Author: Robert Hyatt
Date: 03:28:45 08/08/98
Go up one level in this thread
On August 07, 1998 at 18:43:12, James B. Shearer wrote: >On August 07, 1998 at 05:18:01, Robert Hyatt wrote (in part): > >>I limit Crafty's choice of opponents to those that will actually have a chance >>of beating it, so that I can learn something about what it is doing wrong. >>There are *far* more 1500 players than 2400 players, and they would certainly >>keep it more busy. But it will roll them up into little balls quite easily, >>which doesn't lead me to finding problems... > > Suppose we assume that you only learn from Crafty's losses. You are >no doubt correct that allowing 1300-1500 rated players to play Crafty would >not produce more Crafty losses since the small chance such players will win >will be outweighed by the chance they will displace a higher rated opponent. >However the case of 2200-2400 rated players is less clear. Such players >should have a 1-10% chance of beating Crafty. This is not terrific but if >even the best humans win less than 25% of the time it is not too bad. You >might get more losses allowing such players to play crafty particularly if >you could figure out some way to give priority to higher rated players. > Btw you said no grandmaster scores even 25% against crafty. This >suggests some do considerably worse. I realize you don't want to name names >but could you give some more details about how crafty has been scoring >against humans in blitz games. > James B. Shearer I actually will probably make this change. I did the >2400 formula back when crafty was playing non-stop, as that raised the quality of the opponents. But at present, with an ever-increasing number of crafty clones on ICC, playing 2200-2400 players is better than not playing at all, of course...
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.