Author: Robert Hyatt
Date: 20:57:56 01/22/00
Go up one level in this thread
On January 22, 2000 at 22:04:19, Len Eisner wrote: >On January 22, 2000 at 20:23:42, Robert Hyatt wrote: > >>On January 22, 2000 at 20:19:19, Robert Hyatt wrote: >> >> >>There is another way to handle this that might be better against >>humans: >> >>Do a minimax search to some depth, but probe the tablebases at each ply. >>Make the static eval return something like N-sum(opponent's non-losing >>moves). The idea is that you would like to follow branches where the >>opponent has only one non-losing move at several points, hoping for one >>blunder to lose the game for him. >> >>I had code written for this, but never really finished testing it as the >>current approach is simpler... > >I like it . . . > >In playing through 5-piece endings in GM games, I noticed that even GM?s don?t >always find the single move that draws in those endings where only one move >holds the game. It is incredibly difficult to find those moves without a >tablebase. > >Len Sometimes yes, sometimes no. IE sometimes the only move is very obvious. Which is a real problem... ie you _could_ reach a point where you choose between (a) and (b). (a) has more positions where there is only one move that holds, but to a human they are very obvious. (b) has fewer positions that hold, but they are much harder to find. The problem is detecting which is which. IE which kind of positions are hard for the human to find the one right move in, and which are easy. That isn't something that I think is very easy, yet...
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.