2011-12-05T00:00:30 *** Antimony has joined #aichallenge 2011-12-05T00:04:19 *** pairofdice has joined #aichallenge 2011-12-05T00:07:26 *** Antimony has quit IRC (Ping timeout: 244 seconds) 2011-12-05T00:07:50 *** u_ has quit IRC (Quit: u_) 2011-12-05T00:11:13 *** delt0r_ has joined #aichallenge 2011-12-05T00:11:22 *** Antimony has joined #aichallenge 2011-12-05T00:17:48 dammit, a1k0n! You blew the whistle on my combat strategy (which I'll never get around to implementing) 2011-12-05T00:17:51 :) 2011-12-05T00:18:12 and you guys didn't make a team! 2011-12-05T00:19:08 amstan: That's because I'm a rubbish coder and a1k0n had the good sense to avoid me dragging him down :) 2011-12-05T00:20:59 *** Areks has joined #aichallenge 2011-12-05T00:26:38 hi guys, who wants to help me write a bot? 2011-12-05T00:27:18 I do! 2011-12-05T00:27:29 aren't you disqualified? 2011-12-05T00:27:33 McLeopold: hey, where were you? 2011-12-05T00:27:59 I'm right here 2011-12-05T00:28:04 for the meeting 2011-12-05T00:28:18 a1k0n: oh, totally, but I still get to have fun 2011-12-05T00:28:28 why are oyu disqualified? 2011-12-05T00:28:37 amstan: totally forgot, sorry 2011-12-05T00:28:45 how'd it go? 2011-12-05T00:28:51 McLeopold: well, i included you in all the emails 2011-12-05T00:28:57 there's a log somewhere in there 2011-12-05T00:29:00 I was with the family at a christmas thing 2011-12-05T00:29:08 yeah 2011-12-05T00:29:14 good point 2011-12-05T00:29:16 but its just the begs of december still 2011-12-05T00:29:30 jstrong: cause as an admin, I get to read everyone's code 2011-12-05T00:29:52 McLeopold: you could do just that, it's an interesting exercise 2011-12-05T00:29:57 McLeopold: examining people's styles 2011-12-05T00:30:07 trying to descipher what they're doing 2011-12-05T00:30:14 amstan: I don't want to look at the code, I want to try myself first 2011-12-05T00:30:27 I just want some help with exploration 2011-12-05T00:30:30 "yes... i know what some of these statments do" 2011-12-05T00:30:53 McLeopold: do map prediction! that's the first thing i would do 2011-12-05T00:31:09 *** rajanaresh has joined #aichallenge 2011-12-05T00:32:23 a1k0n: did you do fast furier on that yet? 2011-12-05T00:32:46 *** TheLinker has quit IRC (Quit: Bye) 2011-12-05T00:33:46 amstan: if aarossig gets a bank account and corp going, that would be awesome 2011-12-05T00:34:17 although, we should use a swiss bank account...cause it sounds cool 2011-12-05T00:34:30 heh no 2011-12-05T00:34:39 it's overkill 2011-12-05T00:34:50 McLeopold: implement my hill finding strategy :) I want to know if it works 2011-12-05T00:35:03 *** Jak_o_Shadows has joined #aichallenge 2011-12-05T00:35:13 do you really think hill finding is needed? 2011-12-05T00:35:19 symmetry detection? 2011-12-05T00:35:29 hmm i bet there's an efficient dynamic programming solution to the symmetry problem tho 2011-12-05T00:36:08 mmmm, eggnog 2011-12-05T00:36:21 McLeopold: can you reply to that? it would be nice to have your support in writing 2011-12-05T00:36:28 McLeopold: the main problem is getting jeff to aproove it 2011-12-05T00:36:53 anyway 2011-12-05T00:36:56 g2g to bed 2011-12-05T00:37:00 6:30 wakeup tomorrow 2011-12-05T00:37:05 ok, cya 2011-12-05T00:37:18 a1k0n: I can do it if I see the whole map :) 2011-12-05T00:38:09 oh yeah i wanted to look at that 2011-12-05T00:38:27 a1k0n: so, for each bot, pick one ant, and give it one of 8 orientations 2011-12-05T00:38:54 ants.py already contains all possible symmetries for spawning food eh 2011-12-05T00:38:59 each other bot has a starting ant with an orientation as well 2011-12-05T00:39:13 a1k0n: I think you're making a mistake updating the distributions one ant at a time 2011-12-05T00:39:13 so, for 2 players, there are 8 possibilities 2011-12-05T00:39:31 what? why? 2011-12-05T00:39:34 for 3, I think they all have to be the same 2011-12-05T00:39:35 (bmh) 2011-12-05T00:39:58 Why don't you pick an ant, and all of the ants within (combatRadius+2), and then assign an action: push NSEW, attract, repel 2011-12-05T00:40:15 could 2011-12-05T00:40:28 but lazy 2011-12-05T00:40:45 I think it will do a better job avoiding local maxima. 2011-12-05T00:41:18 i have a version which tries those actions up front before random sampling 2011-12-05T00:41:20 a1k0n: my combat strategy is similar to yours 2011-12-05T00:41:24 with two exceptions: 2011-12-05T00:41:36 mmm, top players discussing combat strategy 2011-12-05T00:41:40 1) I don't work out the best enemy move, I play against all of them simultaneously 2011-12-05T00:41:43 which has pros and cons 2011-12-05T00:41:45 right 2011-12-05T00:41:57 cause your enemy effectively multiplies 2011-12-05T00:42:04 2) I don't use your random sampling technique, I use simulated annealing 2011-12-05T00:42:26 yeah, i have a simulated annealing branch too 2011-12-05T00:42:37 i implemented that first actually 2011-12-05T00:42:49 my rough hack to match up the scoring is that killing a potential enemy is worth 1/5th as much as my ant dying 2011-12-05T00:42:52 before fixing the bugs in combat 2011-12-05T00:42:56 random sampling being monte-carlo method of possible enemy moves? 2011-12-05T00:43:10 and possible local moves 2011-12-05T00:43:26 haha i also use 5:1 scoring 2011-12-05T00:43:31 oh and although you don't mention it I assume you are also salting your score with ancillary goals 2011-12-05T00:43:45 yeah 2011-12-05T00:43:49 a1k0n: like UCT? 2011-12-05T00:43:59 not really like UCT 2011-12-05T00:43:59 so that the combat can also make "global" moves while in combat 2011-12-05T00:44:03 no tree searches 2011-12-05T00:44:07 5-1! whoa 2011-12-05T00:44:14 i only use 2-1 (not done with my normal form yet) 2011-12-05T00:44:21 jstrong: because I play my actual ants vs their possible ants (5 possible moves for each) 2011-12-05T00:44:26 does anyone do more than 1-ply 2011-12-05T00:44:37 ah 2011-12-05T00:44:42 so you're doing a normal form type thing 2011-12-05T00:44:51 actually I vary my ratios slightly depending on circumstances 2011-12-05T00:44:56 aggressiveness, location, how many ants, etc 2011-12-05T00:45:04 nice 2011-12-05T00:45:12 rmmh: i think a couple players attemp that but not eg xathis 2011-12-05T00:45:30 the main known bug with mine is that sometimse I'll make a move where it would be devastating against some countermove 2011-12-05T00:45:34 which the enemy won't actually make 2011-12-05T00:45:35 I -think- a good normal form algorithm is as good as you need 2011-12-05T00:45:39 but sicne they don't I just lose an ant 2011-12-05T00:45:42 yeah 2011-12-05T00:45:44 *** antimatroid has joined #aichallenge 2011-12-05T00:45:45 it's a problem 2011-12-05T00:45:55 well 2011-12-05T00:46:00 why not pick best worst solution 2011-12-05T00:46:04 instead of best best :P 2011-12-05T00:46:07 especially a problem with pseudo minimax simulated annealing 2011-12-05T00:46:08 if you know what I mean 2011-12-05T00:46:15 which is the reason i sample 2011-12-05T00:46:23 I also used to solve combat globally (involving all ants that are in proximity to enemy ants) 2011-12-05T00:46:25 ok gotta read books to my kid, bbiab 2011-12-05T00:46:45 but when I fixed my scoring to be more precise (original scoring only evaluated weakness 1 and >1) 2011-12-05T00:46:46 a1k0n: reading them Norvig or MacKay? 2011-12-05T00:46:52 I need to change my loop so it's not every ant making global decisions... 2011-12-05T00:46:56 the scoring got expensive so I started splitting combat into islands 2011-12-05T00:47:01 which actually yields some interesting peripheral info 2011-12-05T00:47:07 like local strength ratios 2011-12-05T00:47:08 it's dumb when I have 5 of my ants all go towards the same piece of food 2011-12-05T00:47:09 which I didn' tknow before 2011-12-05T00:47:09 yeah islands is the way I want to go myself 2011-12-05T00:47:13 but I don't know how to do it 2011-12-05T00:47:23 I know it will involve some sort of BFS thing but haven't figured it out yet 2011-12-05T00:47:30 islands? 2011-12-05T00:47:32 someone (animatroid?) described an island method 2011-12-05T00:47:33 yeah 2011-12-05T00:47:36 bmh: haha, not quite 2011-12-05T00:47:36 i just figured it out :P 2011-12-05T00:48:13 look u pconnected component labeling and union-find 2011-12-05T00:48:14 the thing I find hard to take into account are multiple enemy ants 2011-12-05T00:48:22 basically you start with your ant, then you add all the enemy ants within (combat_radius + ~2), then repeat for the ants that you added until you stop adding ants 2011-12-05T00:48:24 have you ever been so bad at spelling that a spell checker can't even help you? 2011-12-05T00:48:29 jstrong: find the connected components of the graph of ants within 3 distance on your bfs map 2011-12-05T00:48:38 er 2 distance from your ants, or whatever 2011-12-05T00:48:57 yeah, my original combat also tried to shortcut distance cals 2011-12-05T00:49:02 but it involved the wrong set of ants 2011-12-05T00:49:12 then I made it exactly right and discovered that had other flaws 2011-12-05T00:49:27 McLeopold: once I typed nonsense into YouTube search (my fingers were shifted on the keyboard) and it gave me the correct result. But, no, in general my spelling is fine :) 2011-12-05T00:49:39 a1k0n: exactly what I was thinking 2011-12-05T00:49:43 has anyone tried making concaves around their base on maps like the maze 2011-12-05T00:49:52 if my bfs goes past 2-3 squares from nearest enemy 2011-12-05T00:50:01 concaves? 2011-12-05T00:50:09 then it should cut off because the squares after are no longer part of that battle 2011-12-05T00:50:11 keep in mind that combat involves a mix of euclidean and manhattan distances 2011-12-05T00:50:16 initiate from every ant 2011-12-05T00:50:30 perfectly defensible structures 2011-12-05T00:50:31 but OBVIOUSLY remove ants from being checked in the loop if you already know which island they are part of 2011-12-05T00:51:06 the visualizer was really critical to verifying islands were correct 2011-12-05T00:51:20 yeah good point 2011-12-05T00:52:12 * antimatroid is worried about the larger maps 2011-12-05T00:52:26 the size just got reduced again 2011-12-05T00:52:28 antimatroid: does the 25000 max area help? 2011-12-05T00:52:31 check the new specs 2011-12-05T00:52:46 dunno, that's still up to like 150x150 2011-12-05T00:52:53 but probably 2011-12-05T00:54:11 I know I'm kinda late to the bot writing party, but could someone help me with exploration? 2011-12-05T00:54:19 McLeopold 2011-12-05T00:54:21 I'd be happy to help 2011-12-05T00:54:23 McLeopold: ask...? 2011-12-05T00:54:34 I don't have the best bot ever but I can definitely at least help 2011-12-05T00:54:35 so, what's the best way to spread out? 2011-12-05T00:54:49 I'd say exploration is separate from spreading out 2011-12-05T00:54:58 I use a lastSeen system 2011-12-05T00:55:04 it feels the same 2011-12-05T00:55:06 exploration is keeping track of what you've ever seen (or seen recently for some versions) and trying to get 100% coverage 2011-12-05T00:55:17 i collect unseen border info when generating closest friend hill info 2011-12-05T00:55:18 one is concerned with unseen, the other with non-visible 2011-12-05T00:55:34 then move ants towards that border, bombing it when i collect a location 2011-12-05T00:56:03 i used to just get closest hill info across the map, but i'm just doing it to the unseen border every time i need to update, cause i want maximal time on larger maps 2011-12-05T00:56:23 mcleopold: if you have a working food gathering system you could add a new kind of "food" which is just locations seeded in a grid on the map 2011-12-05T00:56:29 *** Palmik has joined #aichallenge 2011-12-05T00:56:35 (hence my questions about larger maps, i don't want to make "suboptimal" decisions to make my bot run on the specified dimensions and then have the contest barely even use them :\ 2011-12-05T00:56:37 at a rate of about the square that fits inside the vision circle 2011-12-05T00:56:38 BenJackson: I tried that, but it sucked 2011-12-05T00:57:02 oh I thought you were just trying to get started :) 2011-12-05T00:57:21 That's the only thing I tried 2011-12-05T00:57:30 and I tried tweaking it a bunch of ways 2011-12-05T00:57:31 I acutally calculate the perimeter of the explored space and add every one of those squares to a goal set 2011-12-05T00:57:46 but that's purely for exploration, not for spreading out 2011-12-05T00:57:55 my spreading out is ridiculously brute force 2011-12-05T00:58:02 i don't spread out at all 2011-12-05T00:58:26 and despite my spreading out logic I still don't blend in strategy in a way that doesn't sometimes let me collapse 2011-12-05T00:58:30 which can be awesome or terrible, depending 2011-12-05T00:58:42 finding the *last* enemy hill and having all my ants suddenly surround it is great 2011-12-05T00:58:49 if it's a huge battle and I just gave up tons of vis, oops 2011-12-05T00:59:30 do you have any special order functions for searching? 2011-12-05T00:59:33 even if it's a bfs? 2011-12-05T00:59:41 like a priority bfs 2011-12-05T00:59:43 BenJackson: to spread out do you just have each ant find its nearest neighbor and walk away from it? 2011-12-05T00:59:50 not yet 2011-12-05T01:00:05 i meant BenJackson :P 2011-12-05T01:00:08 bmh: for every ant I work out what new squares it would see for every move and what squares it would stop seeing 2011-12-05T01:00:19 and then I solve that for all the ants simultaneously to maximize total seen squares 2011-12-05T01:00:29 actually seen is wrong because it's by walking steps and not seeing steps 2011-12-05T01:00:48 oh more like terrirory? 2011-12-05T01:00:50 when I wrote it I just wanted to see how it would work 2011-12-05T01:00:58 since I couldn't think of a clever/fast way to do it 2011-12-05T01:01:02 and if it sucked I didn't want to keep thinking about it 2011-12-05T01:01:09 but it worked ok AND it didn't actually cost as much as I expected 2011-12-05T01:01:31 however my ants are generally too dense 2011-12-05T01:01:42 do you ever path find while avoiding potential enemy battles? 2011-12-05T01:01:48 that could be good but they don't really exhibit an "antibody" behavior like I'd hoped. 2011-12-05T01:01:53 individual enemy ants are just too hard to kill 2011-12-05T01:01:55 too slippery 2011-12-05T01:02:06 ie. calcualte the soonest battle with an enemy for each location and don't walk there if current path length is at least that soon 2011-12-05T01:02:38 no, my pathfinding ignores ants at the moment 2011-12-05T01:02:51 soonest enemy battle info is a bit of a pain to calculate, you need to do like the bfs but for each search location bomb the attack radius 2011-12-05T01:02:53 although on my list is to actually use weighted pathfinding 2011-12-05T01:07:53 McLeopold: are you worried about time or being smart enough? 2011-12-05T01:08:11 BenJackson: you iteratively collect food right? 2011-12-05T01:08:24 ie. don't send multiple ants to the same food? 2011-12-05T01:08:52 I pick specific ants for each food but I can pick the same ant for multiple food 2011-12-05T01:08:53 how do you do that? iterated bfs, simultaneous bfs? messily update info? 2011-12-05T01:09:02 same ant after its been moved? 2011-12-05T01:09:05 bfs from every individual food 2011-12-05T01:09:10 antimatroid: smarts 2011-12-05T01:09:14 the clever bit is what order to do the searches 2011-12-05T01:09:36 do you mean move an ant then re move it? 2011-12-05T01:09:42 or just consider an ant for multiple food items? 2011-12-05T01:10:08 i used to iteratively collect food with a* while reusing ants, but it's too slow on larger maps 2011-12-05T01:10:17 with a* or bfs 2011-12-05T01:10:18 remember there's not much fodo 2011-12-05T01:10:20 food 2011-12-05T01:10:27 so try to make the complexity scale with how much food you see 2011-12-05T01:10:32 not how many ants 2011-12-05T01:10:33 yeah i know 2011-12-05T01:10:39 hence searching from food to ants 2011-12-05T01:10:51 searching from food to ants and reusing ants is a pita though no? 2011-12-05T01:11:05 cause you need to keep searching after finding an ant until the current path is greater than the best path found 2011-12-05T01:11:14 Err 2011-12-05T01:11:20 well if you bfs from food the first ant you find is the closest 2011-12-05T01:11:27 That^ 2011-12-05T01:11:38 not if some of your ants considered for moving have already collected food and are "available" later than now 2011-12-05T01:11:45 if you bfs from food and the first thing you find is *other food*... 2011-12-05T01:11:48 didn't you say you're using one ant for multiple targets? 2011-12-05T01:11:55 ahhhh, i get you now :P 2011-12-05T01:12:06 i played with that last night 2011-12-05T01:12:13 i thought it didn't work, but i'll play again 2011-12-05T01:12:14 hence the importance of the order you do the food searches 2011-12-05T01:12:28 IMO the only time this is important is in the very early game 2011-12-05T01:12:29 do you order based on the number of adjacent food? 2011-12-05T01:12:42 later food pickup is dominated by other things 2011-12-05T01:12:55 and do you make the targets to food locations or the locations adjacent to food? 2011-12-05T01:13:04 I think working out the order is left as an exercise for the reader here... 2011-12-05T01:13:12 i think i might just search to adjacent locs cause otherwise it's a pita to correct for not actually walking on to targets 2011-12-05T01:13:14 antimatroid: I use the same technique I did for pw 2011-12-05T01:13:26 yeah sorry, i'm just poking around to see what others are doing :P 2011-12-05T01:13:30 I can bfs to future ants 2011-12-05T01:13:31 I use the food because using adjacent never becamet hat important 2011-12-05T01:14:02 I use the adjacent locations, because it's just as easy 2011-12-05T01:14:09 it's easier imo 2011-12-05T01:14:12 not that i'm doing it 2011-12-05T01:14:16 for my food pickup i simply assign one ant per food 2011-12-05T01:14:20 seems to work for me 2011-12-05T01:14:31 like I said, the food thing only really matters at the very beginning 2011-12-05T01:14:31 i don't want fine :P 2011-12-05T01:14:37 i understand that 2011-12-05T01:14:40 efficiently collecting the initial food is important 2011-12-05T01:14:57 after that almost anything works as long as it doesn't try to move ALL of your ants toward the food 2011-12-05T01:15:08 i don't know if i like your food thing, what about a...*..*..a? 2011-12-05T01:15:48 i don't need your solution, but are you correctly assigning closest ants to food? 2011-12-05T01:15:55 antimatroid: it depends on where the hill is 2011-12-05T01:16:03 i meant ben jackson 2011-12-05T01:16:19 my understand is he'd bfs from the food, find another food and not collect both? 2011-12-05T01:16:26 so then end up sending the right a to collect both? 2011-12-05T01:16:58 I think I would send both ants in with my code, but it might be worth it to harvest slower 2011-12-05T01:17:00 hm, the "past other food" condition is commented out of my code 2011-12-05T01:17:11 so I'd send one ant for a..a...*...* 2011-12-05T01:17:15 McLeopold: bfs'ing from future ants I do by ordering BFSSearchItem by current path yeah? 2011-12-05T01:17:18 that's what i'm doing anyway 2011-12-05T01:17:19 but both for a...*..*...a 2011-12-05T01:17:33 but only 1 for a.*..*........a 2011-12-05T01:17:57 i don't think i get what you're doing :P 2011-12-05T01:18:24 you search from the two *'s yeah? so what happens when they find each other? :P 2011-12-05T01:18:36 nothing, that condition is commented out 2011-12-05T01:18:40 antimatroid: another food item is a potential ant at some point, you just increase the distance by the detour amount, which is sometimes 0 2011-12-05T01:19:00 so you're sending one ant to a single food item? 2011-12-05T01:19:08 McLeopold: no it's not 2011-12-05T01:19:16 adjacent to the food in 1 direction is an ant at some point 2011-12-05T01:19:25 you don't actually need to walk onto the food and that no longer spawns into an ant 2011-12-05T01:19:33 meh, I assumed you'd account for that :) 2011-12-05T01:19:51 that's my point about not actually walking onto targets for collecting food 2011-12-05T01:20:01 you want to know where the ant is going to be adjacent to the food, not where the food is 2011-12-05T01:20:03 Not much of a point 2011-12-05T01:20:20 *** ltriant has quit IRC (Quit: Computer has gone to sleep) 2011-12-05T01:20:26 antimatroid: okay, any point is a potential ant 2011-12-05T01:20:31 it is when you're using info about where ants will be in the future :P 2011-12-05T01:20:39 *** zyberkiddy has joined #aichallenge 2011-12-05T01:20:40 Because when you walk next to it, the target goes away, so you're not actually walking on top of it 2011-12-05T01:21:07 but i need to know where my ant will be in the future 2011-12-05T01:21:15 and that's one of the locations adjacent to the food, not the food 2011-12-05T01:21:16 you do need to allow for suddenly appearing food 2011-12-05T01:21:21 yeah 2011-12-05T01:21:24 and food that starts next to you initially 2011-12-05T01:21:25 i just do that at the start 2011-12-05T01:21:51 in a local test I had a bot lose a game when it got stuck trying to walk AROUND initial food right by the hill 2011-12-05T01:21:56 into a crevice 2011-12-05T01:22:01 i figure if food suddenly appears in my path, I get a free food faster 2011-12-05T01:22:02 or alternatively just tell it to move onto food, although i think the bot test thing fails you for that? 2011-12-05T01:22:06 then it wouldn't come out because it didn't want to step on the hill again 2011-12-05T01:22:15 McLeopold: yes, but you need to stay still to collect it at the end of the turn 2011-12-05T01:22:19 true 2011-12-05T01:22:23 if you path find blocking food you'll move diagonal to it 2011-12-05T01:22:45 I think a1k0n lost a game with that, the initial food was next to the hill 2011-12-05T01:22:59 he only loses 1 turn to get the second ant? 2011-12-05T01:23:13 i think it'd be more an issue when an enemy can cause you not to get it 1 turn later 2011-12-05T01:23:33 I just remember swearing and a game link :) 2011-12-05T01:23:36 I've had two equally matched bots get off to radically different starts when one chose to step back into the "cave" the hill was in rather than the first ant going OUT 2011-12-05T01:23:37 :) 2011-12-05T01:23:41 so subtle 2011-12-05T01:23:51 but then that ant lost the race to every chokepoint 2011-12-05T01:24:19 Oh, now I know exactly what I'll do about food 2011-12-05T01:24:29 do you set up a perimeter around your hills or just try to maximise vision around them? 2011-12-05T01:24:56 i'm doing the former 2011-12-05T01:26:57 I used to nail 4 ants at the corners 2011-12-05T01:27:04 now it's more vague but I haven't really evaluated it 2011-12-05T01:27:09 nah i don't like that, too smaller perimeter 2011-12-05T01:27:23 i am pretty proud of my way to do it 2011-12-05T01:27:28 although it's actually simple in the end 2011-12-05T01:27:40 the tricky part of defense like that is knowing when to start defending 2011-12-05T01:27:48 initially you have to abandon your hill to explore 2011-12-05T01:27:55 yeah 2011-12-05T01:27:57 and assume you will defend with new ants due to feeding 2011-12-05T01:28:09 that's part of the reason that low-food multi-hill maps are so random 2011-12-05T01:28:09 antimatroid: do I need to pass parameters to generate multi_hill_mazes? 2011-12-05T01:28:14 one option is to only defend moving outwards 2011-12-05T01:28:21 from a hill 2011-12-05T01:28:34 McLeopold: not sure :P 2011-12-05T01:28:36 let me have a look 2011-12-05T01:29:08 GreenTea is charging up the ranks again 2011-12-05T01:29:13 McLeopold: it has min and max hill parameters 2011-12-05T01:29:24 BenJackson: watch out for dmj111 2011-12-05T01:29:29 anyone want to vote on the min/max values? 2011-12-05T01:29:37 i noticed he had a bot in the top 30 before, and by the looks of it didn't see the contest right away 2011-12-05T01:29:43 1 :P 2011-12-05T01:29:49 i hate the random tie break rule for spawning 2011-12-05T01:29:52 antimatroid: I swear I was going to just let the contest go at this point but then I broke into the top 10 2011-12-05T01:29:54 it's my least favourite thing about the game 2011-12-05T01:29:56 and it's on again 2011-12-05T01:30:22 i very much dislike things being determined randomly 2011-12-05T01:30:41 :p 2011-12-05T01:31:02 1 definitely should be the minimum 2011-12-05T01:31:44 BenJackson: i wont be happy until i finish top 10 after tron, that was devestating 2011-12-05T01:31:59 although i'm pushing it for time on this 2011-12-05T01:32:05 I chalk it up to quantum ant effects 2011-12-05T01:32:10 http://aichallenge.org/visualizer.php?game=161855&user=3473 2011-12-05T01:32:15 yes but i want to know where future ants will spawn :P 2011-12-05T01:32:23 I'm going to need to finally detect cul-de-sacs if those maps are popular 2011-12-05T01:32:27 it would be much nicer if i could just execute spawn orders 2011-12-05T01:32:45 those maps are my favourite so far 2011-12-05T01:32:52 that's what i wanted to make, but failed 2011-12-05T01:33:48 McLeopold: I wouldn't be opposed to some smaller width passages through those water walls 2011-12-05T01:34:09 just so long as a 3x3 block can traverse between all hills 2011-12-05T01:34:28 I think janzert (?) was saying that would be true 2011-12-05T01:34:40 that many of the existing random walks are ok despite the 1-square passes 2011-12-05T01:34:41 what was true? 2011-12-05T01:34:56 yeah some of them are carved with 3x3 i think 2011-12-05T01:35:03 but allow 1 width passages to turn up 2011-12-05T01:35:04 *** bmh has quit IRC (Quit: Page closed) 2011-12-05T01:35:11 but i wouldn't guarantee it, or that all map gens are doing that 2011-12-05T01:35:21 map gens ended up being used that i didn't expect to be used :P 2011-12-05T01:35:49 so far all I've really done since my last submission is make my bot about 15% faster 2011-12-05T01:35:56 i worked out how to do the translationally symmetric maps with pencil and paper, that felt good to solve :P 2011-12-05T01:36:39 I removed several bad random walks earlier 2011-12-05T01:36:46 the list is in a github commit 2011-12-05T01:36:52 the random walk mapgen should really be re done 2011-12-05T01:37:01 right now random_walk generation passes about 40% 2011-12-05T01:37:19 i would love to good map gen done with perlin noise 2011-12-05T01:37:24 i could never get that working well either 2011-12-05T01:37:43 that doesn 2011-12-05T01:37:45 t surprise me :P 2011-12-05T01:37:56 to see a* 2011-12-05T01:39:03 are any organizers still here? 2011-12-05T01:39:21 yeah 2011-12-05T01:39:45 i barely count :P 2011-12-05T01:41:46 Have you decided on a minimum distance between hills? 2011-12-05T01:42:01 I read 24, but then one map has them only 11 apart 2011-12-05T01:42:06 it's 20 2011-12-05T01:42:12 the spec page has been updated with it 2011-12-05T01:42:28 Antimony: the bad maps should have been removed 2011-12-05T01:42:29 ok thanks 2011-12-05T01:42:40 if you see one in a game, let us know 2011-12-05T01:42:56 and that is only enemy hills 2011-12-05T01:43:03 you can have your own hills within 20 2011-12-05T01:43:46 oh cool, lots of new restrictions 2011-12-05T01:44:00 maybe that will make my hill location inference slightly more efficient 2011-12-05T01:44:30 so the earliest possible hill razing is on turn 20? 2011-12-05T01:44:46 if you're stupid, yes :P 2011-12-05T01:44:52 Or lucky 2011-12-05T01:44:56 true 2011-12-05T01:45:21 Are those specifications likely to change before finals? 2011-12-05T01:45:30 antimatroid: does the mapgen.py do 3 or 5 players? 2011-12-05T01:45:38 Antimony: not likely 2011-12-05T01:45:41 but... 2011-12-05T01:45:42 ok good 2011-12-05T01:45:44 for translationally symmetric maps it should 2011-12-05T01:45:57 I guess I need to download the new maps for testing 2011-12-05T01:46:09 the ones in the starterpackage no longer meet the specifications 2011-12-05T01:46:37