The first came in the diagram position when Deep Blue, playing White, moved its rook from e1 to c1. The move appears pointless, yet conceals a good deal of venom. First, Black is discouraged from playing ...cxb4, because his queen would then be on the same open line as the white rook; second, White opens the possibility of playing a later c4 himself, again embarrassing the queen.
When Kasparov replied by closing the game with ...c4, it was clear that Rec1 had served its purpose. As every experienced Ruy Lopez player knows, once Black has played ...c4, he not only deprives himself of Q-side counterplay, but also gives White the opportunity to play f4, gaining control of d4 if Black replies with ...exf4. Deep Blue must have worked all that out for itself before playing Rec1.
After 23.Rec1 c4, the machine again showed the septh of its calculations by playing 24.Ra3. Humans know the plan of Ra3, and doubling rooks on the a-file before playing axb5. Deep Blue once again worked it out from first principles.
The next surprise came a dozen moves later.
In this position, White hs a clear advantage thanks to his passed d-pawn and chances of attacking b5. Everyone expected Deep Blue to play 36.Qb6, with the double threat of 37.Qxd6 and 37.axb5. Black would then have to gamble on 36...Rd8 37.axb5 Rab8 38.Qxa6 e4 followed by Qe5 with counterplay.
Scorning the materialistic approach, however, Deep Blue played the patient 36.axb5 axb5 37.Be4, keeping Black squashed and leaving his own attack for later. It was a very human sort of decision for a machine to make.Reuse content