Author: Jay Scott
Date: 12:19:26 05/22/98
Go up one level in this thread
On May 21, 1998 at 20:11:57, Don Dailey wrote: >Some other details. The pawns were fixed at 100 points as a reference >point. It's amazing how often this "obvious" trick is overlooked. It helps convergence by stabilizing the meanings of the other values and eliminating a meaningless degree of freedom. >The real problem as I see it is that there is a lot of imprecision in >the actual fitness function. Any given individual can get seriously >underrated or overrated by having a little bad or good luck. I've always liked the idea of keeping "error bars" for the fitness measurements, as in Glicko ratings. Then you have the weight of statistics behind your decisions. :-) You can do things like play games until you can show that the worst player is significantly worse than the best one, and only then kill it off. Jay
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.