function email() { document.location=("mailto: ddye" + "r\@real" + ""); }

Computer Go vs. Computer Chess

Why is computer Go hard, why is computer Chess easy?

The real stumbling block in Go is developing a workable set of heristics to model the playing process. However, this is the problem because it's a given that brute force is out of the question.

In chess programs, brute force is the only strategically or tactically significant factor in playing strength. The cleverness in chess programs is 95% directed toward doing brute force searches more efficiently. This *does* involve significant knowledge about chess, but in the end the knowledge is used mainly to order the search tree more efficiently. If the knowledge is faulty, the bad effect is that the search becomes less efficient. The moves chess programs ultimately make are almost completely dominated by counting the wood at the bottom of the search tree. In effect, the search is used to distill the rough knowledge into high potency play.

On the other hand, in go programs the knowledge is all there is, because the search is impossible; so every little flaw in the the playing heuristics remains a visible flaw in the program.

comments/suggestions to:

my home page