Author: David Eppstein
Date: 11:30:49 08/17/99
Go up one level in this thread
On August 16, 1999 at 21:28:56, Robert Hyatt wrote: >On August 16, 1999 at 19:03:23, Jesus de la Villa wrote: >>This "Non-drawish style" must be rewarded in the score, for example: >>A win scores 1.5 >>A draw scores .5 >>A lose scores 0 > >This has been discussed many times. But in general, no one really wants >to change. The issue is usually brought up by "GM draws"... Along with inertia, there is also a point here related to computer chess. In order to use alpha-beta, it must be the case that the evaluations are treated as a zero-sum game, so that a line that is good for you should be equally bad for your opponent. The scores above are not symmetric in this way, so it would be difficult to reflect that in an evaluation. Note that this does not mean that the evaluation should always be zero for a draw. It would be perfectly reasonable to have an evaluation that estimates the probabilities of a win/loss/draw from a given position, and uses that to calculate the expected number of rating points gained or lost by playing from the position, for instance -- this would only give a draw score of zero when playing someone with an equal rating. But this depends on having a rating system that acts as a zero-sum game.
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.