Steve B wrote:spacious_mind wrote:
Larry Kaufmann's 1995 rating is the CCR:
I guess he was wrong as well. 49 points better.
Kaufmans lists are not wrong
they are simply not intended to show a rating at long time controls
while the SS lists and SSDF lists were
the CCR lists are intended to show a blended rating culled from other rating lists and combining them with results of games with faster time controls
His ultimate rating is a mean of different ratings ...
one of the columns is from Selective search(CCNS)
PLY is the SSDF
the 30' column is a rating obtained in 30 sec /avg games
i think Kaufmans approach was an interesting one but again never intended to rate computers at the longer time controls
Min N Match Regards
Steve
Ok lets ask the question differently. You do agree that Portorose has 1024 KB RAM and Vancouver, Lyon and London have 2048 KB RAM.
At Schachcomputer.Info:
TM Vancouver = ELO 2333
TM Portorose = ELO 2271
That's a difference of 62 ELO. The speed doubling factor between 30 seconds and 180 seconds is a factor of 2.5. Meaning Portorose would have to outperform Vancouver at tournament level by 62 + 18 = 80 ELO points or 32 ELO per doubling.
That would also mean that Vancouver would actually stay stagnant and improve by nothing its strength of play while Portorose performs fantastically. Does that sound realistic? When in fact Vancouver has double the RAM which should naturally mean that at a longer time period it should improve its performance when compared to Portorose. It would also mean that two versions and 2 years later Lang brings out a program that has gone backwards?
The same argument cannot be said for London, Vancouver and Lyon as they all have the same hash therefore I would not necessarily fight it if Lyon were for example stronger than Vancouver. With London we know that Lang tried to build into it more intelligence but at the same time made its style more passive so it is hard to say what improvements there are. They are not seen at 30 seconds per move so clearly but maybe at 3 minutes they kick in therefore you don't see me arguing with you here if London proves to be stronger than even Genius.
The ratings at Schachcomputer.Info come from games where with Portorose I played probably 75% of those games and with Vancouver more than 50% of those games as most of them stem from the Revelation Tournament.
The conditions of the Revelation tournament were 32 top rated programs each played each other 2 twice, meaning that every single opponent played the same opponent resulted in a TM Vancouver rating of ELO 2363 and TM Portorose rating of ELO 2268.
With that same logic of hash difference, TM Portorose would have to catch up 95 ELO points at tournament level against these same common top flight opponents or 38 Points per doubling. And that is not going to happen with its inferior hash.
I must have played 150+ plus hours of chess with each of these two computers or if you wish one month of work for each of them so therefore I seriously doubt that anyone knows their playing abilities better than I do, because I have done the labor and not just assumed from some lists.
Best regards
