Author: Ed Trice
Date: 04:57:22 12/31/03
Go up one level in this thread
Crafty has the variable BADTRADE set at -1.2 pawns when a Rook + pawn is given up for a Knight + Bishop. This discourages the program from giving up two active minor pieces for a pawn and sedentary Rook, which usually does not factor into a game's outcome until the late middlgame/early endgame. I have toyed with the idea of making piece values change as a function of the pawn population. With more pawns on the board, the Knights can overcome the Bishop. Queens and Bishops are generally worth more as pawns come off of the board and their mobility increase. You can tailor the weights of the Rook so that B + N is worth more than R + Pawn earlier in the game (say with 14-16 pawns still present) and then gradually have the value of the Rook increase as pawns are removed. This approach lets you 'tweak' values with more granularity, but you also have to choose your weight very carefully. For example, a Rook can mate a lone king (of course) but two knights cannot. You have to make sure your program will prefer a Rook to 2 Knights by the time the endgame rolls around! You might want to examine the relative ease or difficulty of winning certain pawnless endings to help assign weights to your pieces. Look at B + N vs. King compared to R vs. King as an example. The R win is trivial to execute, the B + N is not.
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.