Author: Robert Hyatt
Date: 06:15:26 12/18/99
Go up one level in this thread
On December 18, 1999 at 06:09:56, Tom Kerrigan wrote: >On December 17, 1999 at 23:55:21, Dann Corbit wrote: >>In particular, I am wondering if there is any computable curve of probability >>for narrowing the candidates. I think (perhaps) this could be used for special >>extension searching. Kind of like null move, but instead after we pass some >>statistical test, we consider only 5 moves. Then after another test is passed, >>we study only 3 moves, finally, for the last ply or so we might look at only two >>candidates. > >I think this is a very bad idea because it eliminates the possibility of finding >a terrific, obscure tactical win. > >However, your observation got me thinking... Let's say a program is oscillating >between move A and move B. It probably doesn't matter which one is picked. Now, >what if you could bias the search so B is significantly worse (0.3 pawns?) than >A? Then the program doesn't waste time by changing PVs often, but if B becomes >significantly stronger than A, the program won't ignore it... > >Ideas? > >-Tom This has already been done. To prevent the oscillation you mention. You can simply 'fudge' up the score for the first move, and adjust alpha/beta to match, or you can leave the score for the first move alone, but search all remaining moves with score+N, score+N+1. I tried it many years ago, but didn't like it. If you believe your eval, then you believe that +.05 is worth something...
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.