Talk:Thorn

From Robowiki
Revision as of 12:00, 11 October 2009 by Skilgannon (talk | contribs) (→‎MicroBot throne: rerelease, random offsets and codesize)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

MicroBot throne

Just by a hair =), but congrats on retaking the MicroBot throne! You've really made an impressive return, great work. --Voidious 14:10, 8 October 2009 (UTC)

Doh! So much for that hair. :-/ Good luck in your throne hunt(s)... --Voidious 02:33, 9 October 2009 (UTC)
If you have MiniBot rated 2nd too, you will be one that got 2nd in three rumbles, while Skilgannon got 1st in the same rumbles =) --Nat Pavasant 02:51, 9 October 2009 (UTC)
Heh, I think that must have been a record for the shortest time holding a roborumble throne. Maybe later I'll try to make a competitive (i.e. wave surfing) mini, but for now I think I'll stick to tinkering around with Thorn. --Kev 03:48, 9 October 2009 (UTC)
I think my time with the MiniBot throne earlier this year might've been even less. I updated Komarious to overtake WeeksOnEnd by a hair, but overnight while I slept and my RR clients ran, CunobelinDC came along to crush us both! I had to laugh, though, as it was one of the most bizarre Robocode experiences I ever had. --Voidious 03:54, 9 October 2009 (UTC)
And both time, it is Skilgannon! Blame him. :P --Nat Pavasant 04:48, 9 October 2009 (UTC)
I couldn't have my triple-throne status being taken! =) Necessity is the mother of invention. It's interesting to note that +-1 pixel of randomness is approximately 0.5% more accurate than a random distribution between +-4. --Skilgannon 10:06, 9 October 2009 (UTC)
Precision is everything, even when polluted. I use +/- 0.6 pixel and even that is good enough. --GrubbmGait 09:21, 10 October 2009 (UTC)
I was so surprised by the gain in points I downloaded Toorkild to see if Skilgannon had changed anything else! But it seems like bad battles might have inflated the rating? Anyway, instead of spending codesize in picking a random pixel length, Thorn just adds the same tiny offset to its firing angle every time it fires (0.0005 radians right now), which seems to work fine against BulletCatcher without losing much precision. --Kev 18:12, 10 October 2009 (UTC)
I also tried 0.3 pixel (which is 0.0005 radians at 600 distance), but around half of the bullets got catched. Choosing a fixed angle-offset varies the deviation by distance, but for codesize reasons I understand (and agree) with your solution. --GrubbmGait 18:58, 10 October 2009 (UTC)
I've re-released, we'll see what happens. And the reason I used a random offset is because of my arms-race with Rsim, if I used a fixed offset it would be possible to detect that and predict it for releases, whereas a random offset makes this impossible. And about the codesize - Toorkild is sitting on 733 bytes right now, so it's not really an issue =) --Skilgannon 11:00, 11 October 2009 (UTC)