er-chess expert, claimed important advances during the year between matches with Kasparov in Deep Blue 's chess knowledge, not merely its ability to examine more positions.
Deep Blue's branching factor at each move (or ply) averages perhaps four or five, after allowing for various procedures, like the alpha-beta algorithm and special continuations that are selectively incorporated into the search. Hence, althoug. the speed of Deep Blue was increased by a factor of two during the year between matches, the search capability increased by only a fraction of a ply. Yet all grandmasters watching the match (as well as Kasparov) commented on the qualitative change in its play, especially its ability to play differently in different kinds of positions. So it is to chess knowledge, not brute-force search, we must look for mos of the improvement. That knowledge appears to take three forms:
A large opening "book," probably containing tens of thousands of game trees from actual games and from Deep Blue's own analyses, comparable to the amount of specific information about openings possessed by Kasparov and other grandmasters, although it
...