Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Is computers winning more with black? ( in Comp. vs Comp )

Author: Don Dailey

Date: 09:22:30 07/29/98

Go up one level in this thread


On July 29, 1998 at 09:06:40, Tim Mirabile wrote:

>On July 28, 1998 at 10:13:25, Don Dailey wrote:
>
>>I played a bunch of games between the latest Crafty and a 10 year old
>>PC program of mine I ported to Unix using the xboard interface.  The
>>main thing I noticed was that having WHITE was a big advantage.  When
>>a program won or drew as black it was a big deal and was quite an
>>advantage, just like in grandmaster matches.
>
>Having the first move is always good in chess, regardless who the players are.
>But we might find that computers lack the color bias that some strong players
>seem to have.  For example, you can find a lot of games where Black came out of
>the opening with a slight edge, and then agreed to a draw, while White would
>certainly have pressed on with a smaller advantage.  Players also often play
>more conservatively with Black in the opening, but in this case the computers
>are using opening books based on human play, so they get this effect built in.
>
>John Nunn gave an example of this in his recent book:  Set up a chessboard, and
>play 1.c3 e5 2.c4 Nf6.  Now who would want to play a move like 3.d3 in this
>position?  Yet it is the most common reply when the colors are reversed.


I'm pretty sure humans have a psychological color bias.  I took a large
database of random positions from master games and grouped them by
score.  I plotted them by results.  So one axis is score (according
to Cilkchess) and the other is actually win percentage in the master
games.  This ended up well centered around zero, a zero score meant
about a 50/50 chance of winning.   I tried the same experiment where
I considered this data from WHITE'S point of view only and found that
white had slightly better winning chances even when Cilkchess saw
it as being somewhat negative.  I followed this up with looking at
everything from blacks point of view with the same search depth to
eliminate odd/even scoring problems.   The result was consistant.
I cannot explain this with any kind of proof but one hypothesis I
formed was that masters have higher expectations of losing or drawing
as black.  For instance if your frame of mind when you set down to play
with black is to get a draw, you may be content with a draw and not look
for winning opportunities.  You may see a draw line and jump on it, not
noticing a win.   Or you might feel "more obligated" to win with white.

I really believe that humans tend to perform to their own expectation
levels, or at least have a some bias in this direction.

- Don



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.