Author: John Merlino
Date: 16:28:52 11/01/01
Go up one level in this thread
On November 01, 2001 at 18:29:16, Thom Perry wrote: >On November 01, 2001 at 15:27:23, J. C. Boco wrote: > >>On November 01, 2001 at 13:15:14, John Merlino wrote: >> >>>On November 01, 2001 at 10:53:58, Kevin Stafford wrote: >>> >>>>I'd be willing to bet that chessmaster has poured the most time and resources >>>>into attempting to make their program accurate with repspect to its Elo ratings, >>>>but I doubt that even they are very accurate. I'm not sure how possible this is >>>>given that the program has to run on so many different levels of hardware. I'm >>>>afraid the only accurate way to determine your rating is to do it the hard way; >>>>earn it in a club or tournament by playing rated games. >>>> >>>>-Kevin >>> >>>I can clarify this a bit. Yes, we have poured quite a bit of time and resources >>>into the problem. And, yes, given all the different hardware and personality and >>>time control combinations, it is pretty much impossible to be reasonably >>>accurate with respect to, for example, a USCF rating. >>> >>>However, I can say with reasonable certainty that if you play enough (i.e. at >>>least 50-100) rated games on Chessmaster, at ONE SPECIFIC time control and on >>>the same machine, then your rating will probably be within 100 points of your >>>supposed USCF rating. >>> >>>That's just my educated opinion, >>> >>>jm >> >>Good point. As long as you play the same time controls and don't upgrade your >>computer, the personalities will be stable. It seems reasonable that even >>though a personality is rated 1500, if you got a faster processor it (the >>personality) would be stronger. > >I guess it depends on how the program limits itself to reduce ELO playing >strength. For example, if 1500 equates to a fixed search of ply = 3, then it >wouldn't matter which hardware were used. But if 1500 = 30 seconds/move then, >yes, the hardware would be a factor and the rating would depend on the speed of >the CPU system. That's definitely true. The program knows not to increase the ratings of "fixed depth" personalities, regardless of the CPU speed. However, it also attempts to adjust how many "bonus points" are given, depending on the personality's "Strength of Play" value as well. In other words, stupid personalities don't get the same bonus due to CPU speed that GMs get. jm
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.