Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: tbs statistics, and further idea to investigate ?

Author: Anthony Cozzie

Date: 05:05:02 12/08/03

Go up one level in this thread


On December 07, 2003 at 20:47:17, Georg v. Zimmermann wrote:

>Anthony wrote this little tool to get information about the winning percentage
>in 6men tableB.
>
>
>knnpkq 28.24%
>kbbpkq 28.27%
>kbnnkq 35.12%
>kbbbkq 35.17%
>kbbnkq 35.76%
>knnnkq 37.74%
>krnpkq 44.44%
>krbpkq 48.83%
>kbnknn 49.39%
>kbbkbb 50.00%
>knnknn 50.00%
>kqqkqq 50.00%
>krrkrr 50.00%
>krnnkq 50.81%
>kqbkqn 52.02%
>[...]
>
>I find this is so interesting it deserved another new threat :-)
>
>Right now I am wondering how many of the positions are won because simply a
>piece is hanging, or not won because a piece is hanging, or won because there is
>an accidential mate ...
>
>This should be a different % for every combination of pieces.
>How it could be tested is obvious.
>
>Another interesting test would be to see how difficult these types are for
>programs.
>One could see from say a random selection of 1000 positions how well
>non-TB-Crafty scores against TB-Crafty. I bet in knnknn for example it does so
>well that we see we do not really need that tbs ...
>
>Georg

Actually, I was thinking about this last night, and I think the most interesting
 stat would be to run through Crafty's enormous.pgn, and find which TBs occur
most frequently.  Maybe use some function of appearance rate & easiness of
winning to select which 6-men to use . . .

anthony



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.