User talk:Jdev/News
- 02.11.09: UltraMarine coming...
Basing on 70-rounds battle results with bots from TOP-25 (on the following image), i guess that now i reach TOP-25 milestone
My statistics:
(place: pass / die (pase rate), avg energy) 1: 1 / 1 (1), 0 2: 1 / 3 (0,333), 8,406 3: 4 / 9 (0,444), 24,13 4: 13 / 7 (1,857), 26,229 5: 21 / 6 (3,5), 31,353 6: 25 / 6 (4,167), 39,044 7: 31 / 6 (5,167), 40,352 8: 38 / 10 (3,8), 44,153 9: 48 / 7 (6,857), 46,728 10: 55 / 15 (3,667), 55,234 Hit rate (319/1498) = 0,213
Next milestone is not last place with TOP-10 bots. Next for my bot is:
- develop gun (now he use HOT and CT)
- develop late phase (1-4 enemies) movement
- review and refactor all code
- tune and optimze all alghoritms
Looking good! (Btw, maybe the above should be on User:Jdev/News, and comments would go here? Up to you.) Be warned that melee benchmarking is a bit like voodoo. It's easy to see that SandboxDT beats Portia in this battle, but Portia far outranks DT in the rumble. But you must be doing something right to hold your own against top bots. Don't be afraid to "release early, release often". Good luck. =) --Voidious 20:23, 3 November 2009 (UTC)
Indeed, nice start. I don't think melee benchmarking is really so much voodoo though, or at least, no more so than 1v1 benchmarking is. After all in 1v1 SandboxDT fares better against the top bots than against the field at as a whole. All it means is that SandboxDT is optimized against the strong bots whereas Portia is more optimized against the lower to middle field bots. The same applies to 1v1 really, and the lesson is exactly the same: always test against a variety of bots. --Rednaxela 21:08, 3 November 2009 (UTC)
I understand, that "melee benchmarking is a bit like voodoo", but it better than nothing=) i think, that i will done at least first 2 tasks (gun and movement) and after it i will release my bot. About testing with different bots: spin bot and walls is my main tests=) --Jdev 07:07, 4 November 2009 (UTC)
