NOTE: I wrote the following article in 1996 during Kasparov's first match against Deep Blue.

A Deep Blue Day

Jon R. Edwards

The most famous article in the old American Chess Quarterly was Bobby Fischer's "Bust to the King's Gambit" in Volume 1, Number 1. Much less known and until today forgotten was Edward Lasker's article, "Automatic Electronic Chess-playing Machines" in Volume 1, Number 2. The issue, published in 1961, is now very hard to find. Forgive me, therefore, for quoting at some length:

"We are used to all sorts of exaggerations when the tendency to add fancy to facts for the sake of drama colors a newspaperman's report on an astonishing new scientific development which lies outside the sphere of the average layman's understanding. But the nonsense that has been written about chess-playing electronic computers represents an all-time high. It would seem dramatic enough that machines designed solely to perform arithmetic calculations should be able to figure out any kind of a move on the chess board, a process no chess player would look upon as merely numerical. Even if these machines could hold their own against only the rankest beginner, it would be a remarkable feat. To make the claim that given enough time and money to perfect its game, a computer could be designed that would be able to defeat a chess master was as unnecessary as it was absurd, though perhaps no more absurd than to say that a person who does not know chess could learn to play like a master without devoting to the task years of study and practice."

I sat in a large room last Saturday in Philadelphia. Maurice Ashley and Yasser Seirawan were on stage. In the audience, at least 500 people, perhaps more. Hans Berliner, a winner of the World Correspondence Chess Championship and programmer extraordinaire was there. So were other notables: David Levy, Frederick Friedel, Mike Valvo. Most of the others in the audience seemed relatively new to chess, but all were aware that they were witnessing a moment of some profound importance to humanity.

In a quiet room down the hall, the world champion Gary Kasparov, perhaps the greatest human chess player of all time, sat in battle against a silicon behemoth, Deep Blue, a 1,400 pound creation from an IBM laboratory with 256 processors capable of assessing more than 4 million positions a second.

I confess to having no love at all for the computer. I have long known that machines can be made to be stronger and faster than humans, but we are uncomfortable with the idea of permitting motorcycles to race in the 100 yard dash or to have forklifts enter weight lifting competitions. Machines perform admirably on the assembly line, but the real point may be that we choose not to define such conduct as "thinking."

Thinking is, after all, a uniquely human endeavor. A machine taking on the human mind? Again, I quote Lasker:

"All the computer can do is add or subtract, store numbers, transfer them from one part of the machine to another, compare quantities, and perform a few other extremely elementary logical operations. Assigning approximate numerical values to the different pieces is simple enough. We do that in every game whenever a combination involves an exchange of pieces which are not identical.... But how can we in a comparable moment evaluate an open file for a Rook, or a solidly secure central position of a Knight, or the possibility of an exchange which leads to a Pawn majority distant from the opposing King, or more space to move in than the opponent enjoys on his King's or Queen's wing?"

The trip in to Philadelphia was an event in itself. A traffic jam caused by road-repairing machines. Many stalled cars in the February cold. A traffic light stuck on red, and a parking lot gate that refused to open. Omens all?

Before the match, I joined up with two of my former opponents in the US 10th Correspondence Championship, David Rubinsky and Michael Mazoch. We contemplated the possibilities. Like Seirawan, Rubinsky expected 1.d4, possibly also 2.Bg5. I predicted a Sicilian. Better, I think, to go after Gary's main defense, perhaps even challenging the most complex of his lines.

Deep Blue vs Gary Kasparov

Philadelphia, February 10, 1996

1.e4 c5 2.c3

Already, a provocative choice. My personal preference is for 2...e5 here, but Kasparov cannot make that move. It involves a speculative piece sacrifice and fabulous tactical complications. Hardly the right choice for the first game of this six game OTB match.

2...d5 3.exd5 Qxd5 4.d4 Nf6 5.Nf3 Bg4

Developing the bishop before playing ...e6.

6.Be2 e6 7.h3 Bh5 8.0-0 Nc6 9.Be3

And here I smiled. Were he still alive, Lasker might marvel that the computer can get this far, but the fact is that white will now wind up with an isolated pawn on d4 and a bad dark-squared bishop. Kasparov will only need to blockade the pawn and reach an endgame. Can it be that easy for the human?

9...cxd4 10.cxd4 Bb4

A novelty played instantly and, as if playing a human, Kasparov rose triumphantly from the board. Facing a new move from the World Champion, most of us would think for at least an hour out of respect. The usual move is 10...Be7 preserving b4 for the idea of Nc6-d5. But the machine keeps a relentless pace and is incapable of respect. For the first time, it thinks, but only for three minutes.

11.a3 Ba5 12.Nc3 Qd6

Kasparov, clearly still in his preparation, moves quickly. The machine does not vary its pace. Every move is another three minutes.


But here, the world champion stops to think, and think. He considers his reply for more than 20 minutes. Perhaps he'll offer to repeat the position with 13...Qd5 14.Nc3. I think he ought to try this, at least once, since the computer might speculate with b4?! rather than take the draw. At least take the opportunity to learn a bit about its programming.

13...Qe7 14.Ne5 Bxe2 15.Qxe2 0-0

Certainly not 15...Nxe5 16.dxe5 when white can clamp down with Nd6.

16.Rc1 Rac8

Seirawan suggests Rfc8, with the idea of operations on the b-file later. Gary looks confident. In control. The computer still has the isolani and the Be3 appears to be quite a liability.


Gary's eyebrows shot straight up! A sign of admiration? Surprise? Whatever, this has to be the right "idea." After exchanging the dark squared bishop, white may be able to push the d-pawn. Very human-like play from this box of bolts.

17...Bb6 18.Bxf6 gxf6

Certainly not 18...Qxf6 19.Nd7


The d-pawn is immune since 19 ... Nd4? 20 Nd4 Bd4 loses a piece after 21 Qg4.

19...Rfd8 20.Nxb6 axb6 21.Rfd1

The computer "sees" that there isn't enough material to support the idea of Rc3-g3.


Proceeding confidently, Gary cuts off white's access to the g4-square and now threatens the d4-pawn as a result. Black will continue to mount pressure on the pawn with ...Qf6.


A bad queen? Not really. Black really has to be concerned here about d5. Wouldn't it be nice to play ...Nd5 here?

22...Qf6 23.d5 Rxd5 24.Rxd5 exd5

Hardly the sort of pawn structure that I recommend for my students. Gary got up from the board, still walking confidently.


Seirawan takes a poll. How many think Gary is winning? Losing? Losing has the clear edge in the room. It certainly had my vote. Yasser says that black actually has an interesting plan here. On 25...Rd8 26.Qxb6 Rd7, black's isolated d-pawn becomes quite dangerous. But will Gary reject it after seeing 27.Re1 d4 28.Re8+ Kg7 29.Qc5 d3 30.Qf8+ Kg6. The fact is, black's king is quite safe and black is better there.


Now even Seirawan is nervous. Kasparov prefers to attack with Rg8, but everyone in the room senses the danger of giving up even a pawn against the machine's cold and cruel capabilities.

26.Qxb6 Rg8 27.Qc5 d4 28.Nd6

After a short time away from the board, Gary returns without his jacket. For me, the rest of the game was a tragic affair, less for the chess itself than what it meant to Kasparov, and perhaps also for humanity. He is visibly upset here, perspiring heavily.

28...f4 29.Nxb7

No human I know would make this move. The knight was already wonderfully posted and the pawn is unnecessary. How cruel that human intuition by itself is not enough compensation.

29...Ne5 30.Qd5 f3 31.g3 Nd3

Kasparov has less than 12 minutes left to make 9 moves. The computer has more than an hour. Seirawan sees 32.Rc6 and knows that the game cannot be saved. For example, 32...Qf4 33.Rc8 is horrific. "Surely," comes a sound from the audience, "Kasparov can pull a rabbit from his hat." "I'm afraid," replies Yasser, "that the rabbit has died." There's sweat on Kasparov's forehead, around his eyes, on his hands. Mine too.

32.Rc7 Re8 33.Nd6 Re1+ 34.Kh2 Nf2

The threat of Rh1# is inconsequential against precise, cold blue calculation.

35.Nxf7+ Kg7 36.Ng5+ Kh6 37.Rh7+ 1:0

There's a mating net after 37 ... Kg6 38 Qg8 Kf5 39 Nf3.

I am pleased to report that there was no applause. I return again to Lasker:

"I was very much surprised to read a statement, imputed to Botvinnik, that the day will come when computers will thoroughly master chess, and FIDE will have to establish grandmaster and master titles for these machines. This is, of course, nonsense, and I am sure that Botvinnik was either joking or said something totally different."

With all due respect to Lasker, who today is pleased not to be alive, I must point out, as I write this, that Deep Blue has a provisional rating of 3175.