About bayeselo and white advantage

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

User avatar
Kempelen
Posts: 620
Joined: Fri Feb 08, 2008 10:44 am
Location: Madrid - Spain

About bayeselo and white advantage

Post by Kempelen »

I use bayeselo with default options to compute my engine testing tournaments performance. I am also using crafty.epd . As I see bayeselo is using 32.8 as ELO adventage for playing back. Do you thing it would be better for that kind of tournaments which has starting positions, to disable that adventage and get a estimation based of pure wins/draw/loses?

I dont have clear that that advantage would be useful when no opening is played.....
Fermin Serrano
Author of 'Rodin' engine
http://sites.google.com/site/clonfsp/
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: About bayeselo and white advantage

Post by bob »

Kempelen wrote:I use bayeselo with default options to compute my engine testing tournaments performance. I am also using crafty.epd . As I see bayeselo is using 32.8 as ELO adventage for playing back. Do you thing it would be better for that kind of tournaments which has starting positions, to disable that adventage and get a estimation based of pure wins/draw/loses?

I dont have clear that that advantage would be useful when no opening is played.....
Alternating colors for the same position will wash this out. And assuming your positions are reasonable, whether you play an opening or not shouldn't matter, since that advantage should apply to the entire game where white moves first?
Vinvin
Posts: 5287
Joined: Thu Mar 09, 2006 9:40 am
Full name: Vincent Lejeune

Re: About bayeselo and white advantage

Post by Vinvin »

Bob, with your enormous database of test, could you measure how big is the advantage for white with crafty (perf_with_white - perf_with_black) ?

Thanks !
Adam Hair
Posts: 3226
Joined: Wed May 06, 2009 10:31 pm
Location: Fuquay-Varina, North Carolina

Re: About bayeselo and white advantage

Post by Adam Hair »

You could determine advantage and drawelo from your tournaments
using advdist and drawdist functions in Bayeselo. Then use the computed
values in place of advantage (32.8) and drawelo (97.3).
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: About bayeselo and white advantage

Post by bob »

Vinvin wrote:Bob, with your enormous database of test, could you measure how big is the advantage for white with crafty (perf_with_white - perf_with_black) ?

Thanks !
I could try, but I am not sure the result would be valid. Remember, my positions are chosen from late opening positions, and when I found anything with a gross imbalance, I removed it. Also, I can't guarantee that the positions I use occur with any similarity to the frequency such positions would occur in real games. Some openings might be under-represented, or some over.
Vinvin
Posts: 5287
Joined: Thu Mar 09, 2006 9:40 am
Full name: Vincent Lejeune

Re: About bayeselo and white advantage

Post by Vinvin »

In fact, you're right, "white or black" doesn't means anything in this situation !
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: About bayeselo and white advantage

Post by bob »

Vinvin wrote:In fact, you're right, "white or black" doesn't means anything in this situation !
There are other issues as well. If you play A vs B and alternate colors, it is still hard to understand the "white advantage" when A might simply be much stronger and win every game with white, and then every game with black as the colors are alternated.