2011-04-20T00:00:20 *** smiley1983 has joined #aichallenge 2011-04-20T00:01:20 If I submitted a pull request, then pushed a new change, do I need to submit another pull request? The first one was not processed, so will it include my new bugfix when they process it? 2011-04-20T00:10:55 *** smiley1983 has quit IRC (Quit: leaving) 2011-04-20T00:25:10 antimatroid around? 2011-04-20T00:31:07 He don't seem to been 2011-04-20T00:31:42 We'll give him an alert message 2011-04-20T00:31:47 animatroid: 2011-04-20T00:36:37 bhasker: I'm certainly not the authority on matters like this, but might I ask why you're looking for him? 2011-04-20T00:36:44 just curious 2011-04-20T00:40:26 was just wondering how battle resolution is happening 2011-04-20T00:40:39 especially when multiple ants are around, 2011-04-20T00:41:02 do i just go ant by ant and hit them with 1/n ants in surrounding attack radius 2011-04-20T00:41:08 and then eliminate the ones that take more than 1 2011-04-20T00:41:24 to my understanding yes 2011-04-20T00:41:28 hmm k 2011-04-20T00:42:14 are you doing engine work? 2011-04-20T00:42:59 nope 2011-04-20T00:43:08 just playing around with a bot that i was writing 2011-04-20T00:45:02 ah I see. What language? 2011-04-20T00:46:22 c++ 2011-04-20T00:47:13 have a basic bot working trying to add a few features 2011-04-20T00:47:23 this time trying to be a bit more methodical about my bot 2011-04-20T00:47:56 what did your last one do? I saw it but I can't quite remember 2011-04-20T00:48:31 that was a simpe one it just hunts for food and if no food around goes after enemy bots 2011-04-20T00:48:38 with some BFS search to find shortest paths 2011-04-20T00:49:02 and some calculations to ensure closest ants go after food 2011-04-20T00:49:07 and not all of them 2011-04-20T00:49:21 but when i meant this time around, i meant this contest:) last contest my bot was a pile of heuristic rules 2011-04-20T00:50:39 oh haha. I have no experience with ai prior to now, but I'd ceratinly like to be around when the next game gets going to help make starters and such 2011-04-20T00:50:46 *** Mathnerd314 has quit IRC (Ping timeout: 252 seconds) 2011-04-20T00:51:16 heh 2011-04-20T00:51:22 its a lot of fun i will tell you 2011-04-20T00:51:27 especially once the contest starts 2011-04-20T00:52:58 *** Frontier has joined #aichallenge 2011-04-20T00:54:01 Oh god. I just discovered a pretty serious flaw in gedit. 2011-04-20T00:54:58 I had it set to insert tabs as spaces, but I got tired of deleting tons of spaces so I changed it back to tabs and now I 2011-04-20T00:55:14 *I'm getting tons of indent errors 2011-04-20T00:55:49 just get it to an autoindent 2011-04-20T00:57:26 I do have it set to autoindent. But it evidently doesn't recognize a tab with a space length set to four as equivalent to four spaces 2011-04-20T00:58:19 I think it only considers a tab of any size as one space, which fine as long as you consistently use only spaces or only tabs, but disasterous if you switch at some point in-between 2011-04-20T00:58:27 *which is fine 2011-04-20T01:02:08 use emacs:) 2011-04-20T01:02:14 all real programmers use emacs ;) 2011-04-20T01:03:03 haha so I've heard. 2011-04-20T01:04:22 I probably will soon. But first I'm gonna report this problem. I seriously doubt it's a common problem and it took me about ten minutes of getting the same indent error to figure out what was happening 2011-04-20T01:07:59 so are you taking part this time? 2011-04-20T01:08:46 In the competition? 2011-04-20T01:11:15 yup 2011-04-20T01:11:58 oh absolutely. I am excited 2011-04-20T01:12:44 this competition is my chance to learn some real programming 2011-04-20T01:13:05 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has quit IRC (Quit: Lost terminal) 2011-04-20T01:38:31 *** janzert has joined #aichallenge 2011-04-20T02:16:14 bhasker: an ant with AT LEAST 1 damage dies 2011-04-20T02:19:49 Indeed. Probably very very important. 2011-04-20T02:51:08 *** _0x47 <_0x47!~slash@p5B25EA95.dip.t-dialin.net> has joined #aichallenge 2011-04-20T03:07:06 *** needsch has joined #aichallenge 2011-04-20T03:07:06 *** Apophis_ has joined #aichallenge 2011-04-20T03:10:50 *** Apophis has quit IRC (Ping timeout: 246 seconds) 2011-04-20T03:15:33 *** Ice_Harley has joined #aichallenge 2011-04-20T03:15:48 *** needsch has quit IRC (Quit: Leaving.) 2011-04-20T03:16:56 <_0x47> hi guys, is there any example video of the new ants game? I don't get how it's going to work. :P 2011-04-20T03:25:10 Don't know about video, but you can download visualizer and see how simple bots are playing there 2011-04-20T03:25:45 visualizer is here: https://github.com/aichallenge/aichallenge 2011-04-20T03:29:08 *** berak has joined #aichallenge 2011-04-20T03:33:36 *** ltriant has quit IRC (Quit: Computer has gone to sleep) 2011-04-20T04:10:27 _0x47: also take a look here: http://marco-leise.homedns.org/antgui/stable/visualizer.php?game_id=2 2011-04-20T04:10:30 Frontier: what's the link to your visualieser :P 2011-04-20T04:10:32 oh :P 2011-04-20T04:10:42 good timing 2011-04-20T04:10:47 i know ;) 2011-04-20T04:10:48 and blast you for not knowing hobart! 2011-04-20T04:12:20 <_0x47> Frontier Ice_Harley thank you, awesome! 2011-04-20T04:16:15 *** sigh has joined #aichallenge 2011-04-20T04:23:16 *** Naktibalda has joined #aichallenge 2011-04-20T04:52:00 What happens in the case that 2 ants from one bot and one from another are all within spawn radius of a food square? 2011-04-20T04:53:56 according to the rules, the food dissappears without spawning an ant 2011-04-20T04:54:34 hmm, that's what I thought. 2011-04-20T04:55:21 I haven't really looked but doesn't it work that way? 2011-04-20T04:56:07 seems to. But I think I would prefer making the situation impossible 2011-04-20T04:57:42 I think the battle radius should always be more than twice the size of the spawn radius. 2011-04-20T04:59:49 It seems unfair that in a situation where one bot would definitely come out as the winner and then be able to claim its food square it can instead by destroyed by a single ant 2011-04-20T05:00:12 *** _0x47 <_0x47!~slash@p5B25EA95.dip.t-dialin.net> has quit IRC (Ping timeout: 246 seconds) 2011-04-20T05:00:12 *** _0x471 <_0x471!~slash@p3EE0B7F2.dip.t-dialin.net> has joined #aichallenge 2011-04-20T05:05:21 In theory a bot that came out poorly in battles could camp right outside the spawn radius of food squares and then move in as soon as an enemy ant was about to claim it, both keeping the bot from getting a new ant and preventing its scoring by killing the other ant. I feel like this must already be at the very least an unspoken rule. Unless my logic is flawed somewhere. 2011-04-20T05:05:57 *** aerique has joined #aichallenge 2011-04-20T05:08:24 if poor bot instead of camping claim food himself stronger bot will not only score for food, but will have more problems to fight with poor one :) 2011-04-20T05:08:45 *lose score for food 2011-04-20T05:18:06 *** boegel has joined #aichallenge 2011-04-20T05:29:30 guys, battles happen before spawning 2011-04-20T05:30:17 if they are within fight range, one player may get to spawn there, but if they aren't in fight range and that happens then the food disappears 2011-04-20T05:38:57 *** yasith has joined #aichallenge 2011-04-20T05:54:28 animatroid: My argument is that there shouldn't be cases in which 2 enemies can both be in the spawn range of food, yet not in each others' battle radii 2011-04-20T06:25:08 jaycobcoleman: But increasing the battle radius does not solve that. It is still possible to end with live ants of 2 parties inside each other battle radius. 2011-04-20T06:29:58 Well of course, that's what a battle is. I'm talking about the food's spawn radius. If it's less than half the size of the battle radius then there's no way that two ants can be in the spawn radius of the same food square, therefore we no longer have to deal with the wierd situation of food occasionally disappearing 2011-04-20T06:30:56 plus I think the gameplay would be more eventful 2011-04-20T06:31:32 from what I've seen people are tending to focus almost completely on food collection as opposed to fighting and it leads to a fairly uneventful game 2011-04-20T06:32:43 decreasing spawn radius and increasing battle radius would both get rid of the strange food despawning situation and increase each bot's need to be able to handle itself in a battle 2011-04-20T06:34:05 plus larger battle radius means more complex battle dynamics and more strategy involved in keeping your ants alive, potentially even when you're going up against an enemy with considerably more ants than yourself 2011-04-20T06:46:19 *** antimatroid has quit IRC (Ping timeout: 260 seconds) 2011-04-20T06:51:38 *** _0x471 <_0x471!~slash@p3EE0B7F2.dip.t-dialin.net> has quit IRC (Ping timeout: 252 seconds) 2011-04-20T06:52:05 *** _0x47 <_0x47!~slash@p3EE0BDA0.dip.t-dialin.net> has joined #aichallenge 2011-04-20T06:52:45 *** delt0r_ has joined #aichallenge 2011-04-20T06:53:08 *** Eruonen has joined #aichallenge 2011-04-20T06:54:09 *** delt0r___ has quit IRC (Ping timeout: 246 seconds) 2011-04-20T07:13:14 jaycobcoleman: No you are wrong, whatever the battle radius compared to the food spawn radius is, it is still possible that 2 enemies enter the food spawn radius simultanuously. 2011-04-20T07:16:40 I can't see any situation where that's possible 2011-04-20T07:20:25 *** FireFly has joined #aichallenge 2011-04-20T07:29:46 Ah, it is a bit hard to explain in text. But the basic setup is that two players, A and B are approaching the food with 2 ants each and player C provides each turn a sacrifice. 2011-04-20T07:30:37 Player A and B's ant see 3 enemy ants each and player C sees 4 enemy ants. 2011-04-20T07:31:15 Player A and B's ant receive eache 2 * 1/3 + 1/4 < 1 damage and C 's ant receives 2 * 2/3 > 1 damage 2011-04-20T07:31:47 So ant C dissappears and A's and B's move closer to the food. 2011-04-20T07:31:50 Repeat 2011-04-20T07:32:12 If these exact logistics do not work, you need some more supporting ants, but basically this is the idea. 2011-04-20T07:35:53 I was under the impression they didn't move until the battle was completely resolved. If this is the case instead then I most certainly agree 2011-04-20T07:38:32 *** Eruonen has quit IRC (Ping timeout: 240 seconds) 2011-04-20T07:40:31 Hm, maybe, so your impression is that there are multiple rounds of resolving fight until no two enemy ants are in the attack radius of each other. 2011-04-20T07:40:54 Might be the case, I thought not, but I have misread things before 2011-04-20T07:41:29 It is kind of important to know this exactly 2011-04-20T07:43:44 *** antimatroid has joined #aichallenge 2011-04-20T07:44:23 animatroid: we need information on battle resolution 2011-04-20T07:48:13 what about it? 2011-04-20T07:49:31 what happened to the logs? 2011-04-20T07:49:37 they used to be text files yeah? 2011-04-20T07:49:48 i liked being able to view them in the browser :\ 2011-04-20T07:50:12 In the case that there is battle and ants from more than one bot left does the battle continue until it's completely resolved or not? 2011-04-20T07:50:40 In other words can there be multiple iterations of the battle resolution method in one turn 2011-04-20T07:50:48 no there cannot 2011-04-20T07:51:28 alright, works for me 2011-04-20T07:51:46 and also I have no idea what happened with the logs 2011-04-20T07:52:07 what was the example that lead to this question? 2011-04-20T07:52:09 what's wrong with the logs? 2011-04-20T07:52:26 my browser downloads the files now 2011-04-20T07:52:36 instead of displaying them in the browser window 2011-04-20T07:52:47 from where? 2011-04-20T07:53:03 http://contestbot.hypertriangle.com/ 2011-04-20T07:53:17 two players, A and B are approaching the food with 2 ants each and player C provides each turn a sacrifice. [03:28] Player A and B's ant see 3 enemy ants each and player C sees 4 enemy ants. [03:28] Player A and B's ant receive eache 2 * 1/3 + 1/4 < 1 damage and C 's ant receives 2 * 2/3 > 1 damage [03:29] So ant C dissappears and A's and B's move closer to the food. [03:29] Repe 2011-04-20T07:53:33 ergh 2011-04-20T07:53:40 half-assed quoting of text 2011-04-20T07:53:41 just give me a small example situation :P 2011-04-20T07:54:01 haha let me think one up 2011-04-20T07:55:19 meh. It can't really get simpler. Just ignore the times and username and what it comes down to is "there's a battle situation and ants from more than one bot remain after its resolution" 2011-04-20T07:56:21 i know it's possible, anyway, point is they stay there at the start of next turn 2011-04-20T07:56:23 antimatroid: Ignore the fact how we get in this situation: ...AACBB.. with A, B and C players 2011-04-20T07:56:24 antimatroid: wtf, it is only today's log file 2011-04-20T07:56:41 sigh: weird aye? 2011-04-20T07:57:07 and there is no difference in the HTTP headers 2011-04-20T07:59:21 jaycobcoleman: btw, I think we can get rid of any "disappearing food" by making the spawn radius 1, which i want to do anyway 2011-04-20T07:59:34 <_0x47> Is there already a local testing tool like the JBotManager (Galcon)? 2011-04-20T07:59:48 no jbotmanager 2011-04-20T07:59:52 i would love for someone to make one 2011-04-20T07:59:59 Thank you! that's exactly what I was hoping for 2011-04-20T08:00:00 but the person who made jbotmanager is quite busy 2011-04-20T08:00:47 also, are you talking about issues with the channel logs? 2011-04-20T08:01:15 <_0x47> hm, this is beyond my skills. :( It would take too much time to create it, especially as I have my exams in May 2011-04-20T08:01:19 there's something strange about the log file for today 2011-04-20T08:01:33 _0x47: I have my exams in june 2011-04-20T08:02:34 Oh my, now I see it. 2011-04-20T08:03:32 antimatroid: But what about the situation ...A*B... with * food and A and B different players? 2011-04-20T08:03:45 battle happens before spawning 2011-04-20T08:04:20 it's harder to block "collisions" 2011-04-20T08:04:35 antimatroid: But this can happen after the battle 2011-04-20T08:04:39 because you might have a*.b, then the next turn aa.b where they can collide 2011-04-20T08:04:42 *** needsch has joined #aichallenge 2011-04-20T08:04:44 *** Ice_Harley has quit IRC (Ping timeout: 252 seconds) 2011-04-20T08:04:52 woudshoo: ah yeah, then it will disappear 2011-04-20T08:05:20 it would certainly make disappearing food much less likely 2011-04-20T08:05:47 yes, that is true, but I don't see the big benefit in that. Why is it an issue if food disappears? 2011-04-20T08:06:09 I just want to avoid disappearing food in general 2011-04-20T08:06:44 it just doesn't help anyone and it detracts from the value of superior strategy 2011-04-20T08:06:56 antimatroid: meh, just download all the logs to local disk 2011-04-20T08:07:58 the alternative is to not "kill" food, but then sections of the map might turn into deadlocks 2011-04-20T08:08:01 which we don't want 2011-04-20T08:08:24 sigh: my point is that sucks :P 2011-04-20T08:10:08 I have all the logs locally anyway, so I can grep them 2011-04-20T08:10:51 bhasker: has your kid made it out into the world yet? :P 2011-04-20T08:13:35 *** Eruonen has joined #aichallenge 2011-04-20T08:31:11 *** mceier has joined #aichallenge 2011-04-20T08:33:03 *** smiley1983 has joined #aichallenge 2011-04-20T08:34:52 smiley1983: you don't make me feel as bad about my sleeping pattern 2011-04-20T08:34:53 thanks 2011-04-20T08:36:05 it's 4:30 AM here 2011-04-20T08:43:54 :P 2011-04-20T08:43:57 it's 10 40pm here 2011-04-20T08:45:04 almost 15pm here 2011-04-20T08:45:34 Ha. "Pattern" is a generous term. 2011-04-20T08:46:01 i forget how sigh referred to my sleeping routine recently 2011-04-20T08:46:06 but it made me chuckle 2011-04-20T08:47:19 I like programming after I've been awake for a while. When there's something interesting to do, this encourages late nights to become later. 2011-04-20T08:48:02 yeah, i am often most productive in the middle of the night 2011-04-20T08:48:32 animatroid: Do you want to disable fighting across water? 2011-04-20T08:48:33 jaycobcoleman: You mention that you feel it is too much about food gathering. My experiments do not bear that out. 2011-04-20T08:49:03 a lot of the same food gathering methods will show up 2011-04-20T08:49:10 the real battle arena is going to be battles 2011-04-20T08:49:25 jaycobcoleman: i do, but i don't get to make the decisions 2011-04-20T08:49:28 The matches I've seen involving more complex bots have spendt most of their time collectin food. I don't think it should be about that at all 2011-04-20T08:49:35 I think it will boil down to: 1- all bots will have very good food gathering. To do it optimally is hard, but I think that is not needed. The battles is where the good bots distinguish themselves 2011-04-20T08:49:45 nobody has done enough to do game trees for battles yet 2011-04-20T08:49:53 it requires more effort, it's easy to play around with food gathering 2011-04-20T08:50:07 what does making game trees entail? 2011-04-20T08:50:11 But I think you see it is all about food because that is where you start implementing a bot. After that is done, battle is next. 2011-04-20T08:50:39 setting up nodes, creating the tree, iterating states, evaluating states, doing backward induction, extracting moves and making them etc. 2011-04-20T08:50:51 deleting all your memory if you're using c++ etc. 2011-04-20T08:51:09 I am playing around with a bot and am just starting to look at battles. This is the hard and really important bit. 2011-04-20T08:52:03 My bot loses quite a bit to Stocha's bot. But that is because Stocha's bot somehow places its ants much better, so my ants die when it encounters it. 2011-04-20T08:52:17 i plan to partition "battle ants" into sets of who can affect each other, then iteratively use a game tree to pick moves for x amount of time 2011-04-20T08:52:19 W.r.t food gathering, both bots are similar. 2011-04-20T08:52:23 then just move the rest towards their closest friend 2011-04-20T08:52:33 woudshoo: I could be wrong, and it's probably my fault, but I feel like you may be misinterpretting me. I think it should be a lot about battles 2011-04-20T08:53:08 jaycobcoleman: I am trying to say, don't worry, it will be about battles, we are just in the phase that everybody is concentrating on food. 2011-04-20T08:53:35 alright, good 2011-04-20T08:53:44 How will you select moves for consideration in the game tree? 2011-04-20T08:54:53 *** Ice_Harley has joined #aichallenge 2011-04-20T08:55:23 That is - do you plan to consider one ant's move to be one node in the tree, or will a node represent a "strategy"? 2011-04-20T08:55:58 I just created a function that for any given point will create a map based on the water information I have that shows the minimum number of moves it will take to get to that point from every other point on the map 2011-04-20T08:56:39 How long does it take to execute on a large map? 2011-04-20T08:57:50 I haven't noticed any significant delay. I don't have my bot running with it yet though, so I've just been testing the function itself. I also haven't tested it with very large maps yet. I'll do so shortly 2011-04-20T08:58:16 I've been considering a similar thing, but haven't gotten that far yet, and I was concerned about how much time it might take. 2011-04-20T08:59:20 I have several similar ideas that haven't quite played out yet, but they seem to go pretty quickly and i can see them as being phenominally useful 2011-04-20T09:00:09 Yes, building a map of pathfinding data early in the game could be exceptionally useful in the late game. :) 2011-04-20T09:00:09 smiley1983: unsure yet, i'm tempted to take the set of non-water moves for each ant then find the set of strategies that don't result in cycles to pick out strategies for each player 2011-04-20T09:00:28 otherwise i usually just move each of my ants over x layers, then move onto the other player 2011-04-20T09:00:47 pathfinding map is easy 2011-04-20T09:01:07 actually, no it's not, 2011-04-20T09:01:26 do you feel like telling your thoughts on that? :) 2011-04-20T09:01:40 on what? 2011-04-20T09:01:44 pathfindig 2011-04-20T09:01:49 *pathfinding 2011-04-20T09:01:55 breadth first search with a queue 2011-04-20T09:02:19 a* would be nice, but i don't see it working well when you don't have a specific target 2011-04-20T09:02:24 but bhasker thinks otherwise 2011-04-20T09:02:30 Is that executed per ant, per turn, or is the data built up over successive turns? 2011-04-20T09:02:53 you can do it from all ants simultaneously if you just want shortest paths to squares from any ant 2011-04-20T09:03:06 or you can search from an ant with a queue until you find something useful etc. 2011-04-20T09:03:59 I've actually experimented with all of these and can't figure out which ones I think will be most useful 2011-04-20T09:04:28 animatroid: again, I know you aren't in charge of the code, but if I make some ants.py additions to add the capability to disable battles over water do you think I could get them pushed? 2011-04-20T09:05:20 atm i plan to do something like the following: 2011-04-20T09:05:20 for up to 50% of the time - move battle ants with game trees 2011-04-20T09:05:20 move the rest of battle ants to their closest friends 2011-04-20T09:05:20 collect food 2011-04-20T09:05:20 pair enemy ants off with friend ants and move ants appropriately 2011-04-20T09:05:20 move ants towards not recently seen squares and other heuristics 2011-04-20T09:05:40 battles are happening over water 2011-04-20T09:06:39 antimatroid: Were you worrying about crossing paths when finding food? 2011-04-20T09:06:48 yes 2011-04-20T09:06:52 and in other situations 2011-04-20T09:07:54 Maybe it isn't really an issue, if a ---> p and b --> q and their paths intersect at x. The triangle inequality tells you that the paths a ---> q and b ---> p won't be longer (in total) of course they might still intersect at x. 2011-04-20T09:07:58 my best idea for that so far is a set > to indicate when a square will be taken 2011-04-20T09:08:16 and use that information while path finding 2011-04-20T09:09:19 *** ginDeVine has joined #aichallenge 2011-04-20T09:09:26 What I am struggling a bit is moving ants simultanuously. that is, move a pack of ants such that I do not create holes in the pack. 2011-04-20T09:09:59 that's tricky 2011-04-20T09:10:26 if you're happy to deal with the occasional "collision", you can move ants onto a square already containing an ant that hasn't moved yet 2011-04-20T09:10:37 animatroid, your ideas are just as crazy and elaborate as my own. The key difference is that you probably know how to implement them to some degree :P 2011-04-20T09:10:56 i had no idea at the start of tron, experience helps :P 2011-04-20T09:12:04 antimatroid: Hm, that sounds bad, suppose you move a pack and you encounter a corridor, suddenly all ants squeeze in the corridor and if you ignore that you could lose half your ants. 2011-04-20T09:12:20 i don't think you would lose half of them 2011-04-20T09:12:27 but you may lose one or two 2011-04-20T09:12:45 Maybe I should first ignore my own ants, prepare the moves and before issuing the orders check for conflicts 2011-04-20T09:12:45 on my squares i have int ant, vector newAnts, deadAnts 2011-04-20T09:13:12 when i move an ant i add it to newAnts, then in my ++ function for state, i process the information from new ants back to ant or dead ants 2011-04-20T09:13:28 that way in the game tree you can move enemies to collide and make a lot of other stuff simpler 2011-04-20T09:14:33 an interesting problem i can't work out how to solve is the following: 2011-04-20T09:15:29 when working out all moves for a set of your ants, you get the set of moves for each ant that doesn't move them to water, then take the cartesian product of those to get the set of all moves, but then you want to efficiently eliminate all elements that result in collisions 2011-04-20T09:16:33 antimatroid: That sounds like something that can be done with bit manipulation (Knuth, the art of computer programming Vol 4). But this is just instinct, no concrete idea. 2011-04-20T09:17:41 How about something like this: The ant pack is moving toward a destination. Sort the pack by crude distance from the target. Move all the ants whose "best path" is not obstructed. Put their previous locations into a list. Now: 2011-04-20T09:18:27 Something along the lines of making the ants on the map in bit vectors, create the state after the moves and count the nr of bits. THis sounds slow, but some the bit manipulations can be very fast. (although 10000 bits is about 312 4-byte words hm) 2011-04-20T09:18:31 Iterate over this list of "holes". Attempt to fill the hole with an ant who will improve its position in relation to the target by doing so. Add this ant's previous location to the holes list. 2011-04-20T09:18:34 here is a nice way of doing it 2011-04-20T09:18:56 iteratively search from all enemy ants and move the closest ant towards them 2011-04-20T09:19:03 that would do quite well 2011-04-20T09:19:10 smiley1983: Yes, I was thinking about that, but it is work :-( 2011-04-20T09:19:29 antimatroid: that is what I do now. But I need to pace my ants, they are now one by one walking into traps. 2011-04-20T09:19:36 yep 2011-04-20T09:19:39 hence the game tree :P 2011-04-20T09:19:51 It requires coding work - it probably wouldn't be very processor-intensive. 2011-04-20T09:20:13 smiley1983: Yes, but my cycles are expensive :-) (and I am lazy) 2011-04-20T09:20:51 Yes. It's not an approach suited to use within game trees. 2011-04-20T09:25:29 i'm still not opposed to doing a crazy engine where maps are on a klein bottle 2011-04-20T09:25:35 I'd like to explore the idea of a school of fish kind of movement, where a group of ants moves as a whole, but forms around enemy ants, just outside their battle radii before simultaniously collapsing onto them. 2011-04-20T09:26:06 look up swarm intelligence ;) 2011-04-20T09:26:24 I knew you'd have something for me :) 2011-04-20T09:26:29 Another possible solution: Attempt to move every ant in the pack, assuming its way will not be blocked by existing ants - only by their destinations. Then sort through this list of orders, discarding all the ones which are impossible because the ant ahead actually couldn't move 2011-04-20T09:27:55 That has problems of its own, though. Maybe it wouldn't work at all. 2011-04-20T09:28:15 it'd work 2011-04-20T09:28:29 you may end up getting stuck with some ants until something else comes into play 2011-04-20T09:29:14 You'd have to iterate over the list until you didn't cancel any orders 2011-04-20T09:29:14 Multiple times, that is 2011-04-20T09:30:23 no, i think you could do it better than that 2011-04-20T09:30:47 you should be able to do it in a single pass 2011-04-20T09:30:50 ... by pre-sorting the list by distance? 2011-04-20T09:31:38 if an ants movement is blocked, revert that ants move and continue reverting moves of any square it lands in 2011-04-20T09:31:51 you may do more than a single pass, but it wouldn't be a long iterative process 2011-04-20T09:32:10 i guess it's just changing where the iteration is 2011-04-20T09:33:02 Yes - it's probably better in many cases, though 2011-04-20T09:33:10 If it finds the target early 2011-04-20T09:33:39 "target" being the offending order 2011-04-20T09:37:34 It's possible that you could build "chains" of linked ants in memory while you make the moves, and then if one gets cancelled, just work your way back along the chain. 2011-04-20T09:38:16 *** sigh has quit IRC (Remote host closed the connection) 2011-04-20T09:38:56 *** UncleVasya has joined #aichallenge 2011-04-20T09:39:27 you could always move them around like a snake 2011-04-20T09:39:32 badger badger badger 2011-04-20T09:39:47 mushroom! mushroom! mushroom! 2011-04-20T09:40:13 good call 2011-04-20T09:47:00 smiley1983: you inquired as to how long my mapping would take on a large map. I tested it on a 120 x 120 and it took slightly more time than I'd like, but I should think psyco should cut down on that? I don't know, but a decent fraction of a second 2011-04-20T09:48:19 *** ginDeVine has quit IRC (Quit: ginDeVine) 2011-04-20T09:48:44 do we have approximate setup/turn times? 2011-04-20T09:48:47 *** bhasker has quit IRC (Quit: bhasker) 2011-04-20T09:48:56 *** ginDeVine has joined #aichallenge 2011-04-20T09:49:12 I doubt that they'll finalise those until the contest begins, or even later 2011-04-20T09:49:22 3 second / 1 second seems likely, though 2011-04-20T09:49:33 *** helo_ has joined #aichallenge 2011-04-20T09:50:58 what are you doing during setup? 2011-04-20T09:51:04 Yeah I agree. I guess I'll have to do some real testing then 2011-04-20T09:52:04 at this point I'm not doing anything significant in setup, but that'll certainly change 2011-04-20T09:52:14 Allocating memory, initialising data structures perhaps... 2011-04-20T09:52:23 yes indeed 2011-04-20T09:52:32 I'm not now, but will be experimenting with how much I can do 2011-04-20T09:52:54 how much memory each bot has? 2011-04-20T09:53:59 I have a considerably faster function that maps the minimum number of moves until an enemy ant can be at each square based on the information I have, which I'll certainly setup at the beginning and then maintain every turn 2011-04-20T09:54:20 saving routing tables from each map location to each should take lot of memory 2011-04-20T09:54:45 Yes, I've been thinking about that 2011-04-20T09:56:21 You don't really need each to each, but identifying a subset of vital network nodes is what I'm currently thinking of 2011-04-20T09:57:22 Then you can do a very short search from current location to nearest node when needed 2011-04-20T09:57:33 smiley1983: I will not save each to each, but I will cache calculated ones 2011-04-20T09:57:47 oh shit yeah, i really need to move back to two templates in my game theory library 2011-04-20T09:58:10 game theory not that useful, huh? 2011-04-20T09:58:17 then people can use it in their ai bots and just make like 2011-04-20T09:58:17 NGame game; (with < defined on state) 2011-04-20T09:59:06 antimatroid: do you use your library in your bot? 2011-04-20T09:59:16 it might be worthwhile to find articulation points for 1x1, {1x2, 2x1}, 2x2, .. blocks 2011-04-20T09:59:44 Ice_Harley: I attempted to use the zero sum simultaneous stuff for planet wars 2011-04-20T09:59:55 but it's too hard to iteratively make a simultaneous game 2011-04-20T10:00:10 and then not obvious what move you should make 2011-04-20T10:00:24 https://github.com/antimatroid/GTL 2011-04-20T10:00:29 a decent amount of it is up there 2011-04-20T10:00:46 except extensive form games or any algorithms for mixed strategy stuff 2011-04-20T10:01:16 there's even a function in NGame.cc for getting LaTeX code for an n-player simultaneous form game using Osbornes sgamevar.sty file 2011-04-20T10:01:54 articulation points is wrong 2011-04-20T10:02:04 find parts that will block those blocks 2011-04-20T10:02:09 in NAlgorithms.h: //returns all mixed strategy Nash equilibria of the given game .... //haha, good luck writing this :) 2011-04-20T10:02:17 due to the wrapping, articulation points are impossible 2011-04-20T10:02:31 Ice_Harley: are you at all familiar with how hard that is? :P 2011-04-20T10:02:56 there is no implementation that does it 2011-04-20T10:03:15 gambit has the best functions, but they aren't perfect 2011-04-20T10:03:44 the wiki stuff might be helpful :P 2011-04-20T10:04:10 i'm still adding it to github and not finding a lot of time 2011-04-20T10:05:34 Can you point me to any reading material about the articulation points you were talking about, or give a very brief run-down on what they are and why they don't work on a toroidal map? 2011-04-20T10:06:16 an atriculation point is a node that separates a graph if you remove it and all edges it is linked to 2011-04-20T10:06:30 with the symmetric maps, i'm pretty sure no articulation points will exist 2011-04-20T10:06:38 and they're slightly painful to find anyway 2011-04-20T10:07:08 Right. Thankyou :) 2011-04-20T10:07:09 i think what you really want to do is try to find "passages" such blocks can't traverse through that perhaps a single ant can? 2011-04-20T10:08:15 an idea i have for going around "traps" is as follows 2011-04-20T10:08:42 for any square adjacent to an enemy, see if an ant would die in that square if you had ants in any possible squares around it at the end of that turn 2011-04-20T10:08:59 if it dies no matter what, mark it as a trap square and treat it like water with path finding 2011-04-20T10:11:03 Will the vision, attack and spawn radii be fixed at some point, or will bots have to cope with a variable input in the final contest? 2011-04-20T10:11:06 *** ginDeVine has quit IRC (Quit: ginDeVine) 2011-04-20T10:11:14 it will be fixed in the final contest 2011-04-20T10:11:19 but bots should not hard code it 2011-04-20T10:11:24 at least not until it is finalised 2011-04-20T10:11:43 the parameters have been added right at the start so people can't bitch about parameters changing if they need to :) 2011-04-20T10:11:57 you might remember some bots broke when we varied the number of planets on maps in pw 2011-04-20T10:11:57 Yes, I like that. :) 2011-04-20T10:12:16 btw how did you like my new maps? :P 2011-04-20T10:12:25 Ha. Yes. 2011-04-20T10:12:25 that was my bad 2011-04-20T10:12:32 New maps for this game or last? 2011-04-20T10:12:37 planet wars 2011-04-20T10:12:51 the point symmetry ones? 2011-04-20T10:12:53 i split the contestants on that one, some loved them, others hated the late change 2011-04-20T10:12:54 yeah 2011-04-20T10:13:09 Sorry, line symmetry. 2011-04-20T10:13:15 I thought it was... interesting 2011-04-20T10:13:30 yeah, i didn't know what line/point symmetry was until after i went and made the map gen 2011-04-20T10:13:36 it's fun solving stuff on your own 2011-04-20T10:13:40 It did change things a bit, but any bots which couldn't cope had hard-coded unreasonable assumptions. 2011-04-20T10:13:47 IMHO 2011-04-20T10:13:58 that was my opinion 2011-04-20T10:14:17 and i had always assumed maps would change, so imo it was unfair on me to let people get away with assumptions they made 2011-04-20T10:14:30 It would have been better to have them from the beginning, but I found it interesting and inoffensive. 2011-04-20T10:14:38 but it was a LATE change, the plan is to not do that again and get variation early on 2011-04-20T10:15:16 i got paid out for doing it in polar coordinates 2011-04-20T10:15:25 pesky muggles 2011-04-20T10:15:31 polar? 2011-04-20T10:15:54 (r, theta) rather than (x, y) 2011-04-20T10:16:05 and i would randomly pick r and theta, then convert to x and y 2011-04-20T10:16:32 but it was like, pick a distance from the center, then pick an angle, then I would do symmetry by translating the angles 2011-04-20T10:17:09 I have a tendency to change my mind about standards in the middle of a project, and do new stuff the new way without changing all old stuff, and sometimes it leads to big headaches. :) 2011-04-20T10:17:10 i much prefer the problem of tiling maps on a torus with this 2011-04-20T10:17:28 i'm like that 2011-04-20T10:17:32 I like the fact that they specified row, column 2011-04-20T10:17:37 Saved me from confusing myself 2011-04-20T10:17:42 you can thank me for that :P 2011-04-20T10:17:49 Thank you :) 2011-04-20T10:17:49 it was x,y and i wouldn't shut up about it 2011-04-20T10:17:51 so it changed 2011-04-20T10:18:02 x,y indexing is inferior and i actually have a valid argument 2011-04-20T10:18:05 *** UncleVasya has quit IRC (Ping timeout: 260 seconds) 2011-04-20T10:18:20 with x,y indexing, you need to transpose the map when inputting/outputting 2011-04-20T10:18:23 In that case, than you indeed. Yes, you do. It's faster to print on the screen 2011-04-20T10:18:26 Yes 2011-04-20T10:18:32 which is confusing as hell 2011-04-20T10:18:53 Indeed, that's how I always get into trouble 2011-04-20T10:19:13