Computer Chess Club Archives


Search

Terms

Messages

Subject: Show it's predictive power.

Author: Roger

Date: 10:24:32 12/10/99

Go up one level in this thread


I don't see any problem with recalibrating the SSDF, though doing it with as
many of the top programs as possible and with as large a sample of games as
possible would be recommended.

Whether it should be done, or not, can be settled empirically. Differences
between the ELOs of two players translate into probabilities of winning and
losing. Accordingly, it ought to be possible to develop the recalibrated ratings
using a pool of, say, twenty games of three programs at 40/120, and then how
that these new ratings predict the pattrern of wins and losses better than the
old ratings.

No one can argue with that. And until it can be shown that the new ratings are
better able to predict that the old ones, then I'd just leave it alone.

Recalibration is inevitable, because GM-computer games will become much more
common.

Roger



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.