Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Objective method of emulating players and determining relative strength

Author: Walter Koroljow

Date: 17:41:15 08/11/99

Go up one level in this thread


Christopher,

I think you have a wonderful idea in principle.

It seems to me that the issue is: who has the time to adjust the "constants" to
achieve a high correlation with a past master.  This is a computational issue
since we would have to evaluate the effect of the constants on the correlation
over a substantial fraction of the master's entire output.  In optimizing a set
of constants,  we would have to evaluate multiple sets of constants, etc. This
implies multiple computer reanalysis of the body of the master's games.  Very
time consuming today.

This is probably achievable in a few years.  It took me forty days and forty
nights (amazing coincidence!)  to analyze (with a P233 mmx) 80 games I had
played long ago.  Application of Moore's Law says that in 20 years this would
take several minutes.  Therefore, I expect your program to be carried out
ultimately, but I do not expect it today.

I actually believe (!) that this will ultimately be done.

By the way, thank you for the advice on instructing youngsters.  I am trying
what I can of it.

Cheers,

Walter





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.