Author: James T. Walker
Date: 19:04:29 12/22/99
Go up one level in this thread
On December 22, 1999 at 20:10:03, Dann Corbit wrote: >On December 22, 1999 at 20:05:34, John Warfield wrote: > >>Hi >> >> I think this is a nice game, Rebel actually outplays CM6555 tactically! >>However CM6555 won the match 4-2. I thought this score was surprising since >>Cm6555 ran on hardware much faster, I expected a six to 0 score. Does anyone >>know what the rating difference would be based on these six games? I know it is >>to few games but what would a 4-2 score suggest? 70 Point difference? >With that number of games played, there is no way to predict any ELO. It's just >not mathematically sound. > >However, it takes an ELO difference of just over 100 points to gather 2/3 of the >expected points. Hence, this result is (I think -- not mathematically but just >'seat of the pants') that the result is about right. >[snip] Well mathematically it is about right but in actual practice I would not bet a wooden nickel that this ELO difference would hold up after a few hundred games. You would have to be very lucky to run a 6 game match and come up with a realistic ELO difference between the two programs/systems. In fact there is also a fairly high probability that the ELO could be reversed after a few more games. I have seen Fritz 6 score 9-1 vs Hiarcs 7.32 in a 10 game set. I can guarantee that Fritz is not that strong. I'm sure you have seen the same thing if you have run many games like that. Jim Walker
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.