Chess engine X beats Y at one time control, but loses at a different time control, averaged over many games of course. Investigate what is going on.
The problem manifested in TCEC, wherein developers struggled to test whether a given change improved the engine's performance at long time controls.
Assuming source code to both engines, how can one create a hybrid that is as good as both?
One possibility for the difference between time controls is parallelization efficiency. Control for this by using only 1 core.
No comments :
Post a Comment