SEE benchmark/test suite

Discussion of chess software programming and technical issues.

Moderator: Ras

osvitashev
Posts: 13
Joined: Tue Sep 07, 2021 6:17 pm
Full name: Alex S

SEE benchmark/test suite

Post by osvitashev »

Hey. I am looking for a benchmark and/or a test suite for static exchange evaluation. Do such things even exist?
I realize that SEE is not supposed to be accurate, because of some of the more esoteric possible scenarios.
And yet, it seems that having a way for comparing accuracy of SEE algorithms would be a useful thing.
User avatar
algerbrex
Posts: 608
Joined: Sun May 30, 2021 5:03 am
Location: United States
Full name: Christian Dean

Re: SEE benchmark/test suite

Post by algerbrex »

osvitashev wrote: Thu Nov 11, 2021 3:56 pm Hey. I am looking for a benchmark and/or a test suite for static exchange evaluation. Do such things even exist?
I realize that SEE is not supposed to be accurate, because of some of the more esoteric possible scenarios.
And yet, it seems that having a way for comparing accuracy of SEE algorithms would be a useful thing.
I've seen several test suites in different engines. Arasan's author Jon Dart has some positions listed here: https://github.com/jdart1/arasan-chess/ ... t.cpp#L163. Although your SEE implementation may differ from his depending on how you implement promotions. So you may have to check some of the positions yourself and modify them to match the correct score for your conception of SEE.