Computer Chess Club Archives


Search

Terms

Messages

Subject: an idea for an experiment about rating

Author: Uri Blass

Date: 16:58:17 04/06/05


take a chess program.

Your target is to find the difference in  rating between the program and a
program that plays random moves.

Call the original program X_0
You change X0 to play random moves in n% of the cases and call it X_n

play a match of 1000 games between X_0 and X_1
play a match of 1000 games between X_1 and X_2
...
play a match of 1000 games between X_99 and X_100

total number of games 100,000 games

Calculate rating for X_0,....X_100 based on the result of the games.

1)What will be the difference between X_0 and X_100 in rating?
2)Is the difference between X_0 and X_100 bigger when X_0 is stronger?

I suggested 1000 games in order to get more significant result but you can start
only with 10 games if you are interested to see fast results and repeat the
experiment later.

In order to get many games you can decide about fixed small depth such that
games will be finished fast.

You can take Crafty at fixed depth of 2 and later Crafty with fixed depth of 4
and it may be interesting if Crafty with fixed depth of 4 will have bigger
difference between X_0 and X_100.



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.