Timeline for BrainF***edBotsForBattling - A Brainf*** Tournament
Current License: CC BY-SA 3.0
30 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Sep 12, 2014 at 12:54 | history | edited | weston | CC BY-SA 3.0 | added 94 characters in body |
| Sep 11, 2014 at 8:10 | history | edited | weston | CC BY-SA 3.0 | added 367 characters in body |
| Sep 8, 2014 at 17:40 | comment | added | weston | @Sylwester sure, I don't use the percentage as the fitness rating, see the bit entitled "Potential bots are ranked by" I'd be careful applying one value to it, your way 101 draws is better than 1 win. | |
| Sep 8, 2014 at 15:44 | comment | added | Sylwester | Mickey is going places :) I'm actually using (win*100+tie)/total since I'm planning to try something genetic and then a tie is better than a loose. | |
| Sep 7, 2014 at 16:17 | comment | added | weston | @Sylwester I would do the float score thing, but I find a percentage is more comparable round to round as the number of available points varies. I think the tournament should have a % column. I've calculated mine as 1124/1260=89.2%. I know LethalLoke got 537/610=88%. I'm still unable to run the actual python tournament on windows, so the proof will be in the pudding. However I only recently added the java tournament runner with the random lengths. It mostly mirrors the results, but something's wrong with it because LethalLoke is far from lethal. So I'm not as confident in that win percentage. | |
| Sep 7, 2014 at 16:09 | history | edited | weston | CC BY-SA 3.0 | added 27 characters in body |
| Sep 7, 2014 at 16:08 | comment | added | weston | @Sylwester no. I do all 21. Take the win loss draws quoted 1124+62+74 = 1260, 21*60 = 1260 | |
| Sep 7, 2014 at 13:05 | comment | added | Sylwester | With 4 second tournament speed you could just do all 21 and make a compatible float score by multiplying 10/21 to make the runs consistent. I hope you don't do 10 random rounds as a fitness check? | |
| Sep 7, 2014 at 12:55 | history | rollback | weston | Rollback to Revision 8 | |
| Sep 7, 2014 at 12:50 | history | edited | weston | CC BY-SA 3.0 | formatting |
| Sep 7, 2014 at 12:33 | history | edited | weston | CC BY-SA 3.0 | added 126 characters in body |
| Sep 7, 2014 at 11:31 | history | edited | weston | CC BY-SA 3.0 | added 63 characters in body |
| Sep 7, 2014 at 10:07 | history | edited | weston | CC BY-SA 3.0 | added 110 characters in body |
| Sep 6, 2014 at 21:48 | history | edited | weston | CC BY-SA 3.0 | V3 |
| Sep 5, 2014 at 18:58 | history | edited | weston | CC BY-SA 3.0 | added 62 characters in body |
| Sep 5, 2014 at 18:14 | history | edited | weston | CC BY-SA 3.0 | added 14 characters in body |
| Sep 5, 2014 at 15:01 | comment | added | weston | @Sylwester It's not clear from my text, but I only trim dead code at the end. My thinking is I play every bot in every length so code that doesn't run, won't be run in the real thing. It stopped my lengths exploding. I do however compact the code prior to testing, so -+ is removed, <> is removed. They stay in the genes though so next gen might evolve to <->. | |
| Sep 5, 2014 at 14:54 | comment | added | Sylwester | @weston Thanks. BTW are you sure it's wise to remove dead code from your algo? I'm thinking dead code are rarely dead in BFJ. eg. +[] is an infinite loop in BF but not in BFJ. Likewise the last loop in [ something ][+] can make it win against Ràn. And we do get mutation in unactivated genes that might get activated in the future with either horrible or beneficial consequences? | |
| Sep 5, 2014 at 11:39 | comment | added | weston | @Sylwester Took them from github, so was the last round. getting updates from here too would be a smarter strategy. Well done on LethalLoke btw! | |
| Sep 5, 2014 at 11:35 | comment | added | Sylwester | When you said you used most did you fetch new versions from here or the versions from last round? | |
| Sep 4, 2014 at 7:33 | comment | added | weston | @LymiaAluysia Thanks, I'll check them out, but I wrote my own very fast java implementation. That was half the fun for me. I just use the python one just to verify how it does in the tournament. | |
| Sep 4, 2014 at 1:01 | comment | added | Lymia Kanokawa | It'd probably be best to use gearlance or another BF Joust interpreter for evolutionary algorithms-- they're much much faster, and work close enough to the interpreter used here to evolve against. Just modify the code so it doesn't run inverse polarity rounds. | |
| Sep 2, 2014 at 22:43 | comment | added | Sylwester | Nice. I'm excited to see how it does tomorrow :) | |
| Sep 2, 2014 at 21:58 | comment | added | weston | @Sylwester I use most of the bots now as my generator is much faster and multithreaded. | |
| Sep 2, 2014 at 21:56 | history | edited | weston | CC BY-SA 3.0 | deleted 341 characters in body |
| Sep 1, 2014 at 8:48 | comment | added | weston | @Cipher thanks, I opened an issue on the github github.com/redevined/brainfuck/issues/2 | |
| Sep 1, 2014 at 8:29 | comment | added | redevined | @weston Maybe we could help you with your python troubles? | |
| Sep 1, 2014 at 7:40 | comment | added | weston | @Sylwester Maybe, I mainly didn't because it's so slow. Took two hours to do this. Also though it evolved slower the more bots it had to cope with. I figured if it beats the top guys, it will do well. I've not been able to run the actual tournament to confirm, python troubles. So thought I'd just post it and see... | |
| Aug 31, 2014 at 23:51 | comment | added | Sylwester | Wouldn't it be better to use all 60 bots? | |
| Aug 31, 2014 at 18:49 | history | answered | weston | CC BY-SA 3.0 |