Moderator: Ras
Not using TCEC data. But the idea of using the same set of openings to test each version is standard testing methodology. The hard part is hardware. Not all games get played on the same hardware, which makes comparing your data where they do use exactly the same hardware a bit different.Ferdy wrote:So you have the results of your previous public released version.
CEGT 40/20.
What I plan is get the games of those matches, extract the opening, then use those opening to test your development version. Openings, opponents, TC and color allocations should be similar as in the previous released version.
Has anyone tried this before?
Right, the only thing that I can do here is to get the equivalent benchmark for my hardware.bob wrote: Not using TCEC data. But the idea of using the same set of openings to test each version is standard testing methodology. The hard part is hardware. Not all games get played on the same hardware, which makes comparing your data where they do use exactly the same hardware a bit different.