Information Technology Reference
In-Depth Information
Figure 10.19.
Chess rating and search depth
?
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
rating
2500
1500
500
4
8
12
14
depth of search
It explored game trees to about depth 14, with occasional very deep searches.
The estimates of the values of game states used 8,000 separate factors.
It had a U.S. Chess Federation rating of 2650 in 1996; the rating of a chess grand-
master is 2500. (For comparison, the world chess champion as of January 2011,
Magnus Carlsen, had a rating of 2814.)
What is perhaps most interesting about the effort in computer chess is that almost
none of it is specific to chess. Instead, it is mostly directed toward speeding up the
game tree search. This is not too surprising, since the U.S. Chess Federation rating of
chess programs seems to be directly proportional to the search depth, as illustrated
very roughly in figure 10.19. (It is worth noting, however, that this effort on search
has not helped so far with
extremely
large games like the game of Go.)
Want to read more?
This chapter looked at the thinking behind playing strategic games. Getting comput-
ers to play these games, and especially chess, has been a mainstay of AI research
from the very beginning. Alan Turing, the father of computer science, tried his hand
at it, and one of the first minimax chess programs was due to Claude Shannon, the
father of information theory. John McCarthy discovered alpha-beta cutoffs in 1956.
