2011-06-19T00:00:39 *** Accoun has quit IRC (Read error: Connection reset by peer) 2011-06-19T00:02:41 *** Accoun has joined #aichallenge 2011-06-19T00:06:10 *** amstan has quit IRC (Read error: Operation timed out) 2011-06-19T00:52:20 *** Zaphus has joined #aichallenge 2011-06-19T01:09:54 *** bhasker has quit IRC (Quit: bhasker) 2011-06-19T01:16:02 *** sigh has joined #aichallenge 2011-06-19T02:19:40 *** sigh has quit IRC (Quit: This computer has gone to sleep) 2011-06-19T02:27:04 *** Palmik has joined #aichallenge 2011-06-19T02:28:31 invasion of the java bots 2011-06-19T02:28:49 _flag: was that you who had someone on a russian forum who couldn't get the c++ bot working in visual c++ 2011-06-19T02:29:20 <_flag> antimatroid: I have no idea what you're talking about, so probably not 2011-06-19T02:29:36 antimatroid: I think that was UncleVasya 2011-06-19T02:29:53 hmm okay, i'll ask him next time he's here 2011-06-19T02:34:55 *** stocha has joined #aichallenge 2011-06-19T02:35:21 a good example of bad bot behavior for contest health : http://aichallengebeta.hypertriangle.com/visualizer.php?game=27891&user=206 2011-06-19T02:36:26 *** antimatroid1 has joined #aichallenge 2011-06-19T02:36:54 hehe, they're accreting 2011-06-19T02:37:09 *** antimatroid has quit IRC (Read error: No route to host) 2011-06-19T02:37:46 anyway my two trys at escaping the ennemy failed miserably 2011-06-19T02:38:21 i see to possible reasons : two few ants, the poor thing don't understand the game (neither do i for that matter). Most probably i went the wrong way. I'll try to upload my other way 2011-06-19T02:38:35 I think it's good to have the coward bots, and see how they do. 2011-06-19T02:48:55 stocha: looking at that game that you claim is bad - you could easily have taken him outby coming around the other side of him as well - if your bot can't figure it out, all the better for his bot - nothing wrong with that. If I only had 2 ants left I'd want them to survive as long as possible, dont see why that is 'bad for contest health' - am I meant to suicide ? 2011-06-19T02:50:07 note, when I say you could easily take him out - remember that I'm just crapping on and havent actually checked to see if it's possible to take him out from the other side - just that there were spaces there you werent using :-) 2011-06-19T02:59:42 Zaphus: the bad behavior comes from that all three winning bots are cowards. 2011-06-19T02:59:57 so the game will tend to extend as long as it can. Even though it's clearly cut. 2011-06-19T03:00:34 the cowardice of stocha bot is not obvious, in that the flee code is a bit broken from two side :p 2011-06-19T03:00:57 but it shows when it's afraid of the lonely two remaining ants, when having hordes just at the door :) 2011-06-19T03:02:02 i tryed to update it with a litle more obvious cowardice fleeing (although it's not yet the whole army flee stuff) 2011-06-19T03:03:33 *** stocha has quit IRC (Quit: Page closed) 2011-06-19T03:08:41 I do like the fractalesque pattern it creates in the end :-) 2011-06-19T03:09:21 if 90% of the bots are cowards, then the aggressive bot will rule the day by a large margin :-) 2011-06-19T03:09:47 hi, who has the starter package on C? 2011-06-19T03:13:19 *** carlop has quit IRC (Ping timeout: 240 seconds) 2011-06-19T03:21:40 zeLark: why you no use c++? 2011-06-19T03:24:19 antimatroid1: i'm using C++, but i prefer C 2011-06-19T03:25:26 :( 2011-06-19T03:25:49 what? 2011-06-19T03:25:53 zeLark: how long ago did you download the c++ bot? 2011-06-19T03:26:00 i made some changed a week or two ago 2011-06-19T03:26:15 zeLark: ignore me, i'm just trying to convince as many people as possible to use my starter bot :P 2011-06-19T03:26:19 2 day ago 2011-06-19T03:26:33 you're all good then 2011-06-19T03:29:21 C++? 2011-06-19T03:29:39 antimatroid1: what changes did you make to it ? 2011-06-19T03:29:44 zeLark: you can find the C starter bot at https://github.com/flagcapper/GoogleAI-Ants-Starter-Pack 2011-06-19T03:29:58 Zaphus: got rid of handling the 'r' lines and fixed the timer to work on windows 2011-06-19T03:30:21 Zaphus: I can tell you what to change for the 'r' lines if you still have that 2011-06-19T03:30:24 ok, tnx 2011-06-19T03:30:47 anyone else trying out FF5 yet? 2011-06-19T03:30:59 *** carlop has joined #aichallenge 2011-06-19T03:31:00 chrome <3 2011-06-19T03:31:37 too many firefox addons I like to give it up for me :) 2011-06-19T03:31:55 i never got dragged in, i couldn't go to tab trees and back though 2011-06-19T03:31:58 so i just haven't 2011-06-19T03:33:04 Zaphus: I think I copied the 'r' changes, I ignore the 'r' anyway and it still seems to work :-) 2011-06-19T03:33:18 Zaphus: let me check a couple of things with you 2011-06-19T03:33:49 in the >> function in state you need to get rid of the 'r' line if statement, in State::reset you need to clear the food vector and in Square::reset you need to make isFood = 0/false 2011-06-19T03:34:54 antimatroid1: yep, got all that 2011-06-19T03:35:35 antimatroid1: do you know SQL ? 2011-06-19T03:35:42 not a single bit :P 2011-06-19T03:35:45 lol 2011-06-19T03:36:04 i have a very lacking background in cs/programming 2011-06-19T03:36:20 self-taught? 2011-06-19T03:36:40 i technically did a cs major through science degree at utas, but "computing" at utas SUCKS 2011-06-19T03:36:56 the only useful class was an advanced algorithms one in the winter 2011-06-19T03:37:03 taken once every 2 years 2011-06-19T03:37:24 and i did the bare minimum to complete the major :P 2011-06-19T03:37:27 antimatroid1: what do you think of ending a game early if one player can reach >95% of squares first for more than say 100 turns, at which point the end game bonus is split between the remaining players? 2011-06-19T03:37:52 it would help incentivize bots to make the final kill as quickly as possible 2011-06-19T03:38:03 janzert: yeah i'm all for something like that 2011-06-19T03:38:11 and end games where a bot has blocked itself into a small area 2011-06-19T03:38:14 anything that enables a larger max turns for legitimately long games 2011-06-19T03:38:44 larger in the scheme of what's feasible, not necessarily larger than current 2011-06-19T03:39:03 I'll have to try and remember to bring it up with mcleopold when I see him next 2011-06-19T03:39:08 yeah 2011-06-19T03:39:09 *** stocha has joined #aichallenge 2011-06-19T03:39:27 i found the C starter bot hard to wrap around my own bot 2011-06-19T03:40:01 so much so, that it tryed to rewrite it. Don't forget to flush the stream :p I spent 3 hours trying to figure that hehe 2011-06-19T03:40:53 berak pasted a pastebin sample (with no flush) that could probably be a good start if you don't like the current c stater bot also. 2011-06-19T03:40:56 janzert: what does "95% of squares first for more than 100 turns" mean ? is it first, or is it for 100 turns ? 2011-06-19T03:41:04 (it is in yesterday log i think) 2011-06-19T03:41:17 I was also sort of thinking about a dynamic game limit, say adjust the turn limit once per day such that if 95% of games ended earlier than the current limit it would lower to the limit that number else if >10% of games hit the current limit it would raise the limit (up to whatever we can handle feasibly) 2011-06-19T03:41:25 Zaphus: the squares in which you can reach fastest over other bots covers 95% of the grid for 100 consecutive turns 2011-06-19T03:41:38 Zaphus: what antimatroid1 said :) 2011-06-19T03:41:57 antimatroid1: but who determines how fast you "could" get there ? does that mean the engine does pathfinding for all ants every turn ? 2011-06-19T03:42:06 *** nux67 has joined #aichallenge 2011-06-19T03:42:16 Zaphus: i have an engine that can do that easilly :) 2011-06-19T03:42:26 Zaphus: shortest path to each square 2011-06-19T03:42:27 the engine basically does a floodfill from each ant 2011-06-19T03:42:31 yuk 2011-06-19T03:42:32 simple to paint the entire grid in one pass 2011-06-19T03:42:43 *** stocha has quit IRC (Client Quit) 2011-06-19T03:42:56 a 'mercy' rule like 100x the number of ants of the other guys is much simpler to undertand 2011-06-19T03:43:34 *understand 2011-06-19T03:43:56 I'd be fine with something like one player has >90% of all ants for 100 turns too 2011-06-19T03:44:20 has to be a rule that doesnt mean the game ends 20 turns in because one bot was better at food finding ! 2011-06-19T03:44:25 i think that would be better 2011-06-19T03:44:35 easier for people to understand and shortens games that stupidly go for ages 2011-06-19T03:45:04 the only catch is that your bot has no way of knowing the situation - since it can't tell how many ants are out there, which means you can't react to it 2011-06-19T03:45:15 but thats a lesser problem 2011-06-19T03:45:19 yeah percentage of total ants seems better all around (although maybe not quite as accurate the difference is almost certainly in the noise) 2011-06-19T03:45:40 Zaphus: sure, but it encourages you to be attacking if you have a lot of bots 2011-06-19T03:45:45 lot of ants* 2011-06-19T03:46:15 so.. would it split the points after that ? 2011-06-19T03:46:38 split the bonus points you mean? yes, that's my thought 2011-06-19T03:46:39 yes 2011-06-19T03:46:52 I think can be a good thing 2011-06-19T03:46:53 otherwise you would just try to have lots of ants without taking out the last few 2011-06-19T03:46:57 and is what causes the incentive to finish off the game 2011-06-19T03:47:24 better than automatically remove each possible perfect blocking location from the map generator 2011-06-19T03:47:40 i still want to do that 2011-06-19T03:47:45 :p 2011-06-19T03:47:55 yeah, I think thats good: so this game : http://aichallengebeta.hypertriangle.com/visualizer.php?game=27933&user=199 would have ended around turn 1151, right ? 2011-06-19T03:47:56 ok, I'm pretty convinced to do it so long as McLeopold doesn't raise a large objection 2011-06-19T03:49:15 whereas this one: http://aichallengebeta.hypertriangle.com/visualizer.php?game=27927&user=16 would have gone to more or less the same place as it finished 2011-06-19T03:49:16 yes 2011-06-19T03:49:36 should the 100 turn counter reset if a player is eliminated ? (give you a chance to go after the next one) 2011-06-19T03:49:46 although looking at the first one at least, I think 90% may be to low 2011-06-19T03:50:26 too low ? 600 vs 60 seems a pretty convincing place to be 2011-06-19T03:51:09 100 turns might be too short 2011-06-19T03:51:18 give them time to actually wipe them out 2011-06-19T03:51:25 it's pretty convincing yes, but not really close to a situation where the 10% are simply huddled in a corner 2011-06-19T03:53:08 but why is the 'huddle' the problem, surely we're just trying to shorten games that have reached a conclusion ? or is this a case of creating a rule to prevent a particular type of bot behavior, rather than a general rule for everyone 2011-06-19T03:53:43 *** carlop has quit IRC (Read error: Connection reset by peer) 2011-06-19T03:53:52 we want to shorten games the have reached a conclusion *and* incentivize the leading player to finish off the game naturally 2011-06-19T03:54:07 we have to make sure the leading player really has that chance 2011-06-19T03:54:14 but we're also punishing the leading player to some degree 2011-06-19T03:54:18 hence ^^ yeah 2011-06-19T03:54:44 you're punishing them significantly *if* the game was to otherwise end, but if it's going to play out as a draw then it's the same result but saving CPU :-) 2011-06-19T03:55:04 or, hand out the points as a percentage (90% of the ants gets 90% of the bonus) 2011-06-19T03:55:42 actually, thats pretty fair... and wouldnt hurt the leader 2011-06-19T03:55:53 not much incentive to finish off naturally then 2011-06-19T03:56:10 there isnt anyway, since I'll win regardless 2011-06-19T03:56:22 so looking at that first game example I'm thinking maybe 95% and 200 moves? 2011-06-19T03:56:41 somewhat less chance you'll win regardless 2011-06-19T03:57:01 i guess the question is, what is our realistic kind of max turns? 2011-06-19T03:57:05 janzert: it would have ended about 50 turns earlier with that rule 2011-06-19T03:57:15 i.e. turn 1390 or so 2011-06-19T03:57:40 yes it still would have ended early 2011-06-19T03:57:58 but your bot also wasn't trying to kill it off very hard ;) 2011-06-19T03:58:33 janzert: no, never claimed it was, but no real incentive to bother either - it would have been happy if the game had been called off 1000 turns ago 2011-06-19T03:58:55 *** carlop has joined #aichallenge 2011-06-19T03:58:59 (but not 1100 turns ago, then it would have lost!) 2011-06-19T04:00:22 If you ignore 1st place for now, these rules have a dramatic effect on 2nd place and below, which do change rankings. blocking is even more valuable if the game will be called off and you get a share of the prize, you'll do better than the bots that died earlier 2011-06-19T04:01:15 and I guess I don't really care too much about the specific numbers just so we try and balance so not too many games where the leader is making clear progess ends up triggering it 2011-06-19T04:01:58 Zaphus: yep, certainly more incentive to hold on even when you're about gone 2011-06-19T04:03:30 antimatroid1: any thought on dynamically adjusting the turn limit as I mentioned above? 2011-06-19T04:04:13 which comment? 2011-06-19T04:04:14 i.e. once a day look at game lengths and if 95% are lower than current drop to that, if >10% are larger bump it up 2011-06-19T04:04:45 yeah something like that could work 2011-06-19T04:04:48 with a maximum it can go up to though 2011-06-19T04:05:15 basically a hard maximum of whatever we can support with the resources we end up with 2011-06-19T04:05:48 it seems like it would help utilize resources with the most efficiency 2011-06-19T04:06:15 it also keeps developers from getting settled into a specific length ;) 2011-06-19T04:07:14 yep 2011-06-19T04:07:27 also I just noticed last night that the game specs specify that values in the pregame are integers :( 2011-06-19T04:07:45 you mean the seed thing? 2011-06-19T04:07:53 no 2011-06-19T04:08:43 rather than saying a value can be any free form string (excluding newlines of course) and then saying what the type is for each one we currently send 2011-06-19T04:09:04 that was certainly what I had envisioned at least 2011-06-19T04:09:17 do you want to change it to what you think works better? 2011-06-19T04:09:21 i don't really mind 2011-06-19T04:09:57 the bad thing is at least the D starter bot is already coded for the way the spec is written, I'm not sure how many others are as well 2011-06-19T04:09:58 i have an actual graph constructed now for path finding 2011-06-19T04:10:24 you mean reads type and value? 2011-06-19T04:10:38 yeah, it always converts the value to an integer 2011-06-19T04:10:40 i just read type and if it's unknown read the rest of the line to a junk string 2011-06-19T04:10:52 problem solved 2011-06-19T04:11:07 right, I think that is the way to do it 2011-06-19T04:11:13 *** locutus2 has joined #aichallenge 2011-06-19T04:11:24 but the current spec implies that the other is fine 2011-06-19T04:12:00 frontier pointed out my code actually makes some kind of sense, i have "getline(is, junk);" :P 2011-06-19T04:12:02 a suggestion, for help newby understanding the battle resolution method can be made immage where calculation are made, eg: 1. how many enemy each ant see ?, 2. max (how many enemy each enemy ant see ?) over every ant, 3. declare who should die (num 1 <= num 2) 2011-06-19T04:12:11 this for few battle situation 2011-06-19T04:12:12 heh 2011-06-19T04:13:01 I think this step to step calculation can help 2011-06-19T04:13:10 carlop: it's not how many enemy ants they see, it's how many enemy ants are within the attack radius 2011-06-19T04:13:32 number of ants within reach ;) 2011-06-19T04:13:37 yes, change "see" to "inside attack radius" 2011-06-19T04:13:52 janzert: ? 2011-06-19T04:14:02 i seem to be a terrible judge of helping people understand 2011-06-19T04:14:04 *** qacek has quit IRC (Quit: WeeChat 0.3.5) 2011-06-19T04:14:22 the attack radius is how for the ant can reach or maybe kick :) 2011-06-19T04:14:45 the food spawn radius is how far the ant can do fine work with it's pincers 2011-06-19T04:15:04 takes a lot of small intricate work to get another ant out of food 2011-06-19T04:15:19 can punch and kick further of course 2011-06-19T04:15:23 and see even further 2011-06-19T04:15:32 story time :) 2011-06-19T04:15:54 janzert: what about spit venom or whatever 2011-06-19T04:15:58 poison ants 2011-06-19T04:16:09 you only spit one direction though 2011-06-19T04:16:19 maybe they're canadian ants? 2011-06-19T04:16:24 :) 2011-06-19T04:17:07 i wish amstan was here 2011-06-19T04:18:26 janzert: was AFK. I'm not a fan of dynamic game lengths. especially since it could cause them to go up and then you'd bring in another rule to stop them going up further, and then a rule to try to bring them back down. 2011-06-19T04:19:14 there'd be a limit on how high it can go 2011-06-19T04:19:41 which is basically what we would set the turn limit too if we don't use a dynamic limit 2011-06-19T04:20:08 why not make a simpler rule that can always be watched, like the 90% one, or the one about how much of the map you control - those I can understand and predict, a dynamic limit means the same match on day 1 ends on a different turn to on day 2, even if everything plays out the same 2011-06-19T04:20:08 and also where we would start it 2011-06-19T04:20:43 it would be in addition to the early end rule 2011-06-19T04:21:32 it allows for better resource usage while only cutting off the 5% of games that would go longer 2011-06-19T04:22:36 also I like the idea that it makes sure the bots are used to working with a variable turn limit 2011-06-19T04:23:13 we've certainly seen that anything that stays the same people will assume that it will stay the same, even if we say that it won't 2011-06-19T04:24:02 seems to be making things complicated when it doesnt need to. I dont mind variable turn lengths between 500 and 1000 turns (or whatever) - but dynamic per day isnt the same thing, it's a set length that slowly moves 2011-06-19T04:24:43 we could update it after each game ;) 2011-06-19T04:24:47 maybe better is having different game lenght on different map (or number of player) 2011-06-19T04:24:59 but why ? it's just adding complexity where it isnt needed 2011-06-19T04:25:19 the need for it is to maximize resource efficiency 2011-06-19T04:25:26 it's a meta need 2011-06-19T04:25:26 and janzert rule or similar can set what's the "optimal" lenght for each map 2011-06-19T04:26:06 carlop: hmm, doing the calculation per map might work really well 2011-06-19T04:26:37 Zaphus: it would be fixed for the final contest 2011-06-19T04:26:43 and all rankings are reset when that starts 2011-06-19T04:26:50 so it's not about letting the games play out, or stopping blocking or coward bots then, its just about resources ? in which case make it wall-clock time as the limit, not turns 2011-06-19T04:27:27 which will hurt the smart bots more, and I'm guessing you're going to say thats unfair :-) 2011-06-19T04:28:12 or go the chess-clock route, each bot gets 2 minutes of wall clock for the entire game, on top of the per-turn limit :) 2011-06-19T04:28:13 not so much unfair as contrary to a primary goal 2011-06-19T04:28:14 or a cowar bot can waste all his cpu time before to say "don't move" 2011-06-19T04:28:37 more complex time controls were very seriously considered 2011-06-19T04:28:46 (I'm playing devils advocate here by the way, someone needs to) 2011-06-19T04:28:57 and maybe we will for the next contest 2011-06-19T04:29:27 so the primary goal is what ? to let the matches reach a conclusion ? 2011-06-19T04:30:06 to efficiently use the resources available to rank the bots, with the primary goal of the contest being to have interesting ai? 2011-06-19T04:30:14 "ai"* 2011-06-19T04:30:31 the primary goal I was refering to about not hurting smart bots was "encourage the development of the smartest program possible" 2011-06-19T04:30:34 right now, a 10-bot match between slow bots (that use all their CPU) and a 2000 turn limit will take up to 5.5 CPU hours, which is a long long time 2011-06-19T04:30:45 i.e. Artificial Intelligence contest 2011-06-19T04:31:01 yes, and that's too longf 2011-06-19T04:31:22 actually, I think that was at 1 second per move, I did the calculation a while ago, but you get the idea 2011-06-19T04:31:23 right, I don't think there is anyway we'll be able to use 2000 turns 2011-06-19T04:32:10 so, why not set it at 1000 then ? half the cost. 2011-06-19T04:32:38 I'm not even completely sure 1000 will be feasible ;) 2011-06-19T04:32:55 although too much below that and something else is going to have to give 2011-06-19T04:33:01 now you're getting to the root of the problem.... 500 turns isnt enough to determine the winner in a strong match 2011-06-19T04:33:21 are you sure? 2011-06-19T04:33:40 a good number of strong matches seem to end around 300 or so 2011-06-19T04:34:14 I'm not really not sure one way or the other 2011-06-19T04:34:23 s/not really/really 2011-06-19T04:34:32 true, but those arent the ones people will complain about :-) 2011-06-19T04:35:12 with a map of 100x100, it takes 200 turns to walk the full length diagonal - so 300 turns would *seem* to be barely enough 2011-06-19T04:35:14 if 98% of the strong matches are naturally decided I'll be quite happy 2011-06-19T04:35:19 *** stocha has joined #aichallenge 2011-06-19T04:35:30 http://aichallengebeta.hypertriangle.com/visualizer.php?game=27933&user=78 turn 500, would you be prepared to call the result at that point ? 2011-06-19T04:35:31 it seems this version will chain produce those silly games : http://aichallengebeta.hypertriangle.com/visualizer.php?game=27976&user=107 2011-06-19T04:36:20 Zaphus: yes, it's not hard to find examples that aren't decided by 500 2011-06-19T04:36:32 I'd really like to see a histogram though 2011-06-19T04:37:07 janzert: agreed... on the other hand, it's a Beta server - set it to 500 turns for a couple of days and see what happens 2011-06-19T04:37:24 I've got the SQL dump from a few days ago, but can't work out how to find the game results 2011-06-19T04:37:37 actually if that game had ended at 500 I'm not sure the order would have changed (maybe 2nd and 3rd would swap) 2011-06-19T04:39:03 who are the people who want to keep the possibility of block off regions? 2011-06-19T04:39:19 i am pretty keen to paint maps with 3x3 blocks to eliminate such possibilities 2011-06-19T04:39:38 carlop: ^? 2011-06-19T04:39:43 here's why you don't want to decide at 500: http://aichallengebeta.hypertriangle.com/visualizer.php?game=27968&user=133 my blocktest would have won :-) 2011-06-19T04:39:47 I don't really care if small regions can be blocked off 2011-06-19T04:40:03 and it may provide 'prettier' or more natural looking maps 2011-06-19T04:40:13 janzert: it's a pretty cheap way of preventing someone from getting the food bonus 2011-06-19T04:40:19 I like blocking off, I see no problem. if water was kept "thin" it would be harder to do as well 2011-06-19T04:40:47 Zaphus: i found carving land out of water much faster than the reverse 2011-06-19T04:40:59 while checking the map was connected etc. 2011-06-19T04:41:12 yeah, more like maze generation that way 2011-06-19T04:42:21 yeah we may need to take the map size into account. 2011-06-19T04:42:32 for a 32x32 map, 2000 seems a bit too much :p 2011-06-19T04:42:40 although for 256x256 it might be okay. 2011-06-19T04:43:48 hmm - looking at my blocktest, I think that shorter games will favor blocking, since it takes a while to remove a block (one that is half-hearted but still a little effective) - you need to pick away at the block one or two per turn, takes time 2011-06-19T04:43:48 antimatroid1: don't take out blockng. At least until the beta illustrate why you should. it didn't happen yet. 2011-06-19T04:44:02 especially if it means hard map generation rule to avoid it. 2011-06-19T04:44:38 stocha: i would just be carving the maps with 3x3 blocks 2011-06-19T04:44:45 Zaphus: how does your blocks endure against the top 5 bots ? 2011-06-19T04:44:52 i am going to at least test carving with nxn blocks next week 2011-06-19T04:45:06 antimatroid1: yeah. it would just be another constraint on map generation. Wich is problematic enough without any constraints 2011-06-19T04:45:17 *** stocha has quit IRC (Quit: Page closed) 2011-06-19T04:45:24 http://aichallengebeta.hypertriangle.com/visualizer.php?game=27851&user=133 2011-06-19T04:45:57 actually ignore that, there was nowhere to block so it just played an average game instead 2011-06-19T04:46:49 my blocker doesnt start blocking until turn 200, and against a top bot it doesnt tend to live that long :-) 2011-06-19T04:48:16 it also doesnt actually block, it just makes diagonal walls, but those have blocking properties that need to be dealt with so it kills dumb bots 2011-06-19T04:54:04 anyone have that link to the block patterns withthe current rules ? 2011-06-19T04:54:24 *** carlop has quit IRC (Read error: Connection reset by peer) 2011-06-19T04:59:42 *** carlop has joined #aichallenge 2011-06-19T05:02:44 i think this are the current perfect block allowed: http://pastebin.com/E7kxhdau 2011-06-19T05:03:09 and for attack2=5 vertical and orizontal line work better than diagonal line 2011-06-19T05:03:36 b = blocker ? 2011-06-19T05:03:43 yep 2011-06-19T05:06:04 ok, if I find time tonight I might do an updated Zaphus_Blocktest to try and do some "real" blocks if it can find them. is there a map with a known blockable place I can test on ? 2011-06-19T05:06:54 *** mcstar has joined #aichallenge 2011-06-19T05:13:31 antimatroid1: damn my visual debugger is tooo fucking slooooow. spending most of the time parsing command strings with regular expressions :p 2011-06-19T05:13:44 nawwwww :( 2011-06-19T05:14:26 Zaphus: symmetric_6 2011-06-19T05:14:59 row 40, col 85 should be blockable with 3 1-corridor 2011-06-19T05:16:14 mleise: i set up my development environment: hunchentoot webserver(in lisp) with cl-who which translates s-exps to html, and with parenscript, which translates lisp to javascript 2011-06-19T05:16:37 i wrote my first html5 canvas yesterday night 2011-06-19T05:16:56 its a rectangle that fill black when clicked on it :D 2011-06-19T05:17:02 s 2011-06-19T05:17:50 antimatroid1: for design map against blocking wiht 3x3 block don't know, maybe could be a good thing have small corridor that can be blocked, giving to one side positional advantage (eg. if you fight hard for conquer certain part of the map, you can easily defend), the only problem that i can see is if perfect blocking can be possible from all side 2011-06-19T05:19:10 mleise: just a heads up and not sure if it's something with my setup, but the visualizer in firefox 5 for me has dark grey squares over the enabled buttons 2011-06-19T05:19:44 janzert: firefox 5, if it exists must be alpha or beta software 2011-06-19T05:20:21 official release is monday, final binaries are already on ftp 2011-06-19T05:20:34 barring any last minute emergencies of course 2011-06-19T05:21:16 err, official release is 21st, Tuesday 2011-06-19T05:22:55 if you do want to check it out early, you can find it at ftp://ftp.mozilla.org/pub/mozilla.org/firefox/releases/5.0/ 2011-06-19T05:24:10 janzert: let me take a look at what my fellow Gentoo repository ppl say about its stability. :p 2011-06-19T05:24:55 cause my current repo snapshot still has ff 4 marked as experimental 2011-06-19T05:25:16 :) 2011-06-19T05:25:35 it's certainly not a big update over FF4 2011-06-19T05:25:42 they jumped really quick from 4 to 5, i don't know why. there cant me many new features 2011-06-19T05:26:11 they're switching to chrome style releases 2011-06-19T05:26:26 but probably a lot of bugs, due to the rush, like with my perfectly bug free visualizer that even runs on IE 9, hehe :-( 2011-06-19T05:26:27 timed, 4 times a year no matter what is in 2011-06-19T05:26:59 so by 2020 they have version 40? 2011-06-19T05:27:18 by next year you're not suppose to be paying attention to the version number ;) 2011-06-19T05:27:29 and we have chrome 12? now 2011-06-19T05:28:39 yeah right, but it is chrome 13.0.124.0 ... so you look at it differently 2011-06-19T05:29:42 once we go through a few cycles of the FF number bumping every 3 months I'm sure people won't be paying attention to it either 2011-06-19T05:29:45 oh actually Gentoo has revoked the stable tag from Chromium 13 and I would downgrade to 12 if I updated my system now 2011-06-19T05:30:02 *** berak has joined #aichallenge 2011-06-19T05:30:10 yeah, 12 is the official stable release currently 2011-06-19T05:30:38 my mother used ff 4 on windows and it was eating to much of her 1 GB ram, i told her to use Opera. They were writing browsers for handhelds for years 2011-06-19T05:31:00 :) 2011-06-19T05:31:14 they've had a good history of low memory use for sure 2011-06-19T05:32:00 although they switch paradigm now to compete with the other browsers. at least with javascript precompilation. 2011-06-19T05:32:06 *** antimatroid1 has quit IRC (Quit: Leaving.) 2011-06-19T05:32:21 *** antimatroid has joined #aichallenge 2011-06-19T05:33:00 ok I'll check out the FireFox 5 release. maybe I can work around the bug or tell FireFox that timed releases suck 2011-06-19T05:35:23 janzert: Chrome had some bad visual artifacts with enabled hardware acceleration. I hope its not that type of bug. That may be difficult to solve on either end :p 2011-06-19T05:35:56 hopefully not 2011-06-19T05:36:39 hmm, no FireBug support yet? 2011-06-19T05:37:06 have to go with the 1.8 beta still :( 2011-06-19T05:37:37 uh... what happened to my buttons? 2011-06-19T05:37:47 they are all blurry and grey 2011-06-19T05:37:53 ahh, so it's not just me then :/ 2011-06-19T05:38:08 the disabled ones look ok ;) 2011-06-19T05:40:03 janzert: It is an FF5 bug. Look at http://philip.html5.org/demos/canvas/shadows/various.html 2011-06-19T05:40:35 Scroll down to 'Image'. In other browsers you see dice with a blue shadow. In FF5 the whole image is blurred. 2011-06-19T05:40:40 or rather... blue 2011-06-19T05:40:44 * janzert nods 2011-06-19T05:40:56 ugh :( 2011-06-19T05:41:43 * antimatroid doesn't understand why they don't standardise crap so web development isn't such a pita? 2011-06-19T05:42:17 the problem here isn't standardization so much as implementation 2011-06-19T05:43:04 html5 is complete due to 2022 2011-06-19T05:46:51 There is a hole in the standard: For the buttons I use one big image and render only parts of it. Most browsers add the shadow after things are drawn on screen so that 'cut out' strategy works. 2011-06-19T05:46:52 *** stocha has joined #aichallenge 2011-06-19T05:47:23 In Safari you can see the shadow of adjacent parts of the big image 2011-06-19T05:47:36 The standard isn't clear here. 2011-06-19T05:49:46 cant ou disable shadow? and then add a full transparent image on top of it, which has shadow 2011-06-19T05:49:52 ou->you 2011-06-19T05:53:22 *** stocha has quit IRC (Ping timeout: 252 seconds) 2011-06-19T05:53:29 stocha: interesting match, I like the way you feed free points to green continuously :-/ http://aichallengebeta.hypertriangle.com/visualizer.php?game=28001&user=78 2011-06-19T05:54:12 *** stocha has joined #aichallenge 2011-06-19T05:54:22 at last a game, that a bot clearly loose because of the scoring system : http://aichallengebeta.hypertriangle.com/visualizer.php?game=27994&user=107 2011-06-19T05:55:06 (stocha escaper code, it end up two in order of elimination but 4 on score :) It outlives bots for hundreds of turn without earning a point :) ) 2011-06-19T05:56:00 turn 200 : only two player left A and B. after 650 turn, B end up in 4th position :p 2011-06-19T05:56:18 so the scoring system worked in that particular case. 2011-06-19T05:56:21 mcstar: I wont work around a regression bug in a browser that hasn't been released yet. 2011-06-19T05:56:49 *** Redgis has joined #aichallenge 2011-06-19T05:56:57 stocha: am I looking at the wrong game ? I see 4 alive at turn 200 2011-06-19T05:57:04 mleise: sry, i didnt know that, it was just an idea 2011-06-19T05:57:05 I'm sure the mozilla team will solve this problem very soon. 2011-06-19T05:57:32 *** Redgis has left #aichallenge 2011-06-19T05:58:11 mcstar: yeah, it wan't even a bad idea since I think IE and Chrome don't support shadows on images at all. 2011-06-19T05:58:20 stocha: in fact there were still 3 alive with a couple of turns to go. I must be misunderstanding 2011-06-19T06:00:42 *** stocha has quit IRC (Ping timeout: 252 seconds) 2011-06-19T06:06:34 *** zeLark has quit IRC (Ping timeout: 252 seconds) 2011-06-19T06:13:39 lol, i typed emacs instead of clementine(music player) 2011-06-19T06:20:48 love the swarming nature of the winner of this fight: http://aichallengebeta.hypertriangle.com/visualizer.php?game=28009&user=78 looks cool sped up a bit 2011-06-19T06:21:09 *** smiley1983 has joined #aichallenge 2011-06-19T06:23:35 antimatroid: you said your graph bot was working. I'm glad to hear that :) 2011-06-19T06:23:48 smiley1983: :) 2011-06-19T06:23:56 it's pretty simple to add in the end 2011-06-19T06:27:42 did you end up going with vector > grid or map grid*** or some other option? or is it secret? :) 2011-06-19T06:29:04 ah, I think you already answered that, just catching up with the log now 2011-06-19T06:29:27 i made GNode my normal Square struct 2011-06-19T06:29:36 and Grid has vector > 2011-06-19T06:29:55 and the squares have a vector neighbours 2011-06-19T06:30:29 they also have a vector nDirections 2011-06-19T06:30:33 couldn't avoid that 2011-06-19T06:30:51 well atm it's vector, but i'm going to change that soon 2011-06-19T06:31:55 yes, that makes sense, precalculating the neighbors to avoid all that repetition of wrapping functions :) 2011-06-19T06:32:06 yep 2011-06-19T06:32:11 and not looking on water squares 2011-06-19T06:32:19 yes, indeed 2011-06-19T06:32:22 can just ignore the existence of water when pathfinding etc. 2011-06-19T06:32:51 someone before said it was redundant, i think it'll be worthwhile 2011-06-19T06:34:34 I don't really know the relative costs of looking up a value in a larger table vs recalculating that stuff, but I would have thought it would save some cycles in later turns 2011-06-19T06:35:05 plus you can do all the allocation and calculation in the first turn, and then update it as as you find water 2011-06-19T06:35:52 exactly 2011-06-19T06:36:21 I think the occupied battle system is clear enough, especially with the code provided to explain 2011-06-19T06:36:34 has it been verified that it solves the blocking issue? 2011-06-19T06:38:05 mcleopold has some tester that he ran on it 2011-06-19T06:38:11 i don't know how exhaustive it is 2011-06-19T06:38:15 it's certainly much better 2011-06-19T06:39:40 mleise: looks like the FF5 problem is probably this bug? https://bugzilla.mozilla.org/show_bug.cgi?id=662898 2011-06-19T06:40:42 janzert: no, that is yet another bug 2011-06-19T06:41:12 but the guy who rewrote the canvas code will suffer next week *evilgrin* 2011-06-19T06:41:19 oh really 2011-06-19T06:41:35 heh 2011-06-19T06:41:49 oh someone says 'toggle shadow off', then it is the same bug of course 2011-06-19T06:42:04 yes, that's what made me think it was 2011-06-19T06:42:29 "Remove unnecessary full context save/restore" <- so much for 'unnecessary', lol 2011-06-19T06:42:56