Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Shredder 8 vs. Fritz 8 (8.0.0.23) Blitz 3 0, 200 game test

Author: George Sobala

Date: 23:39:25 01/31/04

Go up one level in this thread


On January 30, 2004 at 22:08:18, James T. Walker wrote:

>On January 30, 2004 at 12:01:23, George Sobala wrote:
>
>>My conclusion is pretty much like yours - when two engines are even only
>>approximately of similar strength, the randomness of the play caused by the very
>>short time control is far greater than the true difference in strengths. You can
>>measure out to 1000000 games, and all you do is measure the randomness of blitz
>>play.
>>
>>Far better to play fewer but longer games.
>
>If you flip a coin 200 times and get heads 101 times and tails 99 times would
>you conclude that it was because you didn't flip the coin high enough into the
>air?  In my opinion your conclusion about randomness caused by short time
>controls is the same thing.  Two programs of approximately equal strength will
>naturally have very close results.  I feel certain however if you run a million
>games you will have a very good answer as to which program is stronger between
>Fritz8/Shredder8.  Programs today on modern hardware are looking ahead about 10
>ply minimum with extensions in some cases another 20/30 ply at blitz time
>controls.  I would not call this random play.  Programs today play better chess
>at 5/0 than 99% of human chess players playing at 40/2 hours.
>By the way, in my blitz database, Shredder 8 is only 13 Elo ahead of Shredder
>7.04 and 19 Elo ahead of Fritz 8.
>Jim

You are wrong because a chess game is not a coin flip. In a series of 3 minute
blitz games, a fair proportion of the games are decided in a mad end-game time
scramble where the engines are searching very few ply indeed. This is the
"randomness" element that does _not_ reflect on the true strength of the
engines.



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.