Author: Dan Andersson
Date: 10:00:18 01/30/04
Go up one level in this thread
Bad trades come in many flavours. But it really comes down to material correlation. E.ex: Bishops correalate better with rooks than knights do. Knights correlate better with queens than bishops do. There are many ways you can implement bad trade code. Two opposed ways are: A lookup table based on the difference of the totality of exchanged pieces. This has the pro of being relatively small. A lookup table based on the totality of pieces remaining. This has the pro of capturing the actual material correlation. Of these two I prefer the latter since I believe it to be more correct. But the former is quite ok compared with most ad hoc schemes. Both of these schemes can be trained by TD-lambda techniques. Or by statistical analysis of played games. And then using otimization you can get a classification/decision tree with a resolution you can live with. MvH Dan Andersson
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.