RE: [Bug-gnubg] Evaluation inconsistency

[Nis Jorgensen]

On Tue, 16 Sep 2003, Albert Silver wrote:

>
> > It would be great if you posted the position - even if you don't think
> > it is relevant to the problem you are facing.
>
> Naturally it is no problem to show the position, but as I said, it is
> completely unimportant whether my decision was wrong or right.

Well, the "unimportance" is based on your assumption that this is an
inconsistency in gnubg - rather than just a badly understood position,
which is evaluated quite differently by eval and rollout.

> What is important is the difference in the size of errors from (-0.007)
> to (-0.115) when *everything* in the percentages is improved:

Did you read the rest of my reply (which I apparently sent to you only AND
did not save in my sent-mail folder, so I cannot quote it)?

The numbers that are "improving" are the CUBELESS numbers. The numbers
used to judge your decision are the CUBEFUL numbers. The rollout estimates a
much larger value of cube possession than does the evaluation.

Nis