Author: Stan Arts
Date: 11:12:15 06/16/02
Hello, I´m working on my chessprogram, it´s very new and I have just gotten a simple method of alpha-beta working in my program. Other than that it is still a simple fixed level mini-maxing program at the basis. Now my question, when it was still doing a full width scan, my program would usually only see 4 ply deep, 5 if you had a lot of patience. However, the evaluationfunction would cause it to play like it should more or less, I mean the positional parameters I put in it, but now with alpha beta, it can usually search 6 to 7 ply deep, but now it tends to see strange deep threats causing it to play strange countering moves, instead of really listening to the evaluation parameters like with lower plies. For instance, it basicly throws open it´s king side pawns when in the evaluationfunction it says it shouldn´t :) Still, when I limit play to 4 ply even with alpha beta it plays as before, so I know the alpha beta works right. Another problem is for instance that it won´t castle, because it keeps "fishing" for other opportunity´s on deeper ply´s instead, instead of just setling for the castling bonus, constantly delaying a good move by making the opponent counter something. This is strange, and on these bigger plies it constantly wants to do this. Always thinking "i can castle later" for instance. Would this problem only be solvable by extensions? Or can it be solvable by increasing scores in the evaluation, because I think that when I increase penalty scores, it becomes dangerous in the way that it then would start sacrificing to achieve something which isn´t worth that much. = dumb Anyone ran into this similair problem with their fixed level program? Please help out a starting (21 year old) chess programmer, Thanks! Stan
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.