Author: Mark Ryan
Date: 15:21:07 07/09/99
Fritz 5.16 Autoanalysis and Blundercheck fail to detect a bad move in a famous position (details below). Using backward analysis, starting by evaluating the final game position and then working back through the preceding moves, Fritz does not remember its evaluations from move to move; it analyzes each move without reference to its evaluation of the following move. Therefore: 1. What is the point of backward analysis? It seems to be no different from on-the-spot position analysis. 2. Are there any programs that use backward analysis "correctly"? That is, do any programs evaluate the final position first, then analyze the preceding moves using the knowledge that they have gained? Example: World Championship 1972, Game One, in which the white pieces were handled by the brilliant World Champion Boris Spassky, and the black pieces were handled by some other guy. Fritz correctly evaluates White as being ahead all the way back from the final move to move 30, at which point it sees that black's bishop is doomed. However, when Fritz moves back to move 29, it forgets this knowledge, and it does not see that the bishop is trapped; so it incorrectly evaluates the position.
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.