Author: Dann Corbit
Date: 14:11:17 05/15/00
Go up one level in this thread
On May 14, 2000 at 17:02:53, Gian-Carlo Pascutto wrote: >>Which brings up a point... >>I have seen some chess engines do something silly. The have finished a ply >>(say, for instance, ply 14 side to move) and do a fail high. Then *instead* of >>choosing the move that failed high, they stick with the one that was >>completed!!! {in this case, the engine might still play axb5 -- though I doubt >>that Amy has this fault}. >> >>Why would anyone want to do that? It does not make any sense to me. > >My program Sjeng being one of those, I can give at least two explanations: > >a) It's a misfeature which I am/was too lazy to fix. > >b) Because of the goofed up search the program will sometimes discover something >new on the next ply and start failing high on a LOT of moves. If I can't >complete the search in time, I rather trust the last ply's result than >one of the new psuedo-good moves. I think this would make a very interesting heuristic study. Find out for fail high what the probability is that you should choose that move instead of the completed one.
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.