Talk:TormentedAngel

Hey, just a suggestion. Try only dodging the bullets that hit you. This should improve your score dramatically. This is what all the top bots do against 95% of the rumble competitors. Only for the very best bots will you want to account for the waves that don't hit, just to throw their stats out so they can't predict where you'll be. --Skilgannon 11:14, 17 May 2010 (UTC)

Hey, I noticed you put ELO rating in the infobox. ELO is in a hopeless downward drift on Darkcanuck's server - has been for a while and we're not sure why. Most of us just use APS these days, or Glicko2 if we really want a voodoo style rating. =) --Voidious 13:25, 17 May 2010 (UTC)

Well, ELO is something I always used. Is Glicko2 really better? I know nothing about it... But if everyone on the wiki uses it, then I'll just comply. --Lrem 13:41, 17 May 2010 (UTC)

Glicko2 is also in hopeless downward on Darkanuck's server. APS is the best now, I think. --Nat Pavasant 03:03, 18 May 2010 (UTC)

Well, if you don't like the popular rating systems, it's time to design your own. I think that everyone could agree on some simple formula like comparing your ELO to some known landmark. Or in other words just normalize it to assure that SandboxDT has a constant score ;) --Lrem 07:26, 18 May 2010 (UTC)

Is there any smart way of getting backtraces of my bot against random rumble opponents? I've seen score of the newest TA against some bots going from over 80% to under 5%, so I guess there is some serious error. But I can't see any reason for it in new code and it didn't surface anytime I was watching... --Lrem 11:52, 20 May 2010 (UTC)