Suitable sets of test positions ?

Discussion of chess software programming and technical issues.

Moderator: Ras

govert
Posts: 270
Joined: Thu Jan 15, 2009 12:52 pm

Suitable sets of test positions ?

Post by govert »

I am looking for a few good sets of test positions, in order to compare a few toga-equivalent-strength engines. The tests should measure analysis strength, not blitz strength.

I was thinking of 3, possibly 4 main categories:
A) Tactical
B) Positional - Like Tony Hedlunds test suite
C) Endgame - EET?
D) Strategical - anything exists here?

I guess I need at least 100 positions per category in order to draw any conclusions.

Any suggestions of good sets?
Dann Corbit
Posts: 12867
Joined: Wed Mar 08, 2006 8:57 pm
Location: Redmond, WA USA

Re: Suitable sets of test positions ?

Post by Dann Corbit »

govert wrote:I am looking for a few good sets of test positions, in order to compare a few toga-equivalent-strength engines. The tests should measure analysis strength, not blitz strength.

I was thinking of 3, possibly 4 main categories:
A) Tactical
B) Positional - Like Tony Hedlunds test suite
C) Endgame - EET?
D) Strategical - anything exists here?

I guess I need at least 100 positions per category in order to draw any conclusions.

Any suggestions of good sets?
There are some test sets here:
http://cap.connx.com/EPD/

Nose around and take anything you like
govert
Posts: 270
Joined: Thu Jan 15, 2009 12:52 pm

Re: Suitable sets of test positions ?

Post by govert »

"some" is a modest understatement :)
I have already nosed around a bit, and I an thinking of using

speelman, & EET for endgames,
arasan10, and possibly "Brilliante Züge" for tactics,
positional.epd for positional.

Are these OK, meaning that they have the correct best moves/avoid moves?
swami
Posts: 6664
Joined: Thu Mar 09, 2006 4:21 am

Re: Suitable sets of test positions ?

Post by swami »

govert wrote:I am looking for a few good sets of test positions, in order to compare a few toga-equivalent-strength engines. The tests should measure analysis strength, not blitz strength.

I was thinking of 3, possibly 4 main categories:
A) Tactical
B) Positional - Like Tony Hedlunds test suite
C) Endgame - EET?
D) Strategical - anything exists here?

I guess I need at least 100 positions per category in order to draw any conclusions.

Any suggestions of good sets?
Please Try:
http://computerchessblogger.googlepages.com/sts

Chapter 1: Undermining is available for download now...It has 100 positions, all of them fully tested and accepted by comp chess community. Every month I and Dann plan to release new strategic chapters.
govert
Posts: 270
Joined: Thu Jan 15, 2009 12:52 pm

Re: Suitable sets of test positions ?

Post by govert »

swami wrote: Please Try:
http://computerchessblogger.googlepages.com/sts

Chapter 1: Undermining is available for download now...It has 100 positions, all of them fully tested and accepted by comp chess community. Every month I and Dann plan to release new strategic chapters.
Yes I have looked a bit at this one. Rybka2.2 came out on top, with Stockfish 1.0b as second. I will definitely include your suites in my testing.
Dann Corbit
Posts: 12867
Joined: Wed Mar 08, 2006 8:57 pm
Location: Redmond, WA USA

Re: Suitable sets of test positions ?

Post by Dann Corbit »

govert wrote:"some" is a modest understatement :)
I have already nosed around a bit, and I an thinking of using

speelman, & EET for endgames,
arasan10, and possibly "Brilliante Züge" for tactics,
positional.epd for positional.

Are these OK, meaning that they have the correct best moves/avoid moves?
I am not sure as to the depth of quality. I think that the Arasan test is well researched. I have done some testing on the brilliant move suite and the analysis will be found in the regular CCC forum as follow ups to Walter's posts.

I believe that Yuri Lyapko has done some testing with Speelman so you might send him an email or examine his results.