Author: Dann Corbit
Date: 19:06:09 02/26/99
Go up one level in this thread
On February 26, 1999 at 21:43:38, Christophe Theron wrote: [snip] > >Maybe I should say: not being a good chess player is a good thing when you want >to write a good chess program. > >I'm not sure you will agree on this anyway. > >I notice that strong chess players who tried to write a good chess program did >not succeed. I notice that there are weak chess players able to write the very >best chess programs. > >From a statistical point of view it means nothing, as you can argue that there >are very few good chess players that tried to write a chess program, and many >programmers are anyway weak chess players. > >But I notice that when I tried to include a lot of chess knowledge in my program >I was not successful. When I removed all the stuffs and tried to keep only the >very basic knowledge my program began to get stronger. > >Hard to explain everything in a few lines, but I can give some examples of human >chess knowledge that, IMO, is useless for a chess program: > * forks > * pins > * tempi >Don't you think a good player would try very hard to implement these concepts in >the first place? And I can tell you that he would loose his time... OTOH: * Bad bishop * Hanging piece * Pawn roller Will you write a good chess program if you don't understand these? I also think that a pawn one square from promotion has a value of at least a knight if it is guarded by a piece. That's because it will either tie up a major piece or be promoted or have to be removed at the cost of a piece. The writing of the eval function seems to require a great deal of chess knowlege. However, I think it could be created by a simple least squares fit and millions of data points. I doubt if anyone has done this. >Or maybe, surprisingly, there is NO relation between "being a strong chess >player" and "being the programmer of a strong chess program". I mean that being >strong at chess gives you no special advantage when you write your chess >program. > >Maybe it is easier too agree on this... I think the sample size is too small. If you find the following combination: 1. GM player 2. Has a degree in computer science 3. Has programmed in C, C++ or Assembly for at least ten years You will get a whale of a program if he decides to write one (also doubtful, since there are probably a lot better ways to make money). The problem is (I think) that you won't find any. All the time and effort it takes to be the best in one area will subtract from the other.
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.