|
From: | Hugh Sconyers |
Subject: | [Bug-gnubg] error rates etc |
Date: | Thu, 25 May 2006 09:39:26 -0700 |
hi address@hidden wrote:
the problem in gnubg's severe cube-rating may be related to the computation of the actual/close number of cube decisions. If the error is above very bad (something like 0.160 I think, it's hard-coded somewhere in the code source) the cube decision is not considered actual/close (hence the error per actual/close decision rate will be higher)... maybe the threshold should be higher ...
this might be the source of the problem. i think we should keep the move error computed using the divisor as "unforced moves". imo snowie makes a big error using all moves and this makes it harder to compare 2 match. btw can someone tell me what the definition of a 'forced move'?i have an idea for feature to the analyze matches and money session. when it finds an error in one of your plays, it then test both plays at another ply. so for example you play a match and analyze it at 2 ply. when gnu find you made an error it will test the 2 moves at 3 ply. of course you would want to make this an option.
regards hugh
[Prev in Thread] | Current Thread | [Next in Thread] |