Commentary for white move 28: And also consider the fact, getting back -- you know, making the analogy in the world of chess now, if Kasparov were to play a match with, say, Anatoly Karpov, it's not -- it would not be so unusual for Karpov to go into hiding for a few months, two, three months, whatever, and just study and analyze and possibly even come back with different openings than he had before. Certainly we'll come back with some new moves. And there would be nothing unfair about that. It would simply be his decision. Also, keep in mind that previous incarnations of Deep Blue played plenty of games. The old Deep Thought games. These games are available. Kasparov had these games before the first match. He still has them, unless he lost them somehow.

YASSER SEIRAWAN: I don't think he lost them. JOEL BENJAMIN: So he still has them. He can still look at them. He can still try to learn from them. He just doesn't want to anymore.

YASSER SEIRAWAN: Right. JOEL BENJAMIN: Or he doesn't want to talk about them. So I don't see any unfairness there at all, and I think, you know, we ought to stop talking about it.

YASSER SEIRAWAN: Okay. JOEL BENJAMIN: Doesn't make any sense.

YASSER SEIRAWAN: You made your point. Let me just say that we do have a move by Deep Blue. He has played the move Qe2-f1. I find this an interesting step back. Basically the move protects the square f5, making the incursion Ng3-f5 playable. Also, there's this kind of weird idea of playing Ra1-a2, followed by bringing the queen behind the rook in order to try to force a favorable resolution on the queen-side. So this is a very flexible move in air position with -- a position with boundless possibilities. I think it's a good move as well. I like it. JOEL BENJAMIN: Yeah, looks like a good move. And I'm confident that Deep Blue will play good moves in this position. I don't know if it will play the best move every time, but it will certainly find things to do and it has a good position, objectively. So it's in pretty good shape here.

YASSER SEIRAWAN: Okay, we'll take another question. Yourself, yes? You're the usher. Go ahead.

AUDIENCE MEMBER: I think the word I was curious as to whether if he played -- the computer will play the same match, given the same moves, every time. In other words, I think the word is deterministic, so if you played yesterday's match over, would it behave the same way? JOEL BENJAMIN: Okay, so you're saying that if we -- if he started with the same opening moves, same opening book, would it continue to play the same moves all the way through. Not necessarily. And I'm not really a scientist. I play one on TV. But I do know that it has something to do with the fact that it's a parallel system and the way that all the different processors kick in and at what moments and so on, that it is possible that the, you know, a move could change, if you gave it exactly the same situation. So it doesn't -- it's not completely deterministic. You don't get the same response every single time. But of course, you know, we may decide to play a different move here or there anyway.

YASSER SEIRAWAN: Does the Deep Blue team go into a match and say, "Well, look, if Garry plays this particular opening or this particular defense, we don't want to just choose a random opening, we want it to play this defense, or this opening?" And you make that choice before the game? JOEL BENJAMIN: Yeah, that can certainly be done.

YASSER SEIRAWAN: Another question from our audience? In the far back. Yes?

AUDIENCE MEMBER: Hi. While I agree that it's fair that Kasparov didn't have access to Deep Blue's games, it seems to me that the way the program can be tuned in between games that in effect Kasparov could be playing a different program between these games. Yes? JOEL BENJAMIN: Kasparov could be --

AUDIENCE MEMBER: Playing a different program based on what am -- altercations you might make to the program during the course of the match.

YASSER SEIRAWAN: Yes, what the gentleman might be saying is, yes guys have all your research isn't it nice and you /TEURPBD it -- turned it over to Garry and after you turned over all your research to Garry somebody said, "Oh, great, let's do this, we'll tweak the program, and actually what Garry is getting is false information because the latest version, or rather, this week's version was better than last week's version. JOEL BENJAMIN: Well, again, I don't see any problem with that because Garry can also change -- he could change his style of play completely.

YASSER SEIRAWAN: Which in fact he has done. JOEL BENJAMIN: He doesn't need programming to do that. He can do that any time he wants. He can do that within the same game. He can play a different opening if he just feels like it. He can play something different from what he had planned before the game. So we can certainly get a different Garry Kasparov from game to game. I think in game one we saw a different Garry Kasparov than we did in the first match, and I don't see anything unfair about that, either. I mean that's just the way it is in sports.

YASSER SEIRAWAN: We have a question?

AUDIENCE MEMBER: Actually you've already come up with this, I guess, but I was saying instead of taking A B, like analogous to the Fischer-Spassky the first game of the match, like R1a2, which seems and Qa1 which seems to win the a-file by force.

YASSER SEIRAWAN: What the gentleman has said is that white would very much like to clarify matters on the queen-side, but he'd like to do it in such a way as to give himself complete control of the a-file. And in the 1992 match between Bobby Fischer and Boris Spassky, Bobby play you had -- played the idea of tripling on the file, bringing his rook up on a -- up a square from a1-a2 and then dropping his queen behind the rooks. The idea is this loaded gun, if you will, on the a-file would force a resolution that would be favorable for white and Deep Blue. Overall again I must say that this particular position I think that Deep Blue has played it very well. Even we have been stating that it is a closed position, and this is the kind of position that Garry was aiming for, if you will, a blocked position. One question, Joel, before you -- maybe this is a little bit unfair, but I know you're an opening expert in unorthodox openings. You've written some books and things like that. Wouldn't it behoove your team to get also other experts in certain openings and invite them to come and help you in the opening research, or maybe you've done that? JOEL BENJAMIN: That might be a good idea.

YASSER SEIRAWAN: Okay. One more question?

AUDIENCE MEMBER: I was wondering, is Deep Blue threatening to win by pawn takes pawn, pawn takes pawn, rook takes rook, rook takes rook, queen takes, bishop takes, pawn takes, bishop, rook, queen takes knight, queen checks, knight takes bishop one.

YASSER SEIRAWAN: You're right. What the gentleman has said is that the pluve Qf1 does have a threat. JOEL BENJAMIN: This is what we call in computer chess a null move. Essentially a null move. It doesn't change the situation. And then exchange.

YASSER SEIRAWAN: The idea is that the move Qf1 set up the threat of axb5, axb5 Rxa8 Rxa8 Rxa8 Qxa8. Now the black queen has been lured from the protection of the knight on f6 so that after bishop takes h6, white could snare a pawn, gxh6 Q Xf 6 with a big advantage. So, yes, that is in the position, and I actually would like to say we'll take a look at the position for a moment, and because of this reason, and not only because of this reason, the move Nf6-e8 by Garry Kasparov may be played in this position. Go ahead. Let's try that. The idea is to side-step the step the gentleman just mentioned and reposition the knight from f6 to the blockading d6 square. A knight usually stands best in front of a passed pawn or blockading pawn. This is the type of move we might also see from Garry. Was there one more question? Yes, I beg your pardon. Uh-oh, we've got another tiger here.

AUDIENCE MEMBER: This is a little back in the position.

YASSER SEIRAWAN: A little bit back? How far back would you like to go? A little bit more? A little bit more?

AUDIENCE MEMBER: Yeah, this. No, put the pawn back. There.

YASSER SEIRAWAN: Okay. And...?

AUDIENCE MEMBER: Well, I thought in this position Qh4 would be a pretty strong move for black.

YASSER SEIRAWAN: Okay. By the way, what's your name?

AUDIENCE MEMBER: /A*EUFRry.

YASSER SEIRAWAN: /A*EUFRry? JOEL BENJAMIN: Okay, so /A*EUFRry has suggested the move Qh4 which attacks the knight on g3. Avery has suggested the move Qh4 which suggests the move Qxh3. I think the move can be handled by a queen move, Qf3 or f2, I'm not sure which one is better. But that seems to defend against both threats. Very good on seeing an opportunity to create a double threat, but I think that there was a way that Deep Blue had planned to defend against it.

YASSER SEIRAWAN: Question? Yes, sir?

AUDIENCE MEMBER: I was curious that with Deep Blue's ability to -- I think I read analyze 200 million per second, what does it -- why does it take so long to determine its next move?

YASSER SEIRAWAN: Well, that's good. JOEL BENJAMIN: Much like people do, the longer it thinks, the more it sees, and even if it sees a whole heck of a lot within a few seconds, if it thinks a couple minutes, it can see even more. And it can find that the variations it thought might play early in that search were not the right ones and that there were better moves. So the fact is that it has two hours to make its first 40 moves, and it knows that, and there's no reason why it shouldn't take at least a good portion of that to try to find the best moves.

YASSER SEIRAWAN: Joel, when I'm losing a game of chess, I mean I'm suffering horribly, obviously, so I'm sitting there thinking to myself "I've got to be able to rescue myself," and here I am struggling trying to save the game, and I'll find myself taking more time when I'm in a worse position as I try to fight my way out. Does Deep Blue say to itself "Oh, boy, I'm worse, I've got to analyze as much as I possibly can to save the position." Will it start taking, like a human, more time in an inferior position? JOEL BENJAMIN: Probably it's more likely to use a lot of time in that situation because it's having trouble finding a satisfactory move, so it will keep looking until it finds one. But the thing that Deep Blue avoids that humans often do is that when Deep Blue has good moves available, it doesn't waste a lot of time. And a lot of time pressure addicts and you and I have both suffered from this problem on occasion, sometimes in the same game, --

YASSER SEIRAWAN: Oftentimes, it seems, in the same game. JOEL BENJAMIN: Yeah, and one of the problems that we get into, one of the reasons we get into time pressure is that we'll have a situation where we see a good move, but we're not so sure -- we spend a lot of time convincing ourselves. We see more than one good move and think a long time about it. If it sees a good move, it's probably not going to have to take extra time.

YASSER SEIRAWAN: Looking at the game position, I just see that Garry's grasping his head, and he doesn't look -- he didn't look to be a very happy camper. I mean this was clearly -- he didn't see what he had prepared for, he's not happy here. We know he's facing this long-term threat of axb5. He could be forced in a defensive mode with knight e8. JOEL BENJAMIN: I think he's amazed that Deep Blue played the Ruy Lopez this well.

YASSER SEIRAWAN: Yeah. And we were very happy about this move Re1-c1. We felt that that was a very sophisticated move. I think you yourself would have been very happy to play that move? JOEL BENJAMIN: Yes, I like that move. Of course, it considered various pawn captures and breaks and so on, and it was thinking about them. It wasn't sure which was the best move, but when it got towards the end it decided the best thing is just to wait and maybe open up the position the next move.

YASSER SEIRAWAN: All right. We'll take another question from our audience. Yes, sir? JOEL BENJAMIN: Maybe we should make this the last one.

YASSER SEIRAWAN: We'll make this the last one and then Joel is going to go back to his den. Yes, sir?

AUDIENCE MEMBER: I'll ask a nontechnical question, or a nontechnical question. I've noticed that even Murray Campbell of the Deep Blue team is recording the moves by hand on a score sheet. Has any thought been given to what will be done with those score sheets? I could see where they might have some historical significance.

YASSER SEIRAWAN: Yeah, the question is that Deep Blue's handlers --

YASSER SEIRAWAN: We do have the move Nf6-e8, avoiding the threat the gentleman pointed out. Why is it that the score sheet, you're writing the moves on the score sheet since we know the positions in Deep Blue's terminal and the like?

AUDIENCE MEMBER: Not only that, but has the Deep Blue team perhaps considered donating those score sheets to say the Smithsonian to historical significance, particularly should Deep Blue win.

YASSER SEIRAWAN: Well, first of all, it's the rules of chess. Maybe it's a rule that we may see go out the door, but it has been a rule of chess that you have to record your games. And the score sheet is actually the official score for claiming victory by time, claiming a repetition of moves, or invoking the 50-move rule. So that has been always been the official scores, and the arbiter, Mike, you know that very well. Yes, clearly should the operator have to take the score? I don't know the answer to that. JOEL BENJAMIN: Well, partly it's a backup because, you know, there's always the possibility that something could go wrong that could be, you know, a power failure or something. I don't know what it would take to actually wipe out the memory, but something could happen. But also mostly it's just convention, it's just, you know, when chess games are played, both participants keep the score of the game, and they sign the score and that's -- they verify that this was indeed the result of the game.

YASSER SEIRAWAN: And the gentleman's further question was that the official score sheet, should Deep Blue win, may be valuable. What happens to the official score sheets of, for example, the Philadelphia match? I mean I'm sure Garry kept his except in the game he may have lost. JOEL BENJAMIN: Well, I know for a fact that somebody has a copy of that score from game one. I'm not going to say who it is. It's not me, but I know somebody has it. Well, who knows. And if you do get your hands on one of the score sheets, I'd hold onto it.

YASSER SEIRAWAN: Grandmaster Joel Benjamin. (Audience applause.) Mike, we've been seeing some moves, and let's just take a poll. Fritz really likes -- green mood.

MIKE VALVO: It's all green.

YASSER SEIRAWAN: Yeah, it's all green. And the green is growing. It's 72 hundredths of a pawn better, in Fritz's opinion. Does our audience agree with Fritz? Has Deep Blue a big advantage? Raise your hands if you think Deep Blue has a big advantage. Okay.

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