‘Rematch’ and a retro chess problem that baffled Kasparov

“A game starts with e4; on the 5th move, a knight takes a rook and checkmate. What are the moves played?” That question is taken verbatim from the TV mini-series, ‘Rematch’ (2024), a dramatisation of the chess match between world champion Garry Kasparov and the IBM super-computer, Deep Blue. While not as wildly popular as ‘The Queen’s Gambit,’ the show is terrific: a psychological thriller that should engage both chess fans and the general audience. Taking a leaf from the earlier hit series, ‘Rematch’ stresses authenticity in its portrayal of the game, including precise recreations of positions seen in the real match. As a problemist, though, what intrigued me most is a scene in which Kasparov is challenged to solve an unusual chess puzzle – to reconstruct a five-move game as stipulated above. That scene, which I have added to YouTube, even includes an animated diagram to represent his thought process. This is actually a genuine retro-analytical problem that was posed to Kasparov in real life. Let’s look at both the fictional and factual circumstances under which he attempted the problem, with opposite outcomes.

[If you wish to solve the problem yourself, note that its solution appears in the video and also at the end of this blog.]

Fictional struggle

In 1996 and 1997, two matches between Kasparov and Deep Blue garnered great public attention, having been marketed as an ultimate showdown between Man and Machine. ‘Rematch’ reenacts these events, and stars the British actor Christian Cooke as a brooding Kasparov. After watching the series, I wondered how much of what occurred was true-to-life, and found a 2003 documentary that was helpful in that respect: Game Over – Kasparov And The Machine (free on YouTube). The program confirms many plot points; for instance, Kasparov really did allege that the IBM team cheated in Game 2 with some sort of human intervention. Another source of contention, IBM’s refusal to release “private” games played by Deep Blue, is accurate as well.

The context of the chess problem scene is that after Game 3, Kasparov felt so much pressure that he contemplated withdrawing from the match (also true apparently). He disappeared from the playing hall and a GM player from the IBM team was tasked with finding him. This character, Xavier Valens – not based on a particular person – tried to goad him with a difficult puzzle, as a sort of reminder that the match is less about Kasparov and more about chess itself.

My favourite part of the clip is Kasparov’s visualisation of the moves as he tries to reconstruct the five-move game. The animated diagram shows plenty of plausible tries or near-misses that almost accomplish the goal; e.g. his first attempt is 1.e4 f6 2.Sf3 Kf7 3.Sh4 Qe8 4.Sg6 e6 5.Sxh8+, which fails only because the king could escape to e7. Presumably IM Malcolm Pein, the chess consultant for the show, was responsible for incorporating this remarkable level of detail on the screen. On the other hand, what might elicit a groan is that when Kasparov solves the problem, the camera pans away from the board and the final moves are not shown. He announces the solution correctly, but it could be missed by viewers not familiar with move notation. I like Valens’ speech at the end about how the focus should be on chess, not Kasparov, but it’s a bit ironic when a few moments ago the show itself shifts its attention from the chessboard to the person at a critical point.

Factual frustration

The real story of how Kasparov – not to mention Karpov, Botvinnik, and their top students – failed to solve this problem is perhaps even more entertaining. The person who “tormented” them with the task is Frederic Friedel, a chess-computer expert and co-founder of ChessBase. He served as an advisor to Kasparov during his match with Deep Blue (a clip of him as an interviewee can be seen in the documentary mentioned).

As Friedel recounts in his article, The infamous 1999 ChessBase Christmas Puzzle, the problem was shown to him by the British GM John Nunn:

In 1986, during the turmoils after Kasparov had won the world championship and was forced to face an immediate rematch, both he and Karpov went to Lucerne to meet with FIDE President Campomanes. I was with them and we had a long car journey together from Zurich to Luzerne. To entertain them I gave the two top players in the world John’s puzzle. It kept them busy during the ride and for the next couple of days in the hotel. They couldn’t solve it.

Before we parted I did the Nunn thing on Garry: I sealed the answer in a hotel envelope and told him to return it unopened with the solution. I didn’t hear from him for many months. Then one day I came home and found a number of messages with a phone number where I should call Kasparov urgently. I did so and found him in a distraught state. “You are a dead man, Fred,” he said, “you have put me in a very embarrassing situation.” Turns out he was running a session of his chess school, together with Mikhail Botvinnik, and he had given the problem to his students. When they couldn’t solve it and asked him for the answer he had told them how important it was to persevere. They should not give up but try for another day. Meanwhile the hunt was on for the envelope, which unfortunately could not be located. When I told him the solution on the phone I could hear Botvinnik gasp in the background. And Garry, who like every one else in the chess school, was convinced I had stated the problem incorrectly, couldn’t believe that he and his students had missed it.

Friedel retells the story with more amusing details in (1) another ChessBase article: Knowing the chess greats - Part I, and (2) a YouTube video by the same name (timestamped at 19:52 for the relevant part).

Source (?) and solution

What is the origin of this short but tough retro problem? It belongs to a genre called synthetic games, in which the solver controls both sides to attain a position that meets a certain objective. The best-known example of the type is the fastest game to stalemate, achieved in ten moves by Sam Loyd. In 1998, the retro expert George Jelliss published Synthetic Games, a large collection of such problems (free download from his site). However, the five-move game that stumped Kasparov is not found in this book.

Since Friedel wrote that the task was given to him by John Nunn, I contacted John and asked if he’d composed it himself. He kindly replied that he did not, and is not aware of who did. Still, he provided a nice anecdote about how he came across the problem:

It was shown to me (and other players) at the 1976 Haifa Olympiad and proved quite a distraction. Some people were amazed by the sight of several grandmasters staring with total concentration at a board on which 1.e4 had been played!

The solution is 1.e4 Sf6 2.f3 Sxe4 3.Qe2 Sg3 4.Qxe7+ Qxe7 5.Kf2 Sxh1 mate. Like most synthetic games, the play is not quite unique; we can swap the order of two white moves, f3 and Qe2, but that’s the only alternative. Although the aim to mate the white king on f2 with a knight’s capture of the h1-rook seems reasonably clear, it is surprising how White doesn’t self-block the king’s escape squares on the e-file, but instead helps clear the file for the black queen to control them. The task could be made slightly easier by specifying that the mating move is executed by Black. Conversely, we could make it even harder by asking for the shortest game that meets the same conditions, without the hint that the answer takes five moves. Such a version of the problem, however, may be too diabolical for practical solving.