Naum 3 FRC tests - 1,000 games - and web site updated

[Spock]

Web site now updated

I prefer to look at the pure list myself
http://www.computerchess.org.uk/ccrl/40 ... _pure.html

Although you need to look at the other lists to see Rybka.

Very consistent performance from Naum 3

Thanks for the new engine Alex

[Tony Thomas]

Ray, did you notice that I gave step by step directions for using Bayesian elo? I guess Alex probably took the number 60 after looking at your FRC list.

[Spock]

Ray, did you notice that I gave step by step directions for using Bayesian elo? I guess Alex probably took the number 60 after looking at your FRC list.

Yes, thanks Tony I saw it , haven't tried yet

[Spock]

I guess Alex probably took the number 60 after looking at your FRC list.

Possibly. The figure to look at will be CEGT 40/20 which should have hundreds of games at standard chess by now. Could be they have done a little better, don't know. Our standard blitz list has nearly +80 ELO so far for 4CPU, but not nearly enough matches or range of opponents to say that's reliable

[Tony Thomas]

I guess Alex probably took the number 60 after looking at your FRC list.

Possibly. The figure to look at will be CEGT 40/20 which should have hundreds of games at standard chess by now. Could be they have done a little better, don't know. Our standard blitz list has nearly +80 ELO so far for 4CPU, but not nearly enough matches or range of opponents to say that's reliable

Is it possible for you guys to obtain say a similar number of matches between the same opponents that played Naum 2.2. For example if Naum 2.2 played 22 matches versus Rybka 2.3.2, then play that many games, and no more. I guess that would increase the accuracy of the list at least 1.1 fold.