[Bug-gnubg] Deterministic vs non-determinstic noise
From:
Bern Hamilton
Subject:
[Bug-gnubg] Deterministic vs non-determinstic noise
Date:
Wed, 16 Sep 2009 17:22:18 +0300
I let two gnubg players play each other in a long money session and also in a few 64 point matches. Both played with 0-ply with 0.040 noise, with one having deterministic noise and the other non-deterministic.
Then, using 0-ply 0.040 deterministic noise analysis settings, I analyzed all the moves and cube decisions. I expected the player with deterministic noise to have 0 error rating in both, chequer and cube play, and the one with non-deterministic noise, some chequer and cube errors.
To my surprise, both had perfect chequer play and both had errors in cube decisions.
So, two questions:
Why didn't the non-deterministic player produce errors in chequer play? Why did the deterministic player produce errors in cube play?
Maybe I'm not understanding what the random noise is intended to achieve.