Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Rybka 1.01Beta 11 vs Rybka 1.01Beta10d Bullet results

Author: Vasik Rajlich

Date: 07:38:58 01/21/06

Go up one level in this thread


On January 20, 2006 at 10:53:03, Keith Hyams wrote:

>First of all - Please don't write back disparaging comments sbout the value of
>bullet games because you will only make me cry!
>
> Just because the 1 min per game + 1 sec per move is a stupid time limit for
>games  between humans it may have well have uses for engine testing.
>
>114 games properly controlled Rybka: 1.01Beta 11 scored +22=54-38. The match was
>stopped here because, under the above conditions, the chances of beta 11 being
>stronger is no more than 8%.
>
>The 32 bit .exe file was used. The results may be skewed because of the CPU, a
>Celeron 2.7GHz with 128kb l2 cache. Rybka runs like a 3 legged horse on these,
>possibly because the large size of .exe completely fills the cache and goes into
>slower RAM. Rybka1.01Beta11 is much larger than Beta10 and so this may be having
>an effect.
>                                   Keith

Keith,

can you please email me a few of the games which went deep into the endgame? I
hope that your notation keeps the depths searched.

There may well be some cache issues on certain computers in the transitions to
the endgame. I test only on Amd64 in 64 bit mode and did not see any serious
problems. You are also the first tester to report this type of result.

Vas




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.