20 I remember though my dad took me to a gaming convention when I was 10, and I had a Wasp. I got smoked and cried. 2011-12-24T01:55:45 haha 2011-12-24T01:55:50 lol 2011-12-24T01:56:37 I called my game, ported from paper, Island battle and then found out about Battle Isle 2011-12-24T01:56:58 It was pathethic. As a noob, I should have had the Warhammer or Battlemaster, but nooooo... 2011-12-24T01:57:05 the hex field map there was a good idea 2011-12-24T01:57:52 it was, but I think it would be difficult to manage in a competition like this 2011-12-24T01:58:12 what exactly? the hex grid?? 2011-12-24T01:58:21 yes, hex grid 2011-12-24T01:58:45 I see no problem with that. It was proposed here earlier and noone screamed 2011-12-24T01:58:47 really only a half phase shift, without much benefit? 2011-12-24T01:58:47 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has quit IRC (Quit: Lost terminal) 2011-12-24T01:59:01 If it's in starter, it could work 2011-12-24T01:59:09 well, you have smoother paths 2011-12-24T01:59:10 _flag: nice writeup 2011-12-24T01:59:34 that's true 2011-12-24T02:00:03 and the distance is still 1 to every neighbor. this is the downside with diagonal moves on a square grid 2011-12-24T02:00:19 I like the paths available in hex maps. I remember playing Desert Commander on NES, and it was squares -- I wished it was hex 2011-12-24T02:00:31 *** dabunny has joined #aichallenge 2011-12-24T02:00:40 If you don't know how to implement though... 2011-12-24T02:01:20 this could be covered by a tutorial 2011-12-24T02:01:23 like xathis tile implementation was cool, and I should have had something like it but didn't (like all the other things that cause win vs lose) 2011-12-24T02:01:40 what did he do? 2011-12-24T02:01:53 I wasn't attending the competition for the last weeks 2011-12-24T02:01:59 nothing special (as far as tiles) just had a tile class 2011-12-24T02:02:18 but you would need something like that in a hex-based game 2011-12-24T02:02:29 and I didn't even think of it 2011-12-24T02:02:43 but what does the tile class do ? :p 2011-12-24T02:03:08 just having a coordinate can't be the magic 2011-12-24T02:04:08 I think it just has the location and references to neighbors 2011-12-24T02:04:26 nah comeon, you didn't think of that? 2011-12-24T02:04:44 FranPeruzzi: Everyone seems to be beating me on tcpants *SOB* I'm still puzzling over the combat engine... but it's also Chrstmas eve so I've put away a bottle of Australian prosecco as err, "thinking juice" to help me along 2011-12-24T02:05:46 antimatroid for example implemented A* this way: he used the tiles as nodes and linked them to their neighbors 2011-12-24T02:05:46 I love the thinking juice. I've put away some Keystone, which is much worse than yours. My combat is nothing to brag about -- I was amazed at the simplicity of the top bots combat. 2011-12-24T02:06:16 so there are many opportunities to discover the 'tile class' 2011-12-24T02:06:35 Right mliese, but in a hex map I think you would have to include an effective implementation in the starter kit to keep the same low barrier to entry 2011-12-24T02:07:06 *** Paradoxial has quit IRC (Quit: :) 2011-12-24T02:07:38 Kettling: my problem with combat was that I couldn't pick out the best moves. Since I'm in Python I tried to prune waaay too much. 2011-12-24T02:08:14 http://tcpants.com/ranking lets play 2011-12-24T02:08:27 I think next year if I program in Python I'll give myself like 5000ms per turn, then once I have a strategy I'm happy with optimize down to get it within tolerance. 2011-12-24T02:08:59 it is not like there wasn't enough round time in ants, a game of PlanetWars took 200 seconds max, ants went far beyond that 2011-12-24T02:09:06 Accoun you and red are already whooping me 2011-12-24T02:09:21 but you had to spread your calculations somewhat over the course of several turns 2011-12-24T02:09:32 yeah, but I think one of my problems is that I started optimizing too early 2011-12-24T02:09:49 FranPeruzzi: that happens to me too 2011-12-24T02:09:54 If I master the algo first, then the speed, things will be better off 2011-12-24T02:09:54 just use C++ and you never have to optimize :P 2011-12-24T02:10:07 I'm bad at just writing an ugly prototype and clean up the mess later 2011-12-24T02:10:21 right, one of the things I want to do this year is make it through my copy of Strousrtroup 2011-12-24T02:10:39 thestinger: You have to know C++ as well to avoid performance killers. C is better ;) 2011-12-24T02:10:42 My code was so nasty compared to the code that was posted 2011-12-24T02:11:00 I have to get my copy of K&R back, or buy another copy 2011-12-24T02:11:09 mleise: well then you avoid the C++ features except for stuff you've actually learned 2011-12-24T02:11:18 well, few people *post* their code in the first place if it makes them look bad 2011-12-24T02:11:27 Mine is so sloppy 2011-12-24T02:12:18 But as far as 'looks bad' is concerned, I'm pretty happy with my final rank -- hope to do better next year though 2011-12-24T02:13:25 The complexity of the next game may increase even more 2011-12-24T02:14:05 *** Belerafon_L has joined #aichallenge 2011-12-24T02:14:10 Looking at "rock, paper, scissors", Tron, PlanetWars, Ants I see a pattern of increasing complexity :) 2011-12-24T02:14:11 mleise: that would be ok with me, as long as it doesn't totally disqualify slower langs 2011-12-24T02:14:51 FranPeruzzi: Antimatroid wants a very simple game again. One that can be solved. 2011-12-24T02:15:07 If it gets to the point where langs like Python, Ruby, etc, just aren't viable, It will be time for a new competition to form 2011-12-24T02:15:12 One where you can find the perfect algorithm 2011-12-24T02:15:50 perfect algo is bad 2011-12-24T02:15:52 FranPeruzzi: they were already barely viable :P 2011-12-24T02:16:10 FranPeruzzi: Well, the language is a tool. There are several tools. Chose the best for the job ;) 2011-12-24T02:16:20 FranPeruzzi: Fluxid is a professional python programmer 2011-12-24T02:16:23 https://github.com/fluxid/flantob 2011-12-24T02:16:24 C will be more work than Python, therefor it is faster 2011-12-24T02:16:43 You trade speed for ease of use 2011-12-24T02:16:57 As a Python coder, I'm starting to feel the same way. Where C programmers were considering all minimax possibilities, I was considering how to prune more 2011-12-24T02:17:03 That said, PyPy gives good results 2011-12-24T02:17:04 python: 933 (48.70%) 2011-12-24T02:17:07 ansic: 738 (38.52%) 2011-12-24T02:17:09 sh: 245 (12.79%) 2011-12-24T02:17:31 mleise: yeah, pypy gets you a 5x speed boost and lower memory usage for each object (ofc the JIT uses a ton of memory, but whatever :P) 2011-12-24T02:18:01 It is a trade off. If you can implement your algo in Python in like the first half, then switch to C for second half, I think that's a good strategy as far as meta is concerned. 2011-12-24T02:18:01 FranPeruzzi: pypy also has optimized ctypes 2011-12-24T02:18:13 and that's much nicer than using a C extension 2011-12-24T02:18:23 I submitted pypy once, without trying it locally 2011-12-24T02:19:20 It crashed because implementation of gc module is different (I guess) and I had specifically made gc happen at beginnnig of turn with gc module, then disabled. 2011-12-24T02:20:14 I'll probably look into it more in the future, although I'll probably concentrate more on doing something similar to Fluxid -- code in py and optimize in C 2011-12-24T02:20:24 well 2011-12-24T02:20:29 *** ruser has quit IRC (Ping timeout: 240 seconds) 2011-12-24T02:20:30 easier just to code the whole thing in C 2011-12-24T02:20:33 than using C extensions 2011-12-24T02:20:42 not sure though because I don't know the comparative PITA factor 2011-12-24T02:20:45 because the boilerplate and ugliness that ends up in the C makes it way worse 2011-12-24T02:20:55 may move to C too. 2011-12-24T02:21:24 probably code algos in py up to about 1/2 through contest, then move to C once I believe in my algos 2011-12-24T02:21:24 or just use modern C++, and don't worry about the 25-30% performance loss over C 2011-12-24T02:21:46 Did you look into D? It has some nice stuff like associative arrays and dynamic arrays, that C lacks 2011-12-24T02:21:50 that's a possibility too 2011-12-24T02:22:07 yeah, D is nice for something like this too 2011-12-24T02:22:37 Personally, I think I have a lot of room for improvement in my algorithms -- Graph theory, general math, etc 2011-12-24T02:23:11 Although I have to say, I'm profiling the code sometimes, because the library implementations are not all at the quality they should be in. BigInt for example is slow. 2011-12-24T02:24:06 I was profiling my python, and found that there were best practices that would improve my performance by 2x + without moving away from py. 2011-12-24T02:24:41 although I feel like I really need to know a compiled language to be serious 2011-12-24T02:24:46 things like that, yeah. where the compiler/interpreter can deduce some optimizations from the way you write code 2011-12-24T02:25:01 cpython barely optimizes :P 2011-12-24T02:25:19 the performance gains are mostly from using the standard library as much as possible and avoiding certain slow things (for loops) 2011-12-24T02:25:22 I needed to use cProfile and do my own optimization 2011-12-24T02:25:25 In D, foreach (i; 0..n) is usually faster than for (i = 0; i < n; i++) 2011-12-24T02:26:10 that reminds me of perl, are the range operations similar in D? 2011-12-24T02:26:45 I don't know Perl, but in D you have this slice operator on arrays (and user defined types) 2011-12-24T02:27:02 *** ruser has joined #aichallenge 2011-12-24T02:27:07 FranPeruzzi: another option is using a fast language, but not necessarily C, C++ or D 2011-12-24T02:27:20 haskell, clojure, scala, etc. will be way faster than python 2011-12-24T02:27:21 something like Go or Java? 2011-12-24T02:27:22 it works with foreach as well, and on arrays you can write [10..$], where dollar means 'end' of list 2011-12-24T02:27:30 but still 2x-3x slower than C/C++ 2011-12-24T02:27:38 FranPeruzzi: Java is way worse than C++11 imo 2011-12-24T02:27:39 I think I need to work on my algos first. 2011-12-24T02:27:50 For people who a bored, my analysis of game #346138 from GreenTea's perspective ==> http://bespoke1.lnx-bsp.net/ants/g346138_gt.mng 2011-12-24T02:28:00 I'd really like to either get into bare C or C++11 2011-12-24T02:28:22 my browser cannot render animated PNGs :( 2011-12-24T02:28:25 And xathis ==> http://bespoke1.lnx-bsp.net/ants/g346138_x.mng 2011-12-24T02:28:36 I think C is as close to the metal as I ever care to get. 2011-12-24T02:28:42 mleise: What??!? no mplayer 2011-12-24T02:29:04 mleise: how do you get through a normal day? 2011-12-24T02:29:07 I don't reboot to Linux now 2011-12-24T02:29:18 *** roflmao has quit IRC (Quit: Page closed) 2011-12-24T02:29:30 mleise: mplayer works fine in windows :P 2011-12-24T02:29:37 smplayer is pretty nice 2011-12-24T02:29:49 but my wind0ws is only for gaming 2011-12-24T02:30:11 ants == gaming :D 2011-12-24T02:30:12 sadly i had to install eclipse to fix some windows visualizer issues with unicode characters 2011-12-24T02:30:34 did you read FlagClapper's writeup? I really identified with that -- where he wanted to have control over everything he did. My bot was more 'emergent' 2011-12-24T02:30:46 MNG didn't quite get out of the box the way other video formats did... however "advmng" is very nice for toony animations 2011-12-24T02:30:47 no, where can i find it? 2011-12-24T02:31:02 http://flagcapper.com/ 2011-12-24T02:31:12 such an egomaniac :) 2011-12-24T02:31:29 I haven't been on linux in like 5y. I really want to get rid of Vista on my current desktip and put linux on 2011-12-24T02:31:41 *** gazpachoking has joined #aichallenge 2011-12-24T02:31:44 Flag? Kid's like 16, right? 2011-12-24T02:32:07 That kid is using to much JavaScript on its website 2011-12-24T02:32:13 FranPeruzzi: I still run Win7 in a VM (under linux of course)... kept away from Vista 2011-12-24T02:32:14 lol 2011-12-24T02:32:18 I .... can... bare..ly ... scro.. ll... 2011-12-24T02:32:28 vista is awful 2011-12-24T02:32:30 I don't notice any speed issues xD 2011-12-24T02:32:37 Use Opera 2011-12-24T02:32:56 you just convinced me not to use it 2011-12-24T02:33:46 maybe opera doesn't handle canvas well yet 2011-12-24T02:34:02 I didn't notice anything unusual on FF when reading his writeup 2011-12-24T02:34:08 Oh it has the most correct canvas implementation I've ever seen 2011-12-24T02:34:24 very accurate gaussian blur on shadow rendering as well 2011-12-24T02:34:44 Now it has settled (the animation stopped) 2011-12-24T02:35:37 I didn't have any problems with it 2011-12-24T02:37:02 also the layout is horrible, text rendering over other text, missing words, gaps 2011-12-24T02:37:20 hmm, maybe something like tower defense would be a good aichallenge. Or maybe something where you control both the towers and the attackers in a 1v1 env. 2011-12-24T02:37:47 I like the "multi agent shooter" idea 2011-12-24T02:38:31 did anyone try the Terrarium game? 2011-12-24T02:38:34 with cover? I hear cover selection is a big ai issue in game developement 2011-12-24T02:38:47 never tried...googling... 2011-12-24T02:39:26 A.I. is taking cover just fine in the shooters I care about 2011-12-24T02:39:31 is this it? http://www.terraria.org/ 2011-12-24T02:39:44 no 2011-12-24T02:39:50 this is v2 apparently: http://terrarium2.codeplex.com/ 2011-12-24T02:40:20 that's way different 2011-12-24T02:40:37 somewhat similar to ants 2011-12-24T02:40:42 but each ant was independent 2011-12-24T02:40:44 kind of 2011-12-24T02:41:09 each ant only had it's own field of vision, and your ants couldn't communicate directly 2011-12-24T02:41:12 I also liked with the ants that there were a large number of pieces on the board, and far to many moves to allow traditional deep tree analysis 2011-12-24T02:41:29 yes, that was a good twist 2011-12-24T02:41:34 that was really cool about ants 2011-12-24T02:41:39 *** arscan has joined #aichallenge 2011-12-24T02:41:44 the fighting was (for me) sufficiently complex, too 2011-12-24T02:42:04 I think it was sufficiently difficult for everyone 2011-12-24T02:42:31 Also I like the time limit on code but I would suggest just run a process with ulimit and let the player read their own ulimit to see what they have left to work with... it might legitimately make sense to put more effort into some moves than others 2011-12-24T02:43:12 ah! 2011-12-24T02:43:24 How about a zombies vs humans game 2011-12-24T02:43:39 where everyone controls a zombie or a human? 2011-12-24T02:43:47 All the ulimit stuff is totally well tested, well documented and debugged, man getrlimit 2011-12-24T02:43:51 *** Manwe56 has quit IRC (Quit: Page closed) 2011-12-24T02:45:02 great times guys -- thanks to the organizers! 2011-12-24T02:45:51 zombies sounds like something that would have appeal 2011-12-24T02:46:55 i haz idea for that 2011-12-24T02:47:03 you only get a point when you survive 2011-12-24T02:47:13 all players are humans 2011-12-24T02:47:21 yes 2011-12-24T02:47:21 to survive you must cooperate 2011-12-24T02:47:28 zombies are computer 2011-12-24T02:47:56 start in august, end 10/31 2011-12-24T02:48:02 ;) 2011-12-24T02:48:32 but how should low ranked players ever win or high ranked players ever lose... hmm 2011-12-24T02:49:00 Do I get to say "economics" one more time? 2011-12-24T02:49:45 first competition is to implement good AI for humans, second competition is to implement better AI for zombies? 2011-12-24T02:50:14 that would have to run in parallel 2011-12-24T02:50:15 more fun to do both at once :) 2011-12-24T02:50:22 Kettling: Is 'economics' guns, food, fuel, etc? :D 2011-12-24T02:50:40 s/fuel/fast food/ 2011-12-24T02:50:54 s/fuel/coke/ 2011-12-24T02:51:10 FranPeruzzi: It's a form of cooperation, my voluntary mutual agreement (and yeah, might involve guns, fuel, etc) 2011-12-24T02:51:15 an economics game would be cool 2011-12-24T02:51:37 would having the engine as Open source ruin things? 2011-12-24T02:51:44 I have to admit that I like to play Zombie Panic! now and then 2011-12-24T02:52:02 I'm a big fan of Boxhead the Zombie Wars 2011-12-24T02:52:23 If you haven't played it then stop chatting NOW! 2011-12-24T02:52:32 Zombie Panic! is like that. You come to share weapons, both humans and zombies are played by humans 2011-12-24T02:52:39 what game? Zombie Panic? 2011-12-24T02:52:43 yes 2011-12-24T02:53:07 and collecting dozens of weapons just makes you very heavy and slow 2011-12-24T02:53:17 it has also some funny comments 2011-12-24T02:53:23