2010-10-03T00:00:21 i've got 11th :( 2010-10-03T00:00:34 aw damn so close 2010-10-03T00:00:50 what is the level of sophistication that is needed to be in the top 100 right now? Top 10? 2010-10-03T00:00:56 not that much 2010-10-03T00:00:57 What sorts of algorithms are we talking here? 2010-10-03T00:01:02 my current one is still mostly heuristic based 2010-10-03T00:01:18 i'm currently working on some kind of approximation to a game tree to try minimax and see how that goes 2010-10-03T00:01:18 Ugh. That is not good. I wonder what we can do to step it up. 2010-10-03T00:01:26 *** yasith_ has joined #aichallenge 2010-10-03T00:01:26 but there's so many choices to be made i have no idea how it will go 2010-10-03T00:01:28 That would do it haha. 2010-10-03T00:01:46 i tried dynamic programming but too many possibe moves and couldn't see and easy way to reduce that 2010-10-03T00:02:11 i think people are going to be using simplex for choice selection at the top towards the end 2010-10-03T00:02:11 I find when I am playing Galcon I am doing a bit of a minimax thing in my head. Trying to figure out how to do the pruning will be tough though. It seems like I base it off of a thousand little heuristics. 2010-10-03T00:02:32 Simplex as in linear programming? 2010-10-03T00:02:38 yep 2010-10-03T00:02:49 fairly simple to set it up for moves to planets you want to make 2010-10-03T00:03:22 I can see how you would do it up. Each decision variable is the number of ships to send between each pair of planets, and the constraints are zero and the number of ships available at each planet. However, what would you use as a fitness function? 2010-10-03T00:03:36 i don't even mind sharing what my strategies are for this, there are so many and the details are many for each that i don't think it even really matters 2010-10-03T00:04:00 not entirely sure yet, Migi32 had an interesting idea 2010-10-03T00:04:07 Do you know what it was? 2010-10-03T00:04:08 *** yasith has quit IRC (Ping timeout: 276 seconds) 2010-10-03T00:04:09 maximise expected number of ships at the end of the game 2010-10-03T00:04:16 You don't say. 2010-10-03T00:04:38 so your ships - their ships + remainingturns(your growth - their growth 2010-10-03T00:04:47 Ahhhh 2010-10-03T00:04:50 Not bad. 2010-10-03T00:05:02 i would probably throw something into the mix about position strength 2010-10-03T00:05:11 And that gets maximized for some time period n turns in the future or what? 2010-10-03T00:05:53 *** delt0r_ has joined #aichallenge 2010-10-03T00:05:56 Like, you have to pick some time t where you evaluate that heuristic. 2010-10-03T00:05:57 yeah, just do it for the end condition if specified, otherwise that one might be a bit wonky 2010-10-03T00:06:11 or you could just try to maximise your number of ships 100 turns into the future or something 2010-10-03T00:06:18 if the end time is unspecified 2010-10-03T00:07:10 *** delt0r___ has quit IRC (Ping timeout: 252 seconds) 2010-10-03T00:07:14 I was thinking that this problem is great for linear programming. I'm just not sure how to choose weights for the maximization function. 2010-10-03T00:07:38 that's where it gets difficult 2010-10-03T00:08:11 i'm still not sure if the simple way of setting the problem removes the possibility of finding an optimal solution for which no feasible integer solution exists 2010-10-03T00:08:19 It seems like you sort of want to be able to say if(i_own_planet_n,some_linear_function,some_different_linear_function) 2010-10-03T00:08:21 i think you can set the constraints up such that that doesn't happen 2010-10-03T00:08:42 7 am ... need to go sleep :( 2010-10-03T00:08:48 night 2010-10-03T00:09:05 mb in my dream i will see better architecture of my bot ;) 2010-10-03T00:09:21 night ult0n 2010-10-03T00:09:30 *** u1tr0n has quit IRC () 2010-10-03T00:09:51 I think there's always going to be a feassible integer solution. You're always free to do nothing: (0,0,...,0) 2010-10-03T00:10:07 yeah, but the planets it spits out to claim as optimal 2010-10-03T00:10:23 i am unsure whether it's possible that there is a real solution without an integer solution to that 2010-10-03T00:10:47 i think it's fine, but i'm not positive 2010-10-03T00:11:13 The biggest problem is that you sort of need conditionals in the fitness function for the LP, which you can't really do. 2010-10-03T00:11:29 whichever way people go algorithm wise, unless it's got a good heuristic side well built into it, it wont do very well this time 2010-10-03T00:11:38 yup 2010-10-03T00:11:42 at the moment, the quality of the heuristics is the dominating factor 2010-10-03T00:11:44 I was thinking that some kind of interior point algorithm might work even if you had conditionals in the fitness expression? 2010-10-03T00:11:45 last time you could do reasonably well with a pretty stragith forward minimax implementation 2010-10-03T00:12:05 even later, more sophisticated strategies have to have strong heuristics underlying them 2010-10-03T00:12:27 because brute force search is just not feasible for almost any aspect of this problem 2010-10-03T00:12:42 I really really want someone to go hardcore on the machine learning to tune their heuristics. It doesn't seem like we've seen anything like that... yet. 2010-10-03T00:13:02 like my idea for minimax is pick some moves from us, them counter that, do that 2 or 3 turns, then estimate their moves with a probability distribution, run each child for a number of turns, evalulate those and run minimax on the tree 2010-10-03T00:13:25 still haven't worked out how to pick possible combinations of moves i and they might make with all of that information 2010-10-03T00:13:26 it's insane 2010-10-03T00:13:28 j3camero: I don't think anyone is at the stage where they know what parameters to use for something like that 2010-10-03T00:13:35 maybe they are, but I wouldn't bet on it 2010-10-03T00:13:41 people are still figuring this stuff out 2010-10-03T00:13:46 i'm playing to let it churn away for a few days at some point trying out different parameter values 2010-10-03T00:14:06 I wonder what I can do to give people a steeper incentive to really create cool stuff. 2010-10-03T00:14:27 i have fucked around with parameters manually a lot, but need to get the computer to really give it a go once things are better set up, no point doing it now 2010-10-03T00:14:28 let us download the st of games 2010-10-03T00:14:34 or the set of top games 2010-10-03T00:14:36 j3camero: prizes :P 2010-10-03T00:14:42 We gotta find some people to do sportscaster-style news coverage of the contest on a daily basis. We could even do Skype interviews with the top contestants and shit. 2010-10-03T00:14:42 that too 2010-10-03T00:14:49 did that person that emailed you the other day pull through with anything? 2010-10-03T00:15:18 j3camero: are we getting a live leaderboard during the final tournament this time? that'd be so much more awesome 2010-10-03T00:15:22 j3camero: I think dmj111 is planning to continue blogging about what he's doing 2010-10-03T00:15:28 Yeah he was quite nice. Memberships to his AI website. He hasn't got back to me yet though. 2010-10-03T00:15:28 even if not being able to view games 2010-10-03T00:15:48 ah, that's not really much of a prize :P 2010-10-03T00:16:10 don't get me wrong, i'm competing anyway, but prizes do motivate people 2010-10-03T00:16:25 antimatroid2: no promises, but the odds are higher this time due to open source. I took it down last time because I was so tired of people crying foul, and I knew people would lose their shit every time their ranking moved and blog about how the cotnest is corrupt. 2010-10-03T00:16:38 o_O 2010-10-03T00:17:04 haters gonna hate regardless 2010-10-03T00:17:13 antimatroid2: yeah, prizes would be so awesome. I wonder if I could put up some prizes... 2010-10-03T00:17:29 have you considered an open paypal account? 2010-10-03T00:17:32 antimatroid2: actually, I found that if people couldn't see the leaderboard, they quieted right down. 2010-10-03T00:17:35 let people donate for prize money? 2010-10-03T00:17:37 i would 2010-10-03T00:18:02 j3camero: it's a lot of fun to see the leaderboard evolve throughout the tournament though 2010-10-03T00:18:02 j3camero: but you also quieten the people interested in the competition? 2010-10-03T00:18:14 antimatroid2: ohhh this is a great idea. Maybe say that half the donations go to prizes, and half goes to a slush fund for the next contests. We are starting to make big plans that require at least a bit of dough. 2010-10-03T00:18:26 yep 2010-10-03T00:18:28 a lot of the chat here revolves around the leaderboard, without it this would be less exciting 2010-10-03T00:18:49 sigh: yeah you're right. This was only during the final tournament after the submission deadline we're talking about. 2010-10-03T00:19:01 contestbot tells us the top 10 every hour, and often that sparks a discussion 2010-10-03T00:19:02 yeah, but that's when it's most exciting :P 2010-10-03T00:19:02 sigh: I'm sorry Dave, I cannot tells. 2010-10-03T00:19:15 :P 2010-10-03T00:19:37 I think it will hurt more than help if you hide the leaderboard 2010-10-03T00:19:57 if the server could handle people also viewing games, that would be awesome 2010-10-03T00:20:11 otherwise maybe pick a few "cool" games each day to release? 2010-10-03T00:20:16 sigh: yeah the odds are higher of the leaderboard staying up for the final tourney this time round, since it's all open source. There is much less chance of people being unreasonable about the ranking algorithm and matching procedure and such. 2010-10-03T00:20:31 cool 2010-10-03T00:20:47 how long are you planning to run the final tourny for? 2010-10-03T00:20:55 i think a week would be awesome with a live leaderboard 2010-10-03T00:21:19 and scores need to be reset at the deadline so that everyone starts from scratch, but i think daniel's onto that one 2010-10-03T00:22:01 sigh: the element of surprise was also a factor though. It's cool to embargo it for a while to build up the anticipation. Otherwise it becomes clear who the winner is pretty early. To tell you the truth, we barely needed the final tourney last time. People were being so crazy about making sure that every top50 person played every other top50 person. After all was said and done, it didn't make much difference. The rankings stayed dead steady for the whol 2010-10-03T00:22:31 also, for next time, would it be possible for people to specify whether they're happy to share their final code and have their profile automatically have it downloadable after the final tournament should they say they're happy for that to happen? would make it a lot cleaner to find shared code at the end of the contest 2010-10-03T00:22:43 google codejam shares everyones code i think 2010-10-03T00:23:09 yeah agreed on the rankings 2010-10-03T00:23:20 i was one of the people bitching about that a lot, but it really was quite steady 2010-10-03T00:23:26 antimatroid2: yeah I think that's a great idea. People could download the zip file they submitted originally, maybe? 2010-10-03T00:23:36 yep 2010-10-03T00:23:48 they're usually pretty small aren't they? 2010-10-03T00:23:52 mine is 2010-10-03T00:24:01 for the most part yeah. 2010-10-03T00:24:34 i wouldn't force people to have their code shared, might discourage some people from participating 2010-10-03T00:24:41 *** FBP has quit IRC (Quit: ChatZilla 0.9.86 [Firefox 3.6.10/20100914125854]) 2010-10-03T00:24:45 do that once the competition is complete viral :P 2010-10-03T00:24:54 We could give people a few options actually: (1) dont release (2) release after the deadline (3) release with a one week delay (4) release with a one hour delay. 2010-10-03T00:25:16 *** pjreddie has quit IRC (Ping timeout: 265 seconds) 2010-10-03T00:26:06 Yeah so today we set up a bunch of workhorse servers on Amazon EC2 to give the bots a bunch more play! 2010-10-03T00:26:14 cool :) 2010-10-03T00:26:22 The goal is to get it so newly submitted entries have their rankings converge in under one minute. 2010-10-03T00:26:30 that would be awesome 2010-10-03T00:26:42 How long do you find it takes your bot's ranking to converge after submission? 2010-10-03T00:26:50 a few hours 2010-10-03T00:26:54 Oh really... 2010-10-03T00:27:06 i'm not too fussed with tcp though :P 2010-10-03T00:27:18 that was yesterday 2010-10-03T00:27:22 my last submission was a week ago 2010-10-03T00:27:30 The TCP model is a good idea I think. 2010-10-03T00:27:32 that took a day or so to converge I think 2010-10-03T00:28:01 I am toying with the idea of just letting people run their own bots. They would just connect to a TCP tournament manager. That way if you want to go set your bot up on a bigass cluster you can do that. 2010-10-03T00:28:03 sigh: you also lost your first game didn't you? 2010-10-03T00:28:07 2nd :P 2010-10-03T00:28:15 j3camero: can't do time limits with that 2010-10-03T00:28:32 tcp has to have a high time limit to account of latency 2010-10-03T00:28:32 antimatroid2: why not? 2010-10-03T00:28:45 like it's about 5 seconds now i heard 2010-10-03T00:28:57 10 sec first move, 5 second for the rest 2010-10-03T00:29:05 Oh yeah latency. Meh. It's not a big issue. Since we're not actually running people's code, I could care less. Your bot could take 1 minute per turn if it wants. 2010-10-03T00:29:21 competition would take ages to run 2010-10-03T00:29:25 oh wait no 2010-10-03T00:29:26 err :P 2010-10-03T00:29:31 Throughput, my good sir. 2010-10-03T00:29:39 yeah okay, so just remove time limits? 2010-10-03T00:29:45 We could run thousands of games concurrently. 2010-10-03T00:30:07 you could set some time limit to remove absurdness 2010-10-03T00:30:16 Yeah we could just remove time limits, or else have one hour time limits or some such. 2010-10-03T00:30:28 i'd say like 30 seconds or a minute 2010-10-03T00:30:43 otherwise the opponent gets stuck in that game for AGES 2010-10-03T00:30:52 absurd time limits is no good I think 2010-10-03T00:30:57 you get less games :( 2010-10-03T00:31:05 sigh: not really 2010-10-03T00:31:10 cause you are always playing someone 2010-10-03T00:31:13 The average use case will be less than one second per turn. It would only be the pwoer users who take ridiculous amounts of time. Since it doesn't hurt the contest, that's fine. They would udnerstand that the tradeoff is that you get less games in. 2010-10-03T00:31:19 as opposed to waiting for the server to be able to play you 2010-10-03T00:31:22 if it's a minute per moves, then potentially you only start a new game every hour or two 2010-10-03T00:31:29 *minute per move 2010-10-03T00:31:41 10 seconds then 2010-10-03T00:31:45 assuming that one bot can only play one game at a time 2010-10-03T00:31:56 allows people using "lesser" languages to not bitch 2010-10-03T00:32:10 Oh, I guess the issue is if you take such a long time that your opponent forfeits by shutting down his bot. Didn't think of that. So we would need some kind of time limit, like 10 seconds. 2010-10-03T00:32:17 although, you might then have lots more timeouts from people with crap connections 2010-10-03T00:32:19 I think the current timeout on tcp is reasonable 2010-10-03T00:32:23 so anyone with a good connection has an advantage? 2010-10-03T00:32:36 Yeah. 2010-10-03T00:32:54 dear uni, may i please have a port opened, no fuck off 2010-10-03T00:32:57 Hardware fairness goes out the window if we let people run their own bots. Someone can go run it on a cluster if they want. 2010-10-03T00:33:03 but, overall I think tcp is awesome for an informal tournament and testing, but not very fair at all 2010-10-03T00:33:30 yeah, i think it's better to keep the final tournament on similar machines with similar time limits 2010-10-03T00:33:40 donations :P 2010-10-03T00:33:42 I think it's a minimal issue, honestly. If the internet is fast enough to play FPS, then latency doesn't matter. 2010-10-03T00:34:20 do there exist places still on dial up? 2010-10-03T00:34:35 can you play fps games on dial up? 2010-10-03T00:34:40 i have no idea for these things 2010-10-03T00:34:42 Yes. But it's unlikely that anyone who would give us grief about latency is using dial up. 2010-10-03T00:35:30 but yeah, not everyone can get on tcp 2010-10-03T00:35:57 yep, a lot of people keep timing out and just give up 2010-10-03T00:36:28 I kept timing out, but that got fixed when I moved my comp from wireless to wired 2010-10-03T00:36:37 but moving to wired was very annoying for me 2010-10-03T00:36:48 2% of people looking at the web page are on dialup. 2010-10-03T00:37:09 isn't it ideal for the competition to be open for everyone? 2010-10-03T00:37:28 what about the ability to upload your submission and have it run by you? 2010-10-03T00:37:34 or run it locally 2010-10-03T00:37:40 so people could upload if they insist 2010-10-03T00:37:46 Yeah I thought of that. 2010-10-03T00:37:49 that way you'd have less to deal with 2010-10-03T00:38:22 Or keep it the way it is, but have an option to connect by TCP for the power users. So if anyone wants to bitch about hardware fairness, I give them the connection details and tell them to rent a cluster. 2010-10-03T00:38:55 that's an option 2010-10-03T00:39:08 i personally think it's better to have them all submitted 2010-10-03T00:40:00 ^^ I second that 2010-10-03T00:40:49 I rather this be about who writes the best algorithm, than who can get the best hardware and connection 2010-10-03T00:42:16 *** keynes has quit IRC (Ping timeout: 255 seconds) 2010-10-03T00:42:17 Hardware hackers are hackers too haha. 2010-10-03T00:42:40 :P 2010-10-03T00:42:41 Of course, if you sample the people who are already taking part (who are all software hackers) then they would prefer hosted bots. :-) 2010-10-03T00:43:48 I wonder if we could do some really intense shit where we let top bots have more resources. I wonder if people who like that. I would love to see some people experimenting with parallel algorithms and stuff. 2010-10-03T00:44:12 i'd like to see some games with more planets if possible? 2010-10-03T00:44:20 my bot can handle variable planets 2010-10-03T00:44:25 i assume most people can 2010-10-03T00:45:00 Yeah we should write a different map generator. The one we have sucks balls. I wrote it on the bus one day just to have maps to test the engine on. 2010-10-03T00:45:11 some purposely degenerate maps would be interesting 2010-10-03T00:45:35 yep 2010-10-03T00:45:37 for example, a regular pattern 2010-10-03T00:45:52 Cool idea. I should generate some maps that have three-way symetry, with only two of the sides occupied by players to begin with. It would be a fight for control of the remaining side. 2010-10-03T00:45:53 or perfect circles etc. 2010-10-03T00:46:10 that'd be interesting 2010-10-03T00:46:10 j3camero: that's a really good idea 2010-10-03T00:46:12 or to be evil, hexagonally packed planets at a distance of 1 unit :) 2010-10-03T00:46:19 start making interesting maps, like starshaped, etc 2010-10-03T00:46:24 circle, square 2010-10-03T00:46:26 stealth fight :) 2010-10-03T00:46:27 square with an X 2010-10-03T00:46:29 yes, or do a square lattice of planets distances 1 apart 2010-10-03T00:46:29 Oh snap that would be interesting. I like the honeycomb map idea. 2010-10-03T00:46:39 lol 2010-10-03T00:46:41 i really want that now 2010-10-03T00:46:45 that can be crazy at 1 unit ? 2010-10-03T00:46:47 Or the square lattice. 2010-10-03T00:46:47 how many planets ? :| 2010-10-03T00:46:52 my bot will die :| 2010-10-03T00:46:56 you never see enemy fleets sent from planets distanced 1 apart 2010-10-03T00:47:02 yeah >.< 2010-10-03T00:47:04 Oh shit yeah if they're one unit apart then it would be intense. 2010-10-03T00:47:23 That's a totally different ball game. 2010-10-03T00:47:33 how hard is it for you to add a map? 2010-10-03T00:47:38 Easy. 2010-10-03T00:47:45 Give me the map, I will add it. 2010-10-03T00:47:52 does it have to have 23 planets? 2010-10-03T00:47:57 No 2010-10-03T00:48:01 Just a text file. 2010-10-03T00:48:04 okay, i'ma make some 2010-10-03T00:48:08 Yup. 2010-10-03T00:48:16 cool :) 2010-10-03T00:48:28 Is anyone here a fan of hard glitch music? 2010-10-03T00:48:33 don't go crazy with number of planets :| 2010-10-03T00:48:58 i wont 2010-10-03T00:49:09 i'll try to keep it below 30 2010-10-03T00:49:13 i WILL* 2010-10-03T00:49:13 ah ok 2010-10-03T00:49:15 that's good 2010-10-03T00:49:18 the square lattice degenerates into a much simpler game. 2010-10-03T00:49:25 i was like worried you put in 100-200 planets :| 2010-10-03T00:49:41 that when i got my bot working, it'd just choke >.< 2010-10-03T00:49:42 j3camero: that's assuming all planets are equal 2010-10-03T00:49:45 get 2010-10-03T00:49:48 more to the game than distance 2010-10-03T00:49:51 it's basically like a chess board where you stack black or white chips on each square. 2010-10-03T00:50:27 hmm, if i do a c++ function to output to a file on windows, is the normal endl going to go fine on the servers? 2010-10-03T00:50:39 No, but I'll fix it. 2010-10-03T00:50:42 i have to edit maps with notepad++ for the save to not fuck it up 2010-10-03T00:50:43 okay 2010-10-03T00:51:33 Crazy thing about the square lattice is that you can actually still go diagonally if you want. 2010-10-03T00:51:46 you can go anywhere 2010-10-03T00:51:49 It only makes sense to do so over longer distances, but you can do it. 2010-10-03T00:52:12 why only over longer distances? 2010-10-03T00:52:33 hopping from planet to planet is always slower than direct 2010-10-03T00:52:41 Oh man. People here should check out "Is You" by D.I.M. Make sure you get the original mix, not the dumb remixes. It's glorious coding music. 2010-10-03T00:53:36 The distance is ceil(x*x+y*y). You would almost never want to jump diagonally by one square, when you could go two sides around the square. Takes the same number of turns, but also gives you the option of turning around. 2010-10-03T00:53:44 this will be done soonish, what kind of lattice ones do you want? 2010-10-03T00:53:52 i will do circular ones going out from the middle after 2010-10-03T00:54:04 j3camero: hopping reduces hte number of units available 2010-10-03T00:54:04 with whatever density for each circle 2010-10-03T00:54:13 becuase of the order that the engine handles things 2010-10-03T00:54:14 i will do 5x5 lattices? 25 planets? 2010-10-03T00:54:24 1 distance apart? want any other distances apart? 2010-10-03T00:55:01 if you go A -> B -> C then B gets stuck with G(A)+G(B) units 2010-10-03T00:55:07 where G(x) is growth of planet x 2010-10-03T00:55:39 whereas if A and B both go direct to C then B only gets stuck with G(B) units 2010-10-03T00:55:55 it's because you make orders before you process growth and arrivals 2010-10-03T00:56:01 where do you want starting positions? top left and bottom right? 2010-10-03T00:56:56 antimatroid2: it shouldn't matter 2010-10-03T00:57:12 yeah, so which do i pick :P 2010-10-03T00:57:25 flip a coin 2010-10-03T00:58:51 Top 10 players: ruilov(3903), davidjliu(3853), Hazard(3817), rsergio(3769), felixcoto(3736), EBraun(3723), cfaftw(3711), dmj111(3711), bix0r4ever(3659), sequoh(3648) 2010-10-03T00:59:25 j3camero: simpler way to explain it - hopping takes an extra turn :P so diagonal is also faster 2010-10-03T00:59:46 yup 2010-10-03T00:59:47 diagonal is two moves 2010-10-03T00:59:55 hoping is just as fast 2010-10-03T00:59:58 I was assuming that all the planet start empty. 2010-10-03T01:00:14 so you can to left then right or diagonally in the same number of turns 2010-10-03T01:00:20 antimatroid2: the travel time is 2 moves, but you are stuck on middle planet for a move 2010-10-03T01:00:36 you can move there 2010-10-03T01:00:36 on the middle turn 2010-10-03T01:00:51 you can't, you issue order -> then arrivals come -> then execute order 2010-10-03T01:00:59 you can't use the incoming arrivals in your order 2010-10-03T01:01:16 yeah you move left, next turn that arrives and you own that planet, so you then move to the diagonal planet 2010-10-03T01:01:17 jut like you can't use the growth 2010-10-03T01:01:20 total trip length is 2 turns 2010-10-03T01:01:28 you don't get growth for that claim though 2010-10-03T01:01:33 on that turn 2010-10-03T01:01:50 what kind of growth and ships? 2010-10-03T01:02:02 random 2010-10-03T01:02:06 okay imma play some fucking galcon now. I can't take it any more. 2010-10-03T01:02:16 heh 2010-10-03T01:02:18 i was thinking it'd be cool to have it constant across at least all neutral 2010-10-03T01:02:31 try both 2010-10-03T01:02:37 okay 2010-10-03T01:02:47 my worry is that constant might be a trivial stalemate 2010-10-03T01:02:59 yeah okay 2010-10-03T01:03:17 i have the necessary function in my bot already anyway 2010-10-03T01:04:41 necessary function? 2010-10-03T01:05:12 def win_now(); 2010-10-03T01:06:40 pick a number for me between x and y 2010-10-03T01:06:54 z? 2010-10-03T01:07:09 x+eplison 2010-10-03T01:07:29 *epsilon 2010-10-03T01:07:35 z*x+(1-z)*y, 0<=z<=1? 2010-10-03T01:09:26 McLeopold's wormbot is a good idea of what happens if you hop: http://72.44.46.68/getplayer?player=McLeopold.WormBot.py 2010-10-03T01:09:36 units get "stuck" on the intermediate planets 2010-10-03T01:09:55 not sure about the effect on travel time anymore 2010-10-03T01:10:35 *** askhader has joined #aichallenge 2010-10-03T01:12:44 *** jimmerton_ has joined #aichallenge 2010-10-03T01:14:46 *** jimmerton has quit IRC (Ping timeout: 240 seconds) 2010-10-03T01:14:46 *** jimmerton_ is now known as jimmerton 2010-10-03T01:15:12 hmmm... I think I'm wrong 2010-10-03T01:17:30 Okay time to do some more art using python. w00t. 2010-10-03T01:17:34 See yall later. 2010-10-03T01:17:45 bye 2010-10-03T01:22:44 took me awhile to understand what the point of planet hopping was 2010-10-03T01:23:08 at first it seemed really dumb, but a lot of the good bots did it... 2010-10-03T01:23:14 I understand the point, just a matter of whether the delay is worth it 2010-10-03T01:23:23 sigh: yes. 2010-10-03T01:23:25 a lot f good bots also go direct 2010-10-03T01:23:55 there are good times to go direct and good times to hop. Also, it depends on what bot you're playing... 2010-10-03T01:24:01 :) 2010-10-03T01:24:29 then there's stuff like everyone hop staging planets 2010-10-03T01:24:41 *hop to 2010-10-03T01:25:31 ok, I've decided I was wrong... in the lattice case hopping doesn't perform worse than direct 2010-10-03T01:25:44 * sigh puts on his dunce cap 2010-10-03T01:29:38 *** boegel has joined #aichallenge 2010-10-03T01:29:52 can i got with 36 planets? 2010-10-03T01:29:56 it's cooler sizer 2010-10-03T01:30:06 do what you like :P 2010-10-03T01:35:44 you guys are setting up another server or something? 2010-10-03T01:37:09 nah, j3camero said he was interested in the idea of some designed maps 2010-10-03T01:38:11 hmm, contestant designed maps? 2010-10-03T01:38:27 Why not? 2010-10-03T01:38:59 because you could tailor them to what your bot is good at 2010-10-03T01:39:21 if you had a special trick for a certain configuration, put that all over the place... 2010-10-03T01:39:31 rwa: not that in that way, just in the sense of some non-random maps 2010-10-03T01:40:05 such as lattices, regular arrangements, etc 2010-10-03T01:41:31 anyway, the point was about having some maps that were specifically designed to be "interesting" 2010-10-03T01:41:41 don't think that's a good idea either. perhaps somebody worked out one of those cases ahead of time, and then gets to submit it as an official map. not fair. 2010-10-03T01:48:22 *** DaTwinkDaddy has joined #aichallenge 2010-10-03T01:49:23 @later tell dhartmei Some of the maps on the TCP server are duplicates: 435 and 436 are the same; 201-300 are the same, too. 2010-10-03T01:49:23 DaTwinkDaddy: As you wish. 2010-10-03T01:49:37 @seen dhartmei 2010-10-03T01:49:37 DaTwinkDaddy: dhartmei was last seen in #aichallenge 3 days, 13 hours, 4 minutes, and 59 seconds ago: project selection problem? 2010-10-03T01:50:14 @rankings 2010-10-03T01:50:15 DaTwinkDaddy: Top 10 players: ruilov(3905), davidjliu(3827), Hazard(3807), rsergio(3757), felixcoto(3722), EBraun(3717), cfaftw(3715), dmj111(3710), bix0r4ever(3647), sequoh(3644) 2010-10-03T01:50:16 rank 112 atm, wondering if my bot could get in the top 100... 2010-10-03T01:50:47 boegel: I was surprised when I hung out in the 60s for a while. 2010-10-03T01:50:57 DaTwinkDaddy: :) 2010-10-03T01:50:59 boegel: Now, I've dropped down to above 100. :( 2010-10-03T01:51:13 DaTwinkDaddy: still, not too bad, right? :) 2010-10-03T01:51:56 *** keynes has joined #aichallenge 2010-10-03T01:52:56 'draw' with dhartmei: http://ai-contest.com/visualizer.php?game_id=5276889 \o/ 2010-10-03T01:57:59 *** Zepp has joined #aichallenge 2010-10-03T01:58:52 Top 10 players: ruilov(3906), davidjliu(3827), Hazard(3808), rsergio(3757), felixcoto(3723), EBraun(3717), cfaftw(3715), dmj111(3710), bix0r4ever(3647), sequoh(3644) 2010-10-03T01:59:05 gah this is retarded 2010-10-03T01:59:29 it's somehow changing the players planets back to neutral for god knows that reason 2010-10-03T02:08:51 *** krokokrusa has joined #aichallenge 2010-10-03T02:12:36 j3camero: map maker http://codepad.org/byUUbJoO 2010-10-03T02:12:44 eample 5x5 map http://codepad.org/DheND3m8 2010-10-03T02:12:57 with distances of 1 between points 2010-10-03T02:14:02 rwa: i kind of agree with you, but it'd be nice to have some maps like this as well 2010-10-03T02:14:51 antimatroid2: I understand the draw. But it's hard to beat random maps for impartiality in a contest setting. 2010-10-03T02:15:23 this is still ranomly picking nuetral ship sizes and growth rates 2010-10-03T02:16:05 true 2010-10-03T02:16:07 you're just specifying dimensions and distances, i've assumed players start with 100 and growth 5, but otherwise it picks a starting location for them too 2010-10-03T02:17:41 i couldn't get it to output the coordinates as actual double values :| 2010-10-03T02:19:28 I'm stuck around 30 at the moment. My last 5 or 6 ideas haven't panned out very much. 2010-10-03T02:20:18 http://codepad.org/5VyYR4hY 2010-10-03T02:20:21 6x6 map 2010-10-03T02:21:15 http://codepad.org/DLeYSPZF that's the better generator too, other one was picking starting locations only in top left 2x2 block for all dimensions 2010-10-03T02:30:48 antimatroid2: you need to prefix the lines with a P 2010-10-03T02:31:35 rwa: yeah, I'm having the same problem 2010-10-03T02:31:55 you're stuck considerably higher than me, though... 2010-10-03T02:32:17 that would possibly explain my issues :P 2010-10-03T02:33:35 h'okay, http://codepad.org/GV3F4XLE, sorry about the spam, use that map generator then the files work 2010-10-03T02:33:42 rwa: not much difference, I am closer to you than say dmj111 2010-10-03T02:35:23 only ~100 elo points 2010-10-03T02:35:32 oh wait, i have one more desired fix :P 2010-10-03T02:37:58 http://codepad.org/TYUySfEO final copy of generator :) 2010-10-03T02:38:30 i'm curious how bots would go against each other on it 2010-10-03T02:39:56 hmmm, i think we could do a map where all planets are distances one apart from the others? 2010-10-03T02:40:04 just put all planets in a ball of radius 1? 2010-10-03T02:40:07 that would be interesting 2010-10-03T02:41:18 that's an awesome idea aha 2010-10-03T02:43:50 my bot can handle a 10x10 map still fairly fast 2010-10-03T02:44:01 and i got a winner in 92 turns 2010-10-03T02:44:14 i think we could maybe do much larger maps and not have the turns necessarily blow out so much 2010-10-03T02:44:37 *** espes has joined #aichallenge 2010-10-03T02:48:37 *** krokokrusa has quit IRC (Ping timeout: 265 seconds) 2010-10-03T02:48:51 sigh: have you played some games on these maps? they're actually pretty cool 2010-10-03T02:49:01 can you paste one? 2010-10-03T02:49:10 what size do you want? 2010-10-03T02:49:17 i am just testing 13x13 but it's pretty slow 2010-10-03T02:49:22 paste the one you thought was most interesting 2010-10-03T02:49:25