Earlier this week I posted the results of a 200 game tournament between
Rybka and Shredder with the Arena mainbook.abk openning book :
Rybka1.1(32bit) 108wins, Shredder8 30 wins, 62 ties.

Today, I just finished another 400 games between Rybka1.1, Rybka1.0,
and Shredder with the d-dahlm-v01__draw.abk openning book:

Shredder8 - Rybka v1.1.w32 38-112-50 32% -131
Shredder8 - Rybka v1.0.w32 60-100-40 40% -70

According to Arena, this corresponds to rating difference of about 131
points! (There is still a lot of uncertainty with 200 games. I
think the uncertainty is about + or - 25 points.) The large difference
in ratings might have been caused by the tournament conditions:

Blitz: 2 min + 1 second per move,
Rybka 79 MB
Shredder 87 MB
Windows XP, AMD 3200, 512 MB ram total.
Opening book: d-dahlm-v01__draw.abk
Ponder Off, Adjudicate games after 250 moves.
I'm not sure if any endgame tablebases were used by Shredder. Rybka
did not use any table bases.

Another factor to take into account:

A computer program can be programmed to play specifically well against
other computers - sometimes at the cost of less good play against humans.
But commercial engines like Fritz and Shredder have other needs to satisfy
their human owners in addition to that: they must be able to play well
against human grandmasters, they must have a variety of playing styles, a
nice plaing interface, a good analysis interface, ability to play on the
web, and so on.

True, the GUI takes a number of these factors into account (the Fritz GUI
is used by other programs such as Shredder and Junior), but not all,
especially the neeed for a variety of playing styles and the ability to
give a reasonable imitation of human play.

So although it may be true that Rybka is stronger than Fritz, it remanins
to be see if this is due only to taking into account positional factors.
Other programs like Junior use more positional knowledge than Fritz which
results in a slightly different playing style but a questionable increase
in strength.

Anyway the performance of Rybka raises interesting issues and the result
should be an all-around improvement of computer chess programs.

Henri

H. Arsenault wrote:
But commercial engines like Fritz and Shredder have other needs to satisfy
their human owners in addition to that: they must be able to play well
against human grandmasters, they must have a variety of playing styles, a
nice plaing interface, a good analysis interface, ability to play on the web, and so on.

Henri,

It may be the case the Rybka was specifically designed to play
against computer players and, perhaps, it is not better than Fritz or
Shredder against humans. I think it would be very difficult to test
this hypothesis because only a few highly rated humans are able to
compete at this level and it would be hard to get them to play enough
games to reach a statistically significant result.

Cheers,
Irchans