Commentary for black move 12:

And I tell you, Yaz, some of the moves have been really strange. To my eyes, I was wondering, for example, before this, after Bb4+ Nd2, it seemed to me that after white castles at some point, let's say after h5 white were to castle, it seems to me white would be able to get in this c3 push without worrying about the d3 pawn and then an automatic d4 push because that pawn has protection, and black would be in a little bit of trouble, being pushed back with these center pawns. So this move Bb4 to my eyes is sort of monkey see check, monkey do check, by the way.

GK MOVE: 12 Qe2

DB MOVE: 12...Qc7 , protecting the knight. Now it just seems, maybe I'm wrong, now it just seems as if c2-c3 can be played and then after some bishop move, let's say back to e7, d2-d4 can be played -- d3-d4, excuse me. And now this knight has to drop back. I don't know which square is best. Maybe this square is best, but it seems these pawns have moved up the board with tempo, control of the board has been assured by white by his central pawns. And I'm not really sure what this move h5 does.

YASSER SEIRAWAN: Okay. Let's talk about that for a moment because this is the whole point of what IBM's concept is. You have this incredibly difficult task called chess. With so many possibilities, how on earth can we even get a computer to even play a decent game of chess? But how's this for a fascinating question? What if we can make a computer that by looking at all of these different chess moves, comes from a completely unconventional and is able to teach Grandmasters something about their game? Well, this would be shocking. A game that is centuries old, millenniums old, and to teach people something about their game of chess through unconventional means. What we have here is a perfect example of what IBM's programmers are trying to do. We can question, as humans, the judgment of Deep Blue. Because he's doing something that is completely unnatural to strong players, like Maurice and myself, we wouldn't consider these moves h7-h5. Now, what Deep Blue says to itself is, "Hey, this move seems to, c3, the moves d3-d4 are coming, pushing me pieces away. I've God to do something about my knight on e5. Let me figure out what I can do." And it says to itself, "Huh. How about that. White has a weak square f4. Of course, white does not have a weak square f4. And Deep Blue says that's the square I'm going off. This is what I'm going to do, folks. I'm going to play h7-h5. And I'm going to play h5-h4." So let's give Garry Kasparov a move. Let's give Garry Kasparov the move castles. New variation. Now we play the move h5-h4. After Garry plays g3-g4, bypassing the h-pawn, then Deep Blue says, "Aren't I brilliant? I now control the f4 square. I'll play Ne5-g6 and when Garry does something silly like c2-c3, well, then my bishop will drop back to d6 and I'll have played perfectly, I've got control of the f4 square." So this is the way Deep Blue is trying to solve its problems. What I had suggested the move Bc5, I thought that that was better than the move Bb4+ precisely because of the problems that the computer now faces, which is c3, attacking the bishop, and d4, attacking the knight. So this is what Garry is now thinking about. Is the move h5-h4 bothersome? And if it is, well, white can always play the move h3-h4. But then Deep Blue solves its problem in another way. It now says, "Oh, it's not the f4 square that's weak, it's the g4 square that's weak."

MAURICE ASHLEY: Well, what if Kasparov tries to handle all of that and plays the variation beginning c2 /450EU6# c3, and after the bishop drops back, plays d4, and after the knight drops back, now plays your move, h4, and that knight didn't find the h4 square and didn't find the g4 square. In fact, it's found no square. What's the deal with that?

YASSER SEIRAWAN: Then Garry would be better! (Audience laughter.) And the strategy wouldn't work.

MAURICE ASHLEY: Well, that /SAOEPLDZ like -- seems like a very simple variation for Kasparov to play. Maybe there's a response that we haven't noted, and I'm sure there are other moves we have not looked at. Deep Blue is looking at all the moves in a given position, but this has come as a surprise to Kasparov and he's going to work out exactly what he feels is the best way to deal with the situation, and I have to say, I'm very suspicious of the some of the moves Deep Blue has played. But as we've seen in so many games, Deep Blue has played some wacky looking moves?

YASSER SEIRAWAN: Absolutely. And he's gotten away with it, too.

MAURICE ASHLEY: And he's gotten away with it. This thing mixes it up like anybody. The game gets wild, game gets sharp, Deep Blue sacks is piece or a pawn, and then Kasparov gets confused --

YASSER SEIRAWAN: And then he gets angry.

MAURICE ASHLEY: -- and we've seen that time and time again. Deep Blue has played some suspicious -- as far as we're concerned, the human side, strategical decisions, for its tactical ability, time and time again --

YASSER SEIRAWAN: Saving the game.

MAURICE ASHLEY: -- turns out fine.

YASSER SEIRAWAN: Absolutely. Throughout the match, three or four games, I think the match score is right, 2-2. But quite frankly Garry has had the edge in three out of the four games. And sometime the computer, even though it makes some questionable decisions, some very, very unorthodox or unconventional decisions -- i.e., it plays like a computer. It still, nonetheless, has managed to save difficult positions that, quite frankly, I don't think any human player could have saved because they were so walk ward. We especially saw that in game four. And maybe it's it's a mixture of intimidation by Garry Kasparov, because Garry Kasparov clearly thought that in the game that he lost, the computer had seen everything and he resigned, as opposed to playing on and playing for that perpetual check. So Garry has maybe given the computer credit, and perhaps too much credit.

Real-time text commentary is made possible by LiveNote, Inc. and Vincent Varallo Associates