Author: Dann Corbit
Date: 17:16:53 05/01/00
Go up one level in this thread
On May 01, 2000 at 20:06:36, Laurence Chen wrote: >I fully agree with Harald Faber and Blass Uri statements that one should not >believe a chess engine evaluation without checking the position, just because a >chess engine post a +3.xx or +10.xx does not mean that the position is a win, it >could be a DRAW. One thing thing, a very important detail they omitted, was the >reason why a chess engine gives such an evaluation. To those of you who >dependings a lot on chess evaluations, you must always add this FACTOR in your >favorite chess engine evaluation, MATERIAL count. Yes, chess engines evaluates >position based on Material VALUE, I don't know the exact value, but, I believe >it is based on this scale, Pawn equals 1 point, Knights and Bishops, 3 points, >Rook equals 5 points, and a Queen equals 8 points. These are rough estimates >and it may vary from different engines. You should always ask why a chess >engine gives a certain value, what is the material count? Using this example >from another post, many chess engine will give a +4.xx evaluation for White. > >[D]6R1/7B/6Pk/8/5K2/8/8/3r4 w - - 0 1 > >Now, before you assume that White is winning, make a material count, using the >scale above, one gets roughly 4 points in material count. So White is +4 >points, and do you know why? It is because it has a pawn and bishop advantage. >So the evaluation of the position is +4.xx because of the material advantage, >however this does not mean that such material advantage can easily be converted >to a win. If the evalutation of the above position was +6.xx then it would mean >that White must have some type of positional advantage besides its material >advantage. Correct understanding of chess engine evaluation will help one to >know when the evaluation is not superficially incorrect based on material value >alone. I hope this helps. Remember the old saying, the blind leads the blind, >and they both fall into the ditch. So beware of this factor. Chess Engine evaluations are (basically) tactical only evaluations. As such, they should be considered as having that sort of value. Hence, unless a checkmate is seen, you can never be sure that a position is winning or losing. That having been said, I think that a GM's evaluation is also subject to error, and perhaps the margin is even greater than a computer's evaluation [IOW, I have seen plenty of positions in books written by GM's where the GM does not give the best possible answer]. Further, a weight should also be given to statistical analysis for any given position. Suppose that a GM pans a position as lost and the chess engine analysis is also negative, but statistically, the position usually wins? It seems it would need deeper evaluation. Hence, the best answer is to have a complete database that contains: 1. Computer evaluation 2. GM evaluations 3. Statistical data And then to consider all of the above + 4. Your own thoughts after careful consideration of all of the above.
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.