Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Evaluation Should Be Winning Probability - Not Pawns

Author: Severi Salminen

Date: 10:30:34 01/16/02

Go up one level in this thread


>It has occurred to me that it is wrong to evaluate a position in terms of
>relative pawns (the "de facto" standard - whereby an evaluation of 2 means that
>you're approximately the equivalent of 2 pawns ahead).
>
>This means that many aspects of evaluation have to be squeezed into a dimension
>which is not appropriate at all.
>
>A better way would be to evaluate "winning probability". If a position was a
>draw, the value would be 0.50 (or 50%). If the player should win 3 out of 4
>times, the eval should be 75%. If the player must win from here, then the
>evaluation should be 100%.
>
>It seems strange when you think about it that all programmers have chosen to
>adopt the traditional "pawn equivalence" standard.

Not at all strange. What's the difference if engine sees move A=1.5 pawns and
move B=-3.6 panwns, or move A=0.76 and move B=0.06?? And in fact you can change
your favorite engine (if source code is available) to scale the score range to
anything you want. In my engine matescore is +MAX_INT, which is something like
2'140'000'000, you can divede all score by 2*MAX_INT and then add 0.5, to get
the desired range, if it makes you happy (if my sources were available...). Or
you can translate it non-linearly (logarithmic or whatever) but the result will
be the same: engine pics the move with highest score, no matter if the scale is
linear or not. Now we can use a scale we don't have to transale at all: it's
fast and clear.

Severi



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.