Rybka 2, the world’s strongest chess program, leads all ranking lists clearly, with ratings well above Elo 3000. Now Rybka 3 is here, with a dramatic improvement of at least 80 Elo points.
Sean Evans wrote:Hi, Chessbase claims the following:
Rybka 2, the world’s strongest chess program, leads all ranking lists clearly, with ratings well above Elo 3000. Now Rybka 3 is here, with a dramatic improvement of at least 80 Elo points.
Is this accurate?
What is the above Elo 3000 based?
What is the highest Elo rating possible?
Thank you
Sean
Ofcourse the rating estimate is true. I'd guess Rybka 3 is > 100 elo better than the 2nd version...Can't believe that Chessbase would be modest once in a while
My guess is that Rybka 3 multi processor is anywhere around 3200-3300.
It's based on comp chess ratings list, ELO calculation program measured the rating based on the games it had played against the other stronger engines...It's nothing to do with human FIDE ratings list, but still looks more or less similar.
The total numbers level of engine rating lists is arbitrary, as calibrations with human chess federation ratings or with FIDE Elo ratings, are missing.
"Elo" is generated whenever the Elo system is applied, to whatever. It is not like a measuring tape which measures length, which is clearly defined. The Elo system has been applied even to sports like volleyball or football (soccer). Engine rating list's Elos have no mathematical or statistical relation to FIDE Elos.
I am not sure but I think there is no theoretical upper limit, for Elo ratings. In practice, we can keep in mind that 99% performance difference (points) mean +677 Elo.
Also, it is important to know that for example CCRL does not use the normal Elo system but Bayesian Elo, an improved method.
Sean Evans wrote:What is the above Elo 3000 based?
I don't know, the number 3000 is very arbitrary. They could have said 3200 and pointed to a computer rating list.
Ofcourse, I agree. It shows that chessbase doesn't care about computer chess ratings, engines etc, they are mainly interested in Human ratings, Grand Master Tournaments. If they know the real ratings of Rybka from their own tests, then they would rush in to make announcement.
