After more than a year of preparation , an epochal match began, which became an important milestone in the history of chess.

Neither before nor after 1997 the fights between the computer and the person did not attract such public attention.

The stunning result was perceived by the general public as the beginning of an era of new champions.

In the photo - the first moves in the first game of the new match.
')

1st party. Kasparov - Deep Blue. 1: 0

In the first game, Kasparov got a white color - this meant that the latter would have to play with black. It was unprofitable for Harry, because in the case of a draw or a losing account before the final game would have to play to win in black. However, lots are lots.


Kasparov is already surprised from the start - he plays Reti's extravagant debut, and even a strange system with double fianchetto (see diagram). This, it seems, has not yet met in his practice ... Thus, Harry showed that he was really going to play "anti-computer chess". This strategy involves the immediate withdrawal from the popular and well-analyzed opening lines, non-standard maneuvering in one half of the field, a viscous game with a long accumulation of unobvious positional advantages, which are understandable to top-class players, but hardly recognized by the machine that evaluates a position with a given pattern.

The party Kasparov played in the best traditions of hypermodernism. The peculiarity of this game was that not one of Kasparov's figures ever visited the “alien” half of the field (two pawns advanced very far, but they are not figures). And nevertheless, it was an atypical maneuvering within its own camp and brought a deserved victory.


Although the game was completely won, at its end something happened that affected the outcome of the match not in favor of Kasparov. Look at the chart. Black affairs are bad - white pawns cheerfully march to the fields of transformation and will soon become queens. It is necessary to take urgent measures and the most stubborn continuation would be the obvious 44. ... d5-f5 + (indicated by the blue arrow). In this case, the blacks did not escape anyway, but this was the most stubborn method of resistance, and you could still snap at the moves of 15. Instead, the computer played an instantly losing move 44. ... d5-d1 (red arrow). White pawn went forward: 45. g6-g7 . The team Deep Blue decided not to disgrace further and surrender.

Kasparov and his seconds had a reasonable question: why did the car play so poorly and did not choose a stronger way to lose? Can't a supercomputer count? An incorrect conclusion was made that the car chose such a weak move, as it considers it too far-sighted in the case of 44. ... Ld5-f5 + the computer probably found the option where it gets mated after 20 or more moves. Since there is nothing more terrible than the mat, the “dark blue” and chose to play much weaker, but with more distant matte threats. As we will see later, this false conclusion to some extent influenced the result of the next game, and at the same time the result of the whole match.

Meanwhile, the cause of the computer error was different. Even during the preparation for the match, an interesting incident was noticed - the car does not make a difference between a very bad and very, very bad position. If a computer assesses its position as clearly losing, and sees that it can’t be saved, then it chooses not the most stubborn sequel, but just throws a coin and plays at least something. Deep Blue was created not by chess players, but by programmers, and they found such logic quite natural. Immediately after this game, a bug was fixed. Now, even in the case of the most hopeless development of events, the computer, on the basis of its estimates and calculations, went along the most unpleasant path for an opponent. But Kasparov did not know all this.

2nd party. Deep Blue - Kasparov. 1: 0

Spanish party! Kasparov is one of the world's largest connoisseurs of debut and plays it well in both colors. With the same Karpov, they played more than a dozen games with this beginning ... The computer played White in a strict academic manner, but Kasparov, made a blot, weakening his queenside by moving the c -pawn. The car made a delicate mistake, revealing and capturing the extreme vertical. In this case, the invasion on the left edge of the board computer combined with non-standard thin maneuvers on the kingside and center.


Immaculately beat the world champion, at the very end Deep Blue makes two gross mistakes. First, in response to the Shah, the king retreats toward the center (and you had to hide in the corner). But the winnings have not yet been lost. And by the next move, the car gives Kasparov a chance to make a draw. Instead of exchanging queens and moving into a winning endgame, the computer recklessly attacked the queen's rook (see diagram).

Here it is - salvation! Now an unexpected jump by the queen to the camp of the white 45. ... b6-e3 (green arrow) allowed Garik to escape unscathed from the soulless chess killer. The white king cannot hide from the subsequent shahs. Analysis of the position shows that White could try to create a kind of bunker, moving the pawn from h3 to h4 , then transfer the king to h3 , and cover it with the g2-g3 pawn and the bishop Ce4-g2 . But the black pawn moves h6-h5 ensured that the queen could also be given the queen from g4, and the white king would only have a rest.

And then an incredible thing happened: Kasparov extended his hand to the operator Deep Blue and congratulated on the victory! As Harry Kimovich later explained, he perfectly saw the Queen's invasion. However, he considered that the computer couldn’t fail to notice such a chance, which means that there is no perpetual check there, the white king will still hide safely. Such a conclusion was influenced by the final of the previous game, as a result of which Kasparov strengthened the false confidence that the computer was calculating the options very far.

When immediately after the game it turned out that there was an easy draw, Kasparov suspected IBM of dishonesty. Subsequently, he repeatedly accused the team of Deep Blue, primarily based on the unsuccessful rook turn in this game. The supercomputer could not fail to see such an obvious option, missing the victory. And yet, nevertheless, he allowed this development of events, which, in Harry's opinion, indicates that the chess machine did not do all the moves independently.

Perhaps, sometimes the decision to choose the next move is not made by the computer, but by one of the grand consultants? A computer would never allow an eternal check, but, for example, Joel Benjamin could not have noticed the salvation for the black ones, just as Kasparov, who was exhausted by the end of the game, did not notice.

Fen Xu in his book explains in detail the last two erroneous computer moves in this game.

The unfortunate departure of the king from the Shah 44. Kg1-f1 (instead of the correct 44. Kg1-h1 ) was due to the fact that he interpreted his position on the board as being in the process of a close transition to the endgame. In this case it is really better for the king to stay closer to the center of the board - after all, for massive exchanges, the passage of the remaining pawns to the queens is mostly supported by the king. But there is still no endgame and the king should be led to the corner of the board, after which the blacks would not have the possibility of an endless counterattack.

As for the move 45. La8-a6 , finally releasing the win, here is another story. Seeing the “simple” perpetual check is not so easy for a computer. The black queen can play with a sufficiently large number of fields, and the white king can move away from the attack in various directions. Alternate these attacks, cares can be different. A person at a glance at the board understands what's what, but the computer that honestly goes through all the moves unexpectedly encounters a huge tree of options, in which the queen and the king play the same fields (but in different sequences!) mice. " And in order to see the notorious "eternity", the car would have to look deep into about 30 semi-moves. In general, Deep Blue was capable of this, but it would take him more than an hour to go through all the emerging branches. However, programmers programmed “dark blue” for a more rational treatment with time, and up to the eternal pursuit the computer did not count.

Anyway, it was a turning point. According to Harry, the overall result was predetermined in this meeting.

"That beats me. After the second game the match was almost over. ”

3 party. Kasparov - Deep Blue. ½: ½

Kasparov played White strongly, but cautiously.


The computer got a cramped position, did outwardly coarse moves, which in fact are impeccable in terms of defense. Laughing and not finding the win (which was not), the world champion agreed to a draw.

4 party. Deep Blue - Kasparov. ½: ½

Kasparov again surprised by the choice of the debut for black - the defense of Pirc-Ufimtsev, which, like, in black, he never played. In general, Harry managed to replay the car, but the computer stubbornly clung to every opportunity and found all the new resources to hold the position.

At some point, Kasparov played inaccurately and the cyborg managed to wriggle out of the tenacious paws of homo-sapiens. If black, instead of exchanging queens, doubled the rooks on the half-open vertical “f” , White would have to do badly: their pieces were completely disorganized. In the diagram, the winning position for black, which could have occurred if the world champion had gone the right way. White has a pawn more, but their pieces are one worse than the other. At that time, as the black troops took the ideal position. But fortunately for Deep Blue and unfortunately for Kasparov, this option did not take place in reality.

5 party. Kasparov - Deep Blue. ½: ½

Kasparov urgently needs to win - the score is still equal, and his last time is white. Again, the extraordinary "anti-positional" debut of Reti. The computer was surprised already on the 4th move, exchanging the bishop for the knight. The fact is that the electronic chess players of that time gave undoubted preference to the long-range elephants, changing them for little predictable horses only when absolutely necessary. This reinforced Kasparov’s suspicions that playing with him is not fair. “I understand something in chess and computers. This is not a computer game! ”- he said after the game. Although in this game such an early exchange was appropriate, as it allowed Black to quickly develop other pieces.


In a sharp ending, White became the masters of the right side of the board and the passed g -pawn gave a green light. It seemed that Kasparov, now finally win and lead in the long run. Black, his small but well-coordinated forces threw in the last attack, which was personally led by the black king. The final position (on the diagram) is quite spectacular - White can even put the queen right now, but this will not help them. A small but friendly black squad announces the eternal check to the white autocrat.

The miraculous rescue of a computer in this game has completely demoralized Kasparov. No wonder that the last game was a disaster for him.

6th party. Deep Blue - Kasparov. 1: 0

Kasparov again chooses a debut for blacks that is absolutely atypical for himself. This time he applied the protection of Tarakan Caro-Kann. White Kasparov always gladly smashed "karokanchikov" (including Karpov), but for Black this debut is not included in his repertoire.

This beginning allows Black to build a strong but passive position. White is given space, there are plenty of opportunities for tactical attacks, traps and sacrifices (if black will play carelessly, then in the first moves they can even get a checkmate). It was to one of these victims that Kasparov decided to provoke his electronic counterpart.


Harry just looked like a pawn 7. ... h7-h6 (on the diagram), driving away the frisky horse. The theory recommends that whites sacrifice a dashing steed, receiving in return a very dangerous attack. After that, it is extremely difficult for Black to hold the position, but, in principle, it is possible. If, however, it is possible to repel a squall attack, then there is an opportunity to stay with an extra piece and eventually win. Apparently, with such an ingenious plan, Kasparov hoped to still end the match in his favor.

The computer accepts the challenge and gives up the horse to the slaughter - 8. g5: e6 . And the next move Harry makes a crucial mistake. Instead of immediately taking the knight by pawn, he makes an intermediate move by the queen 8. ... d8-e7 , after which he cannot be saved (see the following diagram).


It looks good on a superficial glance, as it allows the black king to take refuge on the queenside. However, it turns out that there is much easier to attack him. If the black king sought salvation in the opposite direction, it would be more difficult to get to him. In addition, the black queen violates the coordination of his own pieces, in particular, does not allow the bishop to be quickly withdrawn from the f8 field, which in turn locks the rook to h8 . As well as the vizier himself can make a successful attempt, which happened in this party.

Deep Blue won with extreme cruelty. On the 18th move, the black queen fell, on the 19th person he admitted defeat. Although Kasparov still has enough material on the board, it’s time to give up. Numerous black troops are completely disoriented and interfere with each other. White calmly alternates between the matt attack and the systematic devouring of enemy pawns and pieces.


The final position of the party and the match. Black's move, but it's hard for them to give good advice.

“This machine has not proved anything to anyone yet!”

At the final press conference, Kasparov threw metal thunder and lightning, hinting that he did not play against the computer, but against the bunch “computer + grandmasters”, which did not leave him the slightest chance.

What made the frankly weak moves in the first and second installments if, obviously, there were much stronger alternatives?
Why did the machine first play poorly and then devilishly ingeniously in games that ended in a draw?
Where did the computer operator go before making some moves in the 4th game?
What are the considerations in the 5th game that the computer so instantly parted with his favorite bishop on the 4th move?
In this connection, after the 5th installment, the printout of the technical protocol was made so late?
How so in the 6th game the computer, without hesitation, went to sacrifice the horse, after all this (according to the theory) does not guarantee victory?
How do you explain that at a critical moment in one of the batches, the operator started typing something very quickly on the keyboard?
Why did the mistakes of Deep Blue not allow the very strong commercial chess program Fritz, when it was given to the analysis of the game of this match?


Last batch. Only a dozen moves have been made, but the loss of black is already inevitable.

After some time, when several emotions cooled down, Kasparov was already refraining from accusations, and offered to arrange another match. The champion complained that he did not have enough time to prepare for the competition. I also complained about the too intense schedule of games, which does not matter for a supercomputer, but is heavy for a person. Yes, and 6 rounds - too short a distance, and Harry proposed to organize a new match, but with more parties (8 or 10). At the same time, he agreed on the prize for the “all or nothing” formula, that is, in the event of a loss, he would not claim any money for participation. And even expressed willingness to compete with the supercomputer to put at stake the most expensive - his champion title.

Harry's arguments seemed convincing to many and the chess community believed that IBM owed humanity another one, this time a decisive match. But the third round did not take place, primarily due to financial considerations. The victory over Kasparov brought the company huge image dividends, shares of the corporation rose in price. But if the computer wins in the next confrontation, then there will be no such powerful public resonance. Repeated victory of the computer is unlikely to be perceived by the public as an extraordinary event. But a possible defeat will negate all marketing gains from winning the second match. Holding an expensive event once again did not promise big benefits, moreover, it carried the risk of loss. Therefore, it was decided that even if Deep Blue is guaranteed to remain the winner of Kasparov.

Kasparov quite painfully perceived an insulting defeat and the refusal of IBM to play another match, even more so against the background of his own suspicions of plotting against him. The next time Harry agrees to fight with a robot many years later. This is the next story.

( To be continued ... )

Additionally

Match table

	one	2	3	four	five	6	Result	Glasses
Deep blue	0	one	½	½	½	one	+2, –1, = 3	3.5
Garry Kasparov	one	0	½	½	½	0	+1, –2, = 3	2.5

Debut statistics

Reti's Debut - 1, 5
Spanish Party - 2
English beginning - 3
Pirc-Ufimtseva protection - 4
Caro-Kann Defense - 6

Move the figures in batches here .
Good Wikipedia article on Deep Blue.


Many thanks to Roman Zhukov (@ Rom77) for the translation of Fen Xiu's book “Following the Deep Blue” (“Behind Deep Blue”). Without this source, many of the nuances that shed light on what was happening behind the scenes of a chess theater would be completely unclear.

Here you can download this book in Russian translation.

PS And also thanks to Roman for the fact that he is again in our literary ranks .

PS2. A year later, Boris Schipkov in the open spaces of GeekTimes expressed an alternative point of view about this match .