2011-07-02T00:01:39 *** SSNikolaevich has quit IRC (Quit: Konversation terminated!) 2011-07-02T00:01:58 *** SSNikolaevich has joined #aichallenge 2011-07-02T00:18:29 *** sigh_ has joined #aichallenge 2011-07-02T00:21:24 *** sigh has quit IRC (Ping timeout: 240 seconds) 2011-07-02T00:24:20 *** sigh_ has quit IRC (Remote host closed the connection) 2011-07-02T00:45:21 *** SSNikolaevich has quit IRC (Remote host closed the connection) 2011-07-02T01:05:36 *** sigh has joined #aichallenge 2011-07-02T01:06:56 *** jmreardon has quit IRC (Quit: jmreardon) 2011-07-02T01:15:08 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has quit IRC (Quit: Lost terminal) 2011-07-02T01:18:48 *** Chris_0076 has joined #aichallenge 2011-07-02T01:33:08 *** okayzed is now known as okay 2011-07-02T01:50:03 *** okay is now known as okayzed 2011-07-02T02:11:52 *** Palmik has joined #aichallenge 2011-07-02T02:30:51 *** mleise has quit IRC (Quit: Leaving.) 2011-07-02T02:41:30 *** Chris_0076 has quit IRC (Quit: Leaving) 2011-07-02T02:56:33 *** gobang has joined #aichallenge 2011-07-02T03:08:27 *** sigh_ has joined #aichallenge 2011-07-02T03:09:58 *** sigh has quit IRC (Ping timeout: 250 seconds) 2011-07-02T03:16:19 *** Palmik has quit IRC (Remote host closed the connection) 2011-07-02T03:30:23 *** nux67 has joined #aichallenge 2011-07-02T03:51:52 *** Mathnerd314 has quit IRC (Ping timeout: 240 seconds) 2011-07-02T03:58:25 *** mcstar has joined #aichallenge 2011-07-02T04:02:43 *** mleise has joined #aichallenge 2011-07-02T04:08:04 yes, it is the sound of my coffee dripping 2011-07-02T04:15:30 *** amstan has quit IRC (Ping timeout: 260 seconds) 2011-07-02T04:31:19 *** smiley1983 has joined #aichallenge 2011-07-02T05:04:59 *** sir_macelon has joined #aichallenge 2011-07-02T05:05:13 hi 2011-07-02T05:05:35 look at this game turn: http://aichallengebeta.hypertriangle.com/visualizer.php?game=52727&turn=42&row=61&col=36 2011-07-02T05:05:47 do you think food has been spawned fairly there? 2011-07-02T05:10:07 certainly not, since flagcapper won 2011-07-02T05:13:23 sir_macelon: pardon my insolence, but why do you expect the engine to be "fair" all the time? maybe ants should believe in god(engine) and pray to him to be fair, hm. 2011-07-02T05:17:39 *** sir_macelon has quit IRC (Quit: Page closed) 2011-07-02T05:18:50 *** sir_macelon has joined #aichallenge 2011-07-02T05:19:41 mcstar: but that's why have provided symmaterical spawning, so it should be fair, but it seems it isn't in this kind of symmetry 2011-07-02T05:22:02 just because spawning is symmetric it doesnt imply it is fair too 2011-07-02T05:22:24 but these differences should average out 2011-07-02T05:22:41 maybe in some other situation your bot will get food similarly 2011-07-02T05:23:43 if the engine pays too much attention to this, and precisely controls spawning, it will control the game eventually 2011-07-02T05:23:56 id rather leave it more random 2011-07-02T05:25:26 i mean it assumes the food should be spawned equaliy distant from starting location, but in some situations, like the one i showed, it spawns a group of food, which luckily favours someone who is luckily around there 2011-07-02T05:27:46 if you would have this kind of symmetry for 4 players, there could be a group of 4 adjacent food spawned together, i don't like such spawning as it brings too much luck 2011-07-02T05:28:47 so your problem is, that foods spawned too close together? i agree thats not ideal 2011-07-02T05:29:58 this is exactly, what i mean 2011-07-02T05:30:06 look at this game: http://aichallengebeta.hypertriangle.com/visualizer.php?game=52240 2011-07-02T05:30:20 most of the food is spawned in the middle 2011-07-02T05:30:52 luckily my ant was there first and got it all, dominating the game and a very quick win in under 100 turns 2011-07-02T05:31:03 ok, thats a good point, i thought you had a different problem 2011-07-02T05:31:56 i only care here about adjacent food spawning which is not fair in my opinion 2011-07-02T05:32:16 in prev. version of the food spawning alg. wasnt there a check to not spawn food too close together? 2011-07-02T05:32:26 i dont really remmeber seeing such situtations 2011-07-02T05:32:30 *** foRei has joined #aichallenge 2011-07-02T05:33:05 why dont you @later tell mcleopold? 2011-07-02T05:34:23 ok, i'll do 2011-07-02T05:36:16 contestbot: later tell mcleopold look at this game: http://aichallengebeta.hypertriangle.com/visualizer.php?game=52240 I think this kind of food spawning is not fair, I am against such adjacent food spawning, see my earlier comments about it 2011-07-02T05:36:16 sir_macelon: Yep. 2011-07-02T05:36:28 *** sir_macelon has quit IRC (Quit: Page closed) 2011-07-02T05:49:33 *** gobang has quit IRC (Ping timeout: 255 seconds) 2011-07-02T05:50:25 mcstar: the reason you didn't see it previously was because the tile symmetry never has symmetrically equivalent squares close like that 2011-07-02T05:50:41 i hadn't thought of this problem, i'm not really sure there's anything we can do about it while keeping symmetry 2011-07-02T05:51:26 ? 2011-07-02T05:51:31 i will at least be able to hide where these "middle" parts of the map are with translations 2011-07-02T05:51:44 just call a proper distance function on the foods-to-be-spawned 2011-07-02T05:51:55 not that simple 2011-07-02T05:52:03 then you're excluding some land squares from ever spawning food 2011-07-02T05:52:05 why? if its too low, just ignore it 2011-07-02T05:52:05 is that fair? 2011-07-02T05:52:28 we might have to do that though 2011-07-02T05:52:30 i guess its better than now, dont you think? 2011-07-02T05:52:49 ^^ yeah, i just got introduced to the "problem" :P 2011-07-02T05:53:54 (btw my personal opinion is that symmetric food spawning sucks ass) 2011-07-02T05:53:57 *** sigh has joined #aichallenge 2011-07-02T05:54:23 and ill never write opinion with 2 l's again 2011-07-02T05:55:04 you would prefer asymmetric food spawning? 2011-07-02T05:55:11 yes 2011-07-02T05:55:12 i've always been a big fan of symmetry for ai 2011-07-02T05:55:22 the more random the better 2011-07-02T05:55:24 *** sigh__ has joined #aichallenge 2011-07-02T05:55:41 do you not care about the accuracy of rankings? 2011-07-02T05:56:03 i dont think this will affect ranking in the long run 2011-07-02T05:56:28 we're quite limited on just how many games we can run for each bot during the final contest 2011-07-02T05:56:37 and ranking is already quite iffy 2011-07-02T05:56:43 *** sigh_ has quit IRC (Ping timeout: 240 seconds) 2011-07-02T05:58:02 look, as far as i see, the food spawning problem is significant in the first phase of the game: the expansion phase. and yes, there should be some fairness guaranteed, not to help any bot more than any other one to gain much territory in a short time 2011-07-02T05:58:54 im not sure yet what is the solution to that 2011-07-02T05:59:33 *** sigh has quit IRC (Ping timeout: 276 seconds) 2011-07-02T05:59:37 you can argue that symmatric spawning is, maybe, but id turn it off after a couple of turns 2011-07-02T06:02:37 and obviously the engine can control the expansion rate of the bots by controlling the amount of food it spawns, cotrolling the probability depending on the manhattan distance from each bot 2011-07-02T06:03:02 con-trolling 2011-07-02T06:03:09 that is a nice phrase 2011-07-02T06:03:22 if you troll, i will con-troll 2011-07-02T06:05:54 i'm not willing to pick spawn locations based on the current ants locations 2011-07-02T06:06:05 it opens up the game to influencing the spawn code as well 2011-07-02T06:07:00 no, i dont really mean that 2011-07-02T06:07:37 consider this: the edge of the blob(the area ants can reach) grows with the turns 2011-07-02T06:08:16 and i would change the probability density function with the value of this distance(the edge of the blob) 2011-07-02T06:08:55 and this pdf of course depends upon the distance from the spawn locations 2011-07-02T06:09:11 not to cause a cascade-like expansion 2011-07-02T06:09:29 im not saying that it should depend on individual ant locations 2011-07-02T06:09:50 so there could not be a back-propagation effect 2011-07-02T06:10:45 if you really want a fair game, i think, first of all you have to control the fast initial expansion 2011-07-02T06:12:00 antimatroid1: i got it 2011-07-02T06:12:27 Dijkstra's algorithm cant work with negative edge costs 2011-07-02T06:12:37 id would loop forever on those edges 2011-07-02T06:12:54 because it would always make the cost ever smaller 2011-07-02T06:13:44 and basically the criterion: h(x) <= g(x,y)+h(y) guarantees, that the a* rewritten as dijkstra's cant have negative edge cost 2011-07-02T06:15:50 i should find my linux tux, put it next to my monitor, and talk to it 2011-07-02T06:16:09 its easier than to write this on irc, and would have the same effect 2011-07-02T06:25:06 *** locutus2 has joined #aichallenge 2011-07-02T06:32:07 ah, ok in the properties section it says the same thing 2011-07-02T06:32:24 antimatroid1: are you there btw? i really hate talking to myself 2011-07-02T06:32:42 *** antimatroid has joined #aichallenge 2011-07-02T06:35:11 *** antimatroid1 has quit IRC (Ping timeout: 240 seconds) 2011-07-02T06:35:27 antimatroid: how much did yoi get from that ive written? 2011-07-02T06:35:40 up to "its easier than to write this on irc, and would have the same effect" 2011-07-02T06:35:45 sorry, just got back from the shop :P 2011-07-02T06:36:01 ah, ok in the properties section it says the same thing 2011-07-02T06:36:09 only missed this then 2011-07-02T06:36:22 antimatroid1: are you there btw? i really hate talking to myself 2011-07-02T06:36:31 and this one, but its sort of redundant 2011-07-02T06:37:27 the heuristic thing? 2011-07-02T06:37:30 you are right i think 2011-07-02T06:39:37 what about food spawning? obviously im not asking you to change anything, just asking for your opinion 2011-07-02T06:40:06 i wouldn't preference any of the squares more i don't think 2011-07-02T06:40:12 but i dunno 2011-07-02T06:48:08 *** sergey_ has joined #aichallenge 2011-07-02T06:50:12 *** sergey_ is now known as SSNikolaevich 2011-07-02T07:24:11 reading up on D* and skimming Sven Koenig's page it seems that d* could be an optimal solution, since it reuses previous paths, for new searches, and can incorporate changing environments well, but i havent fully comprehended it though 2011-07-02T07:25:04 sounds interesting 2011-07-02T07:25:17 i just came to ask you a question about the a* with negative weight edges 2011-07-02T07:25:32 wouldn't that imply you need to use the open/closed set way with a non-monotone heuristic? 2011-07-02T07:25:37 making the wikipedia page wrong? 2011-07-02T07:26:08 mcstar: does that sound right? 2011-07-02T07:28:47 wiki says for negative edges there is the bellman ford, which detects negative cycles, and theres a variation for directed graphs with negative edges, but no negative cycles 2011-07-02T07:28:50 Yes. open/closed sets is needed 2011-07-02T07:29:22 but i would really leave this alone cause it got nothing to do with applications 2011-07-02T07:29:30 with our 2011-07-02T07:30:07 negative weights is alowed only in acyclic graphs 2011-07-02T07:30:11 i'm trying to find the exact statement that confused me 2011-07-02T07:31:03 i can't find the comment :\ 2011-07-02T07:31:27 i guess they're allowed always, but algorithms broke down, obviously, but you can say, that if you find a negative cycle, you dont traverse it more than once 2011-07-02T07:31:47 or at all 2011-07-02T07:32:02 i think it depends what you are really representing with edges and costs.... 2011-07-02T07:32:16 yes 2011-07-02T07:32:46 mcstar: there's another way of ordering your search queue items 2011-07-02T07:32:50 the closed set is an improvement over not having them, cause it makes possible not to check again a node if it was added to the closed set 2011-07-02T07:33:19 when there is a draw with f, you have fifo yes?, i say split on h, but then they also suggest making it lifo after that, or without even splitting on h 2011-07-02T07:33:21 antimatroid: the min(g+h) then min(h) didnt help mcuh 2011-07-02T07:33:42 mcstar: but i don't ever search a location twice without the closet set :\ 2011-07-02T07:33:54 once i add a location to the search queue, i mark it as visited and don't touch it again 2011-07-02T07:34:23 i add a square to closed if i added all its nbors to open 2011-07-02T07:34:35 might be the same in essence 2011-07-02T07:35:02 a* & d* both good enought. But weight function is main challenge. It depends on enemy ants around etc. 2011-07-02T07:35:06 i may be wrong, but i figure once you add a location to the search queue, you can't possibly find a shorter path to it 2011-07-02T07:35:12 otherwise it would have already been added 2011-07-02T07:35:29 * antimatroid still doesn't get weighted a*? 2011-07-02T07:35:36 its wrong. 2011-07-02T07:35:37 its just c*h 2011-07-02T07:35:48 ok? just scale up the heuristics 2011-07-02T07:36:03 yeah, but i don't get why it improves it 2011-07-02T07:36:04 etc. etc. 2011-07-02T07:36:06 the criterion that h(target)=0 will sill hold 2011-07-02T07:36:12 i like to wrap my head around the why of things 2011-07-02T07:36:15 but you dont check that many nodes 2011-07-02T07:36:22 but you sacrifice optimality 2011-07-02T07:36:47 you scale up the gap between solutions 2011-07-02T07:36:53 but you dont scale up g 2011-07-02T07:36:59 sorry, between heuristics 2011-07-02T07:37:06 hmm, i kind of see i guess 2011-07-02T07:37:15 i'm going to write myself some testing stuff for path finding 2011-07-02T07:37:18 and test various things out 2011-07-02T07:37:30 sure 2011-07-02T07:37:40 maybe add a graphics library yourself 2011-07-02T07:37:57 i have no experience with guis :P 2011-07-02T07:38:09 gui!=graohics library 2011-07-02T07:38:33 well, graphics libraries then 2011-07-02T07:38:36 and it is not required to make full search. Search with fixed max depth is enought to determine the aproximate direction for ant. I tested this on my test cases 2011-07-02T07:38:58 SSNikolaevich: have you played around with pathfinding with multiple targets yet? 2011-07-02T07:39:12 i have got it working with a*, but mcstar has me interested in d* 2011-07-02T07:40:00 yes. It is my real live work :) 2011-07-02T07:41:08 SSNikolaevich: you work in ai? robotics maybe? 2011-07-02T07:41:17 CAD 2011-07-02T07:41:31 placement & routing 2011-07-02T07:42:05 hm, i never really considered this, but cad/graphics applications has many of these ai algorithms 2011-07-02T07:42:11 SSNikolaevich: what is your take on path finding with multiple targets then? 2011-07-02T07:42:30 i've just been using h(x) = min_T mdist(x, T) 2011-07-02T07:43:21 no. I'm fighting with timeouts in my bot now :( 2011-07-02T07:43:27 which language? 2011-07-02T07:43:44 SSNikolaevich: are you used to supercomputers? :D 2011-07-02T07:43:50 python 2011-07-02T07:44:07 that's going to be really slow for this contest :) 2011-07-02T07:44:18 computational claster in LAN 2011-07-02T07:44:18 that's my claim anyway 2011-07-02T07:44:36 i like clusters too, wrote for them in mpi 2011-07-02T07:44:44 mcstar: http://ants.zeroviz.us/AxIF2/ 5 player symmetry still weirds me out :P 2011-07-02T07:45:35 we using OpenMP, Erlang ant task scheduling 2011-07-02T07:46:00 omg erlang? 2011-07-02T07:46:09 i read that its good for concurrency 2011-07-02T07:46:40 well, we use c, its for physics simulations, no need for functional languages :) 2011-07-02T07:46:42 Bot may be slow because of algorithms, not because language. 2011-07-02T07:47:41 Functional languages is good for multiprocess applications because they have no side effects. 2011-07-02T07:48:33 antimatroid: nice idea on the superlattice 2011-07-02T07:49:09 mcstar: the recursive backtrack maze generation technique seemed easiest :P 2011-07-02T07:49:14 so i had to do that to implement it :P 2011-07-02T07:50:36 Start width search from food is good, if your bot have anought time 2011-07-02T08:12:35 ... is it worth doing the searches in reverse 2011-07-02T08:12:41 just so you can get the path out easier 2011-07-02T08:13:08 i doubt it, i imagine it'd be better to minimise the number of targets if possible 2011-07-02T08:17:43 and how you search path too food? from every ant? From closest subset of ants? The main goal of searching from food it is one search instead multiple 2011-07-02T08:24:50 i search from all (my) ants to all food 2011-07-02T08:24:52 or vice versa 2011-07-02T08:25:21 i'm going to use a deque of moves for my paths, so it doesn't matter which way i do it 2011-07-02T08:25:34 then i might get it to split which way it does it based on the size of the sources and target sets 2011-07-02T08:25:35 i have an idea: just add all the h together h_f1+h_f2+h_f3 and use this as a heuristic function searching for food from 1 ant 2011-07-02T08:25:43 its very expensive in term of time 2011-07-02T08:25:57 SSNikolaevich: how? 2011-07-02T08:26:04 i would have expected it to be quicker.. 2011-07-02T08:26:18 for just bfs's it would surely be quicker 2011-07-02T08:26:30 my heuristic function is a bit iffy though when there is a large number of targets 2011-07-02T08:26:51 oh, maybe not for bfs 2011-07-02T08:26:52 hmm, dunno 2011-07-02T08:26:53 width search from food is too expencive in python. I'm search path to food from 5 closes ants 2011-07-02T08:26:57 i'm writing up test stuff now 2011-07-02T08:27:00 closes=closest 2011-07-02T08:27:10 c++ <3 :) 2011-07-02T08:27:22 ;) 2011-07-02T08:29:42 C++ is good. But is require free time. I have work to do, family, which don't allow me to write bot in home. That's why i use python: rapid development and prototyping 2011-07-02T08:33:20 i find c++ easier to program in anyway, as it's what i know best 2011-07-02T08:33:36 my python skills are minimal at best 2011-07-02T08:34:36 I'm familiar with C++, python and erlang. But don't have anought free time to write in C++ :( 2011-07-02T08:36:42 games is stoped now. Anybody knows why? 2011-07-02T08:38:04 yeah, i know that was a stupid idea in itself, but ill improve on ti 2011-07-02T08:40:09 *** SSNikolaevich has quit IRC (Quit: Konversation terminated!) 2011-07-02T08:57:41 *** olexs has joined #aichallenge 2011-07-02T09:06:11 *** jbrechtel has joined #aichallenge 2011-07-02T09:14:57 *** jbrechtel has quit IRC (Read error: Connection reset by peer) 2011-07-02T09:21:42 *** choas has joined #aichallenge 2011-07-02T09:33:43 *** jmreardon has joined #aichallenge 2011-07-02T09:45:04 *** jbrechtel has joined #aichallenge 2011-07-02T09:45:17 *** onensora has joined #aichallenge 2011-07-02T10:03:14 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has joined #aichallenge 2011-07-02T10:05:08 anyone know how i'd do a comparison operator for structs in c++ to do l/fifo for priorit_queues 2011-07-02T10:07:34 i know this isnt what you want, but why not just write a function which test all .elements ? 2011-07-02T10:07:48 (sure you can even overload =, cant you) 2011-07-02T10:09:38 what do you mean? 2011-07-02T10:09:47 i want a priority queue of search items 2011-07-02T10:10:06 and i want to be able to set the comparison operator with variables when calling a search 2011-07-02T10:11:06 i think i'ma need to mark numbers for order in which i found a square 2011-07-02T10:11:22 i can overload = and == 2011-07-02T10:12:11 yeah, i meant == 2011-07-02T10:12:47 im just gonna use a list, and always sort it by f 2011-07-02T10:13:09 when the list is almost sorted, the sort is very fast 2011-07-02T10:13:27 (for the better algorithms at least) 2011-07-02T10:14:09 *** choas has quit IRC (Ping timeout: 255 seconds) 2011-07-02T10:24:18 *** antimatroid1 has joined #aichallenge 2011-07-02T10:26:26 *** antimatroid has quit IRC (Ping timeout: 240 seconds) 2011-07-02T10:31:47 *** pcavs has joined #aichallenge 2011-07-02T10:32:36 I'm sure this is asked 100+ times a day, but I was wondering if there was a clojure ai challenge starter kit? 2011-07-02T10:33:06 If not, is anyone working on one and would like some help, or if I were to create my own, what would I need to read up on in terms of rules? 2011-07-02T10:33:29 *** O01eg has joined #aichallenge 2011-07-02T10:33:55 pcavs: it isnt asked a 100 times a day 2011-07-02T10:34:22 and im sure it would be welcomed if you were to write one 2011-07-02T10:34:46 and read up on hte github wiki 2011-07-02T10:34:50 all the rules are there 2011-07-02T10:35:02 yeah, specs wiki page is the best place to look 2011-07-02T10:35:05 or you can take a look at a starter pack already written 2011-07-02T10:35:14 if you know c++ i'm of the biased opinion that my starter bot for that is worth basing bots off :) 2011-07-02T10:36:02 well, he is writing in closure, so pls rather take a look at the common lisp version 2011-07-02T10:36:08 clojure 2011-07-02T10:36:42 *** olexs has quit IRC (Quit: Leaving.) 2011-07-02T10:37:06 yeah, I think I'll take a look at the CL bot I guess 2011-07-02T10:37:23 Thanks for being helpful, I will probably end up bothering this group with more questions in the future 2011-07-02T10:37:43 i guess the channel was created for this exact purpose :) 2011-07-02T10:37:52 haha, indeed ;) 2011-07-02T10:39:52 there's also a scala bot, which might be worth looking at. Someone named tnbd was asking about a clojure starter pack recently, but from grepping the channel logs it seems there hasn't been any mention of one being developed yet 2011-07-02T10:40:58 sigh__, smiley1983: what's the best way for me to make a priority_queue in c++ fifo or lifo on elements that are otherwise "equal"? 2011-07-02T10:41:12 is it to just store a variable indicating the order in which i added items and comparing based off that? 2011-07-02T10:42:20 I don't know enough about c++ to answer that 2011-07-02T10:42:28 antimatroid1: yes, that sounds reasonable 2011-07-02T10:43:37 damn, cheers 2011-07-02T10:43:42 i was hoping there'd be something neater :P 2011-07-02T10:44:02 smiley1983: maybe it was asked yesterday? i think i let that question pass me by 2011-07-02T10:44:24 i vaguely remember someone asking something like that recently 2011-07-02T10:44:24 2011-07-01T09:00:18 Is anyone creating a clojure starter pack? 2011-07-02T10:47:02 antimatroid1: cant you just put the pointers to those structs in a vector? (push, or something?) 2011-07-02T10:47:26 mcstar: i want to use normal priority_queues as implemented in c++ 2011-07-02T10:47:35 they're implemented as a heap as far as i know 2011-07-02T10:47:42 (im not saying i know it better, im just curious, i dont mind a small cpp sticking to me) 2011-07-02T10:47:58 i give it a type, something to store it in (a vector) and a comparison operator for determining "max" elements (which can give a min operator if you want) 2011-07-02T10:48:37 you can do what mcstar says and store vectors instead of values, but that is much more code to maintain 2011-07-02T10:48:55 or the quick hacky solution is to store x+k*epsilon for equals values x 2011-07-02T10:49:12 like priority_queue, greater > will give a priority queue of integers, stored in a vector that is "sorted" with >, so when you retrieve the last element (top) then you get the least element in the priority queue 2011-07-02T10:49:47 antimatroid1: you'd pretty much be rolling your own implementation for that : 2011-07-02T10:49:49 sigh__: i wouldn't use +k*e, i just add the code into my comparison operator 2011-07-02T10:50:11 yup, that's normal 2011-07-02T10:50:58 well, i did this with a second comparison 2011-07-02T10:51:06 i don't like that the comparison thing needs to be a struct/class either :P 2011-07-02T10:51:18 it would be cooler if i could just pass it an operator/function, pesky c++ 2011-07-02T10:51:38 i checked for (if (< sum min-sum) (do this and this)) 2011-07-02T10:51:49 (if (= sum min-sum) (do another thing)) 2011-07-02T10:52:05 oh, i could min/max g for fifo lifo? 2011-07-02T10:52:06 where sum is g+h 2011-07-02T10:52:30 and where its =, i check for (< h hmin) 2011-07-02T10:52:34 i think that'd work? 2011-07-02T10:53:01 close enough to fifo/lifo for what i want anyway, mcstar: does that sound right? :P 2011-07-02T10:53:17 sorry, which one? :S 2011-07-02T10:53:42 g(x) := mdist(source, x) 2011-07-02T10:53:56 no it's not 2011-07-02T10:53:59 it's the path distance to x 2011-07-02T10:54:15 that wouldn't do it :( 2011-07-02T10:54:18 poop 2011-07-02T10:54:28 antimatroid1: do you get what i wrote? 2011-07-02T10:54:36 there's quite a few options though :\ 2011-07-02T10:54:46 nope :P 2011-07-02T10:54:50 :( 2011-07-02T10:55:23 *** locutus2 has quit IRC (Ping timeout: 240 seconds) 2011-07-02T10:55:24 i mean you can sort your queue, with <, but ask the sort to behave differently when its = 2011-07-02T10:55:43 yeah, okay sure 2011-07-02T10:55:46 but i already have that 2011-07-02T10:56:05 my problem is what to do with "do another thing" 2011-07-02T10:56:25 or something like that? i dunno 2011-07-02T10:56:31 i get confused too easily :P 2011-07-02T10:57:03 well, i update hmin, min-sum, and i choose that element for the next one from the open set 2011-07-02T10:57:27 i don't know why you bother with open/closed sets :\ 2011-07-02T10:57:31 theres no point in sorting again in the equal elements 2011-07-02T10:57:39 thats just a name fo rit 2011-07-02T10:59:52 mcstar: if you searched from food to ants... you could priotise searchs from squares with the most adjacent food items first? 2011-07-02T10:59:56 would that work :s 2011-07-02T11:00:04 prioritise* 2011-07-02T11:00:59 *** UncleVasya has joined #aichallenge 2011-07-02T11:01:30 you mean when the search runs? 2011-07-02T11:02:01 yeah 2011-07-02T11:02:04 because you could try searching from a food square where are more food squares close by, that makes sense 2011-07-02T11:02:33 let me think 2011-07-02T11:02:51 i think if you did that with fifo on a draw then you'd get the optimal solution? and not with lifo 2011-07-02T11:02:56 if my intuition is correct 2011-07-02T11:04:28 well, if you use strictly f=g+h, where h is the manhattan distance, you can prioritise whatever the way you like 2011-07-02T11:04:57 no i mean like, min f, min h, if both of those are equal, then lifo or fifo 2011-07-02T11:05:07 but since you are searching for an ant now, i dont think you should be preferring food squares 2011-07-02T11:05:44 there cant be many such cases where both are equal 2011-07-02T11:05:52 the open set supposed to be small 2011-07-02T11:06:07 and you put 2 constarints on it, soo 2011-07-02T11:06:35 hey, what can i say: try it, and we will see :) 2011-07-02T11:22:02 *** Accoun has quit IRC (Read error: Connection reset by peer) 2011-07-02T11:24:28 lol, min/max g == max/min h, obviously :P 2011-07-02T11:29:06 *** UncleVasya has quit IRC (Ping timeout: 260 seconds) 2011-07-02T11:32:28 *** Accoun has joined #aichallenge 2011-07-02T11:37:38 only when you have min f 2011-07-02T11:39:36 i mean wtf 2011-07-02T11:39:52 min g and min f is both 0 2011-07-02T11:40:10 max g and max f can be arbitrarily large 2011-07-02T11:41:08 but yeah, for f=c, its true 2011-07-02T11:42:00 0,c and c,0 2011-07-02T11:42:07 yeah, i always plan to min f at the start 2011-07-02T12:14:05 *** olexs has joined #aichallenge 2011-07-02T12:15:48 *** sir_macelon has joined #aichallenge 2011-07-02T12:16:47 contestbot: rankings 2011-07-02T12:16:55 sir_macelon: Top 10 players: sir_macelon(97.7), FlagCapper(95.5), Sto03RepliCrush(84.2), antimatroid(83.2), sto_greed_002(81.1), OlexandrT(80.0), Zaphtest02(79.9), UncleVasya(78.8), Zaphus(78.2), BaronTrozo(77.5) 2011-07-02T12:19:06 *** gobang has joined #aichallenge 2011-07-02T12:25:35 another lucky food spawn boosted me for the win: http://aichallengebeta.hypertriangle.com/visualizer.php?game=54058&turn=75&row=0&col=0 2011-07-02T12:29:02 *** pvarga has joined #aichallenge 2011-07-02T12:31:54 *** choas has joined #aichallenge 2011-07-02T12:44:00 *** gobang has quit IRC (Ping timeout: 255 seconds) 2011-07-02T12:46:00 *** sigh__ has quit IRC (Remote host closed the connection) 2011-07-02T12:50:45 *** smiley1983 has quit IRC (Quit: leaving) 2011-07-02T12:57:57 *** O01eg has quit IRC (Ping timeout: 255 seconds) 2011-07-02T13:09:47 *** McLeopold has joined #aichallenge 2011-07-02T13:10:40 *** O01eg has joined #aichallenge 2011-07-02T13:11:05 antimatroid1: what do you think about not spawning food next to each other? That would effectively make the land area control percentage not match food spawning percentages 2011-07-02T13:11:05 *** okayzed is now known as okay 2011-07-02T13:12:11 The dead area would always be on symmetric lines, which would be fair 2011-07-02T13:12:55 The fact that your on land that doesn't spawn may balance with the fact the food could spawn 4 squares apart sometimes 2011-07-02T13:13:31 It seems similar to the "no planets closer than 1" of the last contest 2011-07-02T13:15:04 *** gobang has joined #aichallenge 2011-07-02T13:15:40 *** amstan has joined #aichallenge 2011-07-02T13:15:40 *** ChanServ sets mode: +o amstan 2011-07-02T13:15:58 amstan: hi 2011-07-02T13:16:18 McLeopold: hi 2011-07-02T13:16:29 McLeopold: i don't have much time, i'm only online for email and stuff 2011-07-02T13:16:32 can you look at the log just before you joined 2011-07-02T13:16:39 food spawning issues 2011-07-02T13:16:50 I'm opening a github issues right now 2011-07-02T13:17:01 ok 2011-07-02T13:20:50 amstan: https://github.com/aichallenge/aichallenge/issues/206 2011-07-02T13:21:19 @later tell janzert: please take a look at this and let us know if you have ideas to solve it https://github.com/aichallenge/aichallenge/issues/206 2011-07-02T13:21:19 McLeopold: Yep. 2011-07-02T13:21:36 sir_macelon: https://github.com/aichallenge/aichallenge/issues/206 please added games that show off this issue 2011-07-02T13:21:41 *** McLeopold has left #aichallenge 2011-07-02T13:22:35 McLeopold: ok, i'll do 2011-07-02T13:26:44 *** pvarga has quit IRC (Quit: pvarga) 2011-07-02T13:29:48 *** pvarga has joined #aichallenge 2011-07-02T13:34:24 *** pvarga has quit IRC (Ping timeout: 255 seconds) 2011-07-02T13:34:39 *** VilleH has joined #aichallenge 2011-07-02T13:36:34 Hello, anyone have a clue why I am getting "Test Error: compiled, but failed test cases" after upload on profile page? I'm using C# and bot compiles (in microsoft vs and also with mono) and runs fine on local computer. 2011-07-02T13:37:09 Also I don't get any emails from the server. 2011-07-02T13:40:10 *** qacek has joined #aichallenge 2011-07-02T13:41:05 *** jmcarthur has quit IRC (Quit: WeeChat 0.3.5) 2011-07-02T13:44:51 Hi, at one point I heard that an ant could still "see" for 1 turn after it died. I see the game spec has been updated on the wiki and I don't see any mention of this. Has this idea been axed? 2011-07-02T13:47:28 definitely not true now 2011-07-02T13:49:35 VilleH: its still beta, they dont yet send emails, and try running your bot with test_bot.sh 2011-07-02T13:50:14 and make sure you have the latest tools from github 2011-07-02T13:57:50 mcstar: Just downloaded the tools today, so I think they should be latest. 2011-07-02T13:59:02 *** pvarga has joined #aichallenge 2011-07-02T14:04:06 *** pcavs has quit IRC (Ping timeout: 255 seconds) 2011-07-02T14:06:50 Also I don't see any errors when running test_bot.cmd 2011-07-02T14:07:58 *** choas