Commentary for black move 8: Also, the move Bc8-e6 is a little bit suspicious. What I mean by that is that it invites the move Nf3-g5, attacking the bishop, and in some cases white will then bring this knight from g5 badge to -- back to e4 with a solid advantage. Strategically speaking for just a brief moment, what is white's concept? White's concept is to dominate the super, and most specifically, the d5 square. In this case, the pawn on c4 controls the d5 square, the knight on c3 wants to leap to the d5 square, and it's supported by this long-range bishop on g2. GK MOVE: 8 O O.

YASSER SEIRAWAN: So the move on Nf3 to d5 -- we do have a move so I'll interrupt myself. Two moves, in fact. DB MOVE: 8...Q /2K-7#.

YASSER SEIRAWAN: I think we're about to see the rook f1-e1 by white, or even the move Nf3-g5. Because of the whole strategical line for white -- because of this particular opening, white in general, to gain control of the d5 square, needs to move this /THAO*EUT on f3. So in concert, putting the bishop on e7 is strategically a misfit. The bishop is passive here. To allow the knight on f3 to jump to the square g5 with tempo is also an error. So I would say that after a vair nation like the move Nf3-g5 attacking the light squared bishop, the light squared bishop is forced to move, for example to the move g4. Then will come the move h2-h3 kicking the bishop. Let's go back to the game position for just a moment. What Deep Blue has done with the moves Qd8-d7 and Be6 is to set up what we call a battery. It's lined its pieces up together and what Deep Blue would dealer love to do is to play the move Be6-h3, trying to force an exchange of light squareed bishops.

MIKE VALVO: Could I add a comment here of something about computer strategy or computer psychology. Garry actually played a line that's well known if you leave off a3. And if you leave off a3, the book line is to play b6 and fianchetto the queen bishop. Garry's idea is to get them out of book, they won't know the move and play an inferior line. If you play g3 yourself, black will play g6 and go into the normal variation. And this is something you can try on your home computers too. If you emlate what Garry did, I'm sure the computer will do the same because I've noticed that in thesis positions, -- these positions which I've tested with these computers they're all play Be7 /RA*ER fianchetto, because they think it's weakening the squares around the bishop.

MAURICE ASHLEY: Forcing them on their own to get out of that book, which in this case Grandmaster Joel Benjamin has working with the computer for over eight months, has put in would be helpful, and what you're saying, Yaz, are you saying maybe black is busted?

YASSER SEIRAWAN: No. No, no, no, no. Black is not busted. I just appreciate what Mike has said that Garry, through the mixture of the opening move d2-d3 and a2-a3 has bamboozled the computer in the opening and allowed to -- it to misplace its bishop, giving him a clear and significant advantage. Transforming an advantage to a victory is one of the hardest things in chess.

MAURICE ASHLEY: Well, the move d2-d3, this is not the kind of move that I would tell my students to play.

YASSER SEIRAWAN: No, you wouldn't.

MAURICE ASHLEY: Now that Garry has played did 2-d3 everybody is thinking "Hey, I can use this." What are you saying, has Garry played intentionally poorly in order to make the computer play even worse? (Audience laughter.)

MIKE VALVO: Well, I think --

YASSER SEIRAWAN: Bingo, you hit it.

MIKE VALVO: I think it falls into a category that humans would call "move order," or considering transposition. If he started off with the English on the first move, black would have played e5 and sole kind of four knights Defense. He didn't want that. He wanted them to actually play d6, so he probably tested it out against Fritz or some other kind of machine and he snowed "Gee, if I play this move order, they will play d6. And if later on I play a3, they will play Be7, and he probably prepared a whole bunch of these things for the computer, more than IBM could have anticipated in advance with Joel or anybody else, and he had this waiting for him. Notice that he did not repeat the first game. He got a great opening out of the first game? Why didn't he repeat the first game? He had other things he wanted to try before they could recover.

YASSER SEIRAWAN: You said something, Mike, I find kind of interesting. You said that Garry could take some Mike computer programs, let's say the world's best, stick them on his computer at home and the majority of them all play the opening moves. Now, we noticed ourselves that the Fritz program, our little helper here, wanted to play the move d5. So if Garry had used Fritz to prepare, he would have expected the move d6-d5. Why do computers play similarly? Are the algorithms all the same? Why do computers play --

MIKE VALVO: The backbone of computers playing the chess is the same no matter what the computer is. There are differences, but basically what they do is they generate all the possible positions and they compare them against each other. The secret is, how do they evaluate each position, and then the comparison of course is going to be different as a result. What Deep Blue does is going to be different than what Fritz does. And of course Deep Blue is going to generate many, many more, 200 million positions, remember.

MAURICE ASHLEY: Per second?

MIKE VALVO: Yeah, per second. Against Fritz, which is maybe 10 percent or less of that number of positions. And it's just not going to have the same background material to work with. But by and large they have the same kind of tendencies no matter what computer you're dealing with. They'll all tend to do roughly the same thing.

YASSER SEIRAWAN: I see.

MAURICE ASHLEY: Well, now, that Kasparov has been thinking for some time, and I don't see him ready to move --

YASSER SEIRAWAN: You want to jump back to game two? Before we do that, let's just go for the game position, our expectation is I did the move -- is either the move Ng5 or Rf1-e1 by Kasparov. As I explained, chess Grandmasters will consider the move G e6 /* h3 to be a strat eeje cal threat and with the move Rf1-e1 the idea is that white will meet the move Bh3 by dropping his bishop back to h1, preserving the bishop. Now we're going to try to understand why the final position in game two was a draw.

MIKE VALVO: While you guys are going through this, I think I'll go up to the press room and see what the reaction is up there to what's been going on in the game. So I'll see you in a bit.

YASSER SEIRAWAN: IM Mike Valvo.

MAURICE ASHLEY: Okay, Yaz, let's get back to this crazy game. This possibility Qe3 has still got me in shock. When I first heard the news I got it from any paren -- my friend Josh Waitzkin, who is also an International Master. He was as excited -- very excited. He and other Grandmasters all thought it was a draw. I just wanted to insist, this is impossiblement Deep Blue played to beautifully, strategically, the whole game, a masterpiece. Any Grandmaster would be proud to play these moves, and a draw didn't seem like justice.

MAURICE ASHLEY:

YASSER SEIRAWAN: No, it doesn't. And I must say I suffered the same feelings. I have such a profound respect for Garry Kasparov, the world chess champion. To me, he is a Michael Jordan, if you pardon the analogy, of chess, and to even believe that it was possible for him to resign a drawn position, I shook my head and said, "Sorry, guys, the martinis you're drinking are a little too strong." No, I was in disbelief. Let us see what we were looking at just a moment ago. And that was after the move Qe3, the idea of the move Ra6. And this is why Garry resigned. You have to ask yourself "Well, why did Garry resign?" Garry resigned because he saw the move Qe3. He further saw the move Qxd6. He didn't believe that he had compensation for the lost bishop. It was just he was filled with disgust, he had been out played the whole game. He knew he had a lost position. He had convinced himself, actually, that he had a lost position, so he resigned. And then the idea was this very -- the star move, Rb8-e8, calmly saying, "You've won a piece, but so what."

MAURICE ASHLEY: Why that square, why D e8?

YASSER SEIRAWAN: Well, the idea is that black's king is especially vulnerable, and the queen on d6 threatens the rook on b8. So the idea is this rook covers this possibility of Qd6-e6+, it gets the rook out of capture, and it just also allows the possibility of meeting a move like Qd7+ with the interposition rook E -- Re8-e7. A crucial point is that white would dearly love to play the move Ra6-a7+, driving the black king back, followed by the move Qe7 and checkmate Kasparov on the g7 square. But this queen on e3 covers the a7 square. So anyway, this was the setup. Everybody asked me, "Yaz, what about Re8? Isn't that simply a draw?" And I said "no." There's actually a win for white, and I said the win was the move Bf3. And then after the move Bf3, I saw a whole sequence of checks, and this is known as a draw by perpetual check. If one player can perpetually attack the opponent's king, the game is a draw. And the perpetual check is that black plays Qc1+. And I said no, white can avoid this. Kf2, dodging the first check; Qd2+ again. Now we interpose the bishop, Be2, Qf4+. The king drops back to e1. Qf4-c1+. The bishop drops back, and I said, "Brilliant. This is the winning variation for Deep Blue, because then after the move Qe3+, the tricky move Kf1, Qf4+. Now white's king scur riz into the corner. Kg1 Qe3+ Kh1 wins. The idea being just to finish there is no further check. There is only one check, that is Qe1+ Kh2. Now, because of this decisive check of Bd1-h5+ it's necessary to capture the bishop, and now it's white's turn. His king is safe, and he can turn his attention to the attack with Ra7+, king has to drop back, Qe7 and it's mate.

MAURICE ASHLEY: Easy line, Yaz.

YASSER SEIRAWAN: Easy line!

MAURICE ASHLEY: Did all this over a cup of coffee, huh?

YASSER SEIRAWAN: I think it was several cups of coffee, actually. But it was such a relief for me. I could put it out of my mind and say to myself "Oh, thank goodness. No, no, no Garry resigned." However obviously this is a deep piece of analysis. His resignation of premature. At least for ourselves, the audience, we should see why he resigns. And in this case we have the full understanding. So I am very pleased with myself, I found the winning variation for Deep Blue. I could quietly go to sleep. This morning I was woken up (Audience laughter.)

MAURICE ASHLEY: And what was the variation?

YASSER SEIRAWAN: Well, it was so smooth, everything was so smooth. We got to this position.

MAURICE ASHLEY: That is after your 51st move Bd1 in your variation.

YASSER SEIRAWAN: Right. I had only just now calculated for you the variation Qe3+ because it seemed like the only sensible decision. The rude awakening is the move Qxc3+. I said, "This doesn't change anything. I am skill going to move over to the king-side with the move Kf1. And now the computer found an absolutely star move. Because of the threat Ra7 check followed by Qe7 which wins immediately, I figured that everything that black does has to be with check, has to be with check. And I thought that after the move Qd3+ Be2, there's simply no perpetual, nothing to be done, so Garry would laws. But, a star move, I didn't see it, I have to confess, the computer -- a computer found the move Qc3-c1. And I've got a problem. This move forces the draw, because after the move Ra7+, king drops back to g8, queen goes to d7, white is set up to mate now we have queen takes bishop, Kf2 Qd2+. We can't play Kg3 by the way because of the unfortunate Qf4 checkmate. So we have to step back. If we go to g1, this is the tricky move. Now black has Qc1+. If he goes to h2, there is the move Qf4+ and I can't escape the perpetual check.

MAURICE ASHLEY: We should point out that g2 g3 can be answered by Qf2+ Kh1 Qf1+ and this will go on forever.

YASSER SEIRAWAN: So now it was time to panic. So all of these lovely variations all had to be written off and we have to go back to the game position and say, "Oh, come on, please say it's not true," and we had to go all the way back to the start. Here we are back at the final position that Garry Kasparov resigns. We know that the correct move is Qe3. We had to question -- beg my pardon. I'm getting ahead of myself. Once again, Qxd6 Re8. Rf8-e8. Now we had to question the move Be4-f3. Maybe that's the wrong move. Can we win the game without that move? And the idea is that white's king is in this accuracy box, both in the center and on the king-side, so there's the stunning move, a paradoxical move of leaving the bishop on e4 to capture and playing the move h3-h4.

MAURICE ASHLEY: Before you continue, Yaz, I just want to get back to this point. Isn't this just a try, this move Qe3? It's just -- you know, you're down, you're busted, you're dead, but there might be a perpet.

YASSER SEIRAWAN: Right.

MAURICE ASHLEY: Isn't this something you would do just anyway? Forget about an audience that we should show it so that the understand understands why I'm losing. I don't think Kasparov is thinking of that when he thinks he's busted on the /PWOERT.

YASSER SEIRAWAN: No, he's not resigning for the audience's pleasure.

MAURICE ASHLEY: No. But just the situation, you give it a last try, you sort of see what happens.

YASSER SEIRAWAN: You don't give up.

MAURICE ASHLEY: Would he have done this against a human, against Anand, Karpov, any of those illustrious players?

YASSER SEIRAWAN: No, I think he would have resigned as well, for the same reason. You see, the computer has an advantage. It doesn't have this body of emotions. We human players get depressed. We simply get depressed. We get beaten, we get depressed. The computer doesn't get depressed, it doesn't have any prejudice, it doesn't carry along any emotional turmoil or upset, it doesn't get an upset tummy. It doesn't "byte" too much. Garry was feeling rotten the whole game because Garry was getting outplayed the whole game. Garry was in a mental framework which said to himself "Man, I hate this game. I'm disgusted it myself. I played like a jerk. I'm going to lose in front of millions of fans. What am I doing here? Why did I wake up today?"

MAURICE ASHLEY: But doesn't he also have the same contempt factor that he may say maybe this guy will miss something, maybe I'll get this strange perpet, because this guy isn't perfect, and maybe he thought the computer was perfect.

YASSER SEIRAWAN: Again, there is that intimidating factor. When you sit there and told that your opponent analyses chess at 200 million moves a second, and all you're looking at is a three or four move perpetual check, you've got to figure that your opponent's seen everything and, you know, the hardy hand clasp.

MAURICE ASHLEY: So to give Deep Blue some due, we don't know that Deep Blue hasn't seen something in that 200 million move per second analysis that we haven't spotted. I mean it isn't likely, but we're not sure.

YASSER SEIRAWAN: Well, C. J. Tan, the lead for the IBM Deep Blue team has spoken to this controversy, and believe me when they come out, we'll get something official, but he said that "No, when Garry resigned, Deep Blue's valuation was a big one, and they were playing for the win. They didn't think -- nothing was untoward until the controversy erupted. Let's get back. This move h4 -- does this move h4 win? And it looks really good. After -- I won't bore you with all the analysis, but the bottom line is that after the move Qxe4 Ra7+ Kg8, an important move now is the move Qd7, defending that f5 pawn with check, and there's simply no perpetual in the position. Let's just see that to be sure. There's the move Qf4+ Kg1 Qc1+ Kh2 Qf4+ Kh3 Qe3+ --

MAURICE ASHLEY: That's the last check.

YASSER SEIRAWAN: And then after Kh3, life is good. So now we'll go back to the position after the immediate move h4. This is what Dokyan and Frederic fried he will talked to Garry Kasparov about, and Garry was very excited that h4 does in fact win the position for white, and then they said, "No, it doesn't because of this crazy move h5. (Audience laughter.) Garry's understanding of chess is such that within literally seconds he went din, ding, ding, bop, bop, he went "oh, my God."

MAURICE ASHLEY: What does this move h5 do?

YASSER SEIRAWAN: What this move h5 does is it turns the king-side into a fortress so that white's king cannot escape by stepping up in a very crucial variation for the g4 square. So now we have the same problems as before, with the move Be4-f3. Because of this crucial little tempo move h5, there comes the move Qc1+ Kf2 Qe2+. Now after Kg3 there comes the move Qf4+ Kh3 Qxf5+ Kh2 Qf4+, and we can't escape the perpetual checks again on the king-side.

MAURICE ASHLEY: Amazing. Garry must have been suffering after hearing this.

YASSER SEIRAWAN: Ha-ha-ha. Garry didn't enjoy his dinner. Gosh, we're getting bogged down in -- into a mountain of analysis, but this is exactly what top professional chess players do. They analyze a position and they keep analyzing, analyzing, analyzing, until they can prove to themselves, satisfy this with.

MAURICE ASHLEY: Did Garry have to see all this to play Qe3?

YASSER SEIRAWAN: No. The problem is if Garry had seen an infinitesimal amount of this he would have played Qe3 instantly. The problem is he was carrying all of this emotional baggage with him into the position. He had convinced himself that he was lost before he got to this position. He had convinced himself that "Well, back on move 37, the move Qb6 won material for Deep Blue," and then he was lost at that point. So for the last five, ten moves he had been suffering with this conception of having a lost position. He was waiting for an adequate moment where it was clear to everybody that the game was lost, and so he resigned at what he felt was the clearest moment.

MAURICE ASHLEY: It's a big argument that fans generally have against chess resignation. Often chess players resign and the fans think why didn't you play on some more? "

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