Kasparov and the Machine
Though the film came out in 2003 and has been available on DVD in Europe for over a year, "Game Over: Kasparov and the Machine" has only just become available on DVD here in the U.S. of A.
The film takes a dramatic look back at Kasparov's infamous 3.5-2.5 1997 match loss to IBM's "Deep Blue" computer from a largely pro-Kasparov perspective. The two culprits of the match loss, from Kasparov's point of view, were (alleged) human intervention and the IBM team's psychological warfare at his expense. I'll discuss each in turn.
1. Human Intervention?
In 1996, Kasparov lost game 1 of his match with an earlier version of Deep Blue, won game 2, and after a pair of draws, won two more games against the computer, and very easily at that. Having won the match 4-2, and having then won the first game of the 1997 match in good style, things seemed to be completely under control.
Then came game 2. The computer outplayed Kasparov on the White side of a main line Ruy Lopez, and after move 26 the following position arose:
Black is in trouble here, as he is getting squished in a position devoid of counterplay. Kasparov has a moment of hope here, as White can now win a couple of pawns with 36.Qb6 Rd8 37.axb5 Rab8 38.Qxa6, but this will allow Black to gain some counterplay with 38...e4. It may or may not be enough, but it gives Black a chance to save the game. Note: there's nothing clearly wrong with 36.Qb6. There aren't any hidden tactical points at the end of a combination that win for Black or even force a draw. White really does win material and maintains an overall advantage, at least as far as the computer can tell; it's just that Black gets some counterplay as a result.
Instead, Deep Blue played 36.axb5 axb5 37.Be4!, immediately snuffing out Black's hopes for counterplay by preventing the ...e4 advance and thereby sealing in the Black pieces. It's an uncontroversially strong move, but what makes it remarkable here is that the computer prefers it to the safe win of two pawns!
That was disturbing enough, but things got worse.
In this position, Deep Blue played the seemingly safer 44.Kf1, preventing the Black queen from coming in to f2, and after 44...Rb8 45.Ra6 Kasparov resigned.
Unfortunately, as Kasparov found out in the evening between games two and three, Black has a draw in the final position: 45...Qe3! 46.Qxd6 Re8! 47.h4 h5! 48.Bf3 Qc1+ 49.Kf2 Qd2+ 50.Be2 Qf4+ with perpetual. Paradoxically, 44.Kh1 would have won easily, as 44...Rb8 45.Ra6 Qe3 46.Qxd6 Re8 47.Ra1! Qxe4 48.Ra7+ Kg8 49.Qd7 is terminal, but after 44.Kf1 it doesn't work: 47.Ra1 Qxe4 48.Ra7+ Kg8 49.Qd7 Qd3+ with a perpetual.
Between these two surprises - the computer's playing a deeply human move on move 37 and then missing a "simple" perpetual check a few moves later - Kasparov concluded that something was rotten in IBM land. Likening Deep Blue's play in game two to Maradona's famous "hand of God" goal in the 1986 World Cup - a blatant accusation of cheating - Kasparov for the first time requested Deep Blue's game two "logs" - printouts of the computer's thinking - in order to make sense of these two shocking moments.
Psychological Warfare
The IBM team was understandably angered by Kasparov's accusation (assuming, of course, that they're innocent!), and was a bit less than forthcoming with those logs. One reason was a concern that it might give Kasparov sufficient insight into the way the computer evaluated different positions, but it also became a way of further unbalancing Kasparov's psyche, as they actually admit. What Kasparov had thought was a friendly match, done in the interest of science (as well as mutual pocket-lining, no doubt), had turned into a mega-corporation's assault on him, the victim, in IBM's efforts to win at any price.
This was Kasparov's perception, at least, and the filmmakers seem happy to encourage this view as well. One silly technique, repeated ad nauseaum, are various shots of a replica of the "Turk", a 19th century gizmo that purported to be a chess-playing machine, but was in fact operated by a human scrunched up in a small compartment under the board. (Just the way Deep Blue seemed to be the one playing 37.Be4, but "really" wasn't. Get it? Subtle, right?)
Film-making techniques aside, do we have any reason to accept this perspective on what happened as the objective truth of the matter? To take the easier part first, I think the psychological warfare charge can be accepted straight away, as even the IBM team acknowledges that they were trying to psych him out. It's a bit funny to me that Kasparov is complaining about such a thing, however, as he was known throughout almost his entire career for both his psychological toughness and for himself psyching out many of his opponents.
(An example: I once saw him give a simul to a group of 30 or 40 players with ratings under 1800 - most of them young children - and even in that setting he worked himself up into a competitive fervor beforehand, pacing like a caged lion. For the first go-round, he'd shake hands and flash his smile on cue for the pictures, but after that his expression was deadly serious. Sometimes he'd even glare, as if he were playing not kids with ratings in the 1400-1600 range, but a team of young Azeri GMs.)
About the cheating charge, I find it very unlikely, and certainly not well-substantiated by the limited evidence of the two iffy moves. One complaint Kasparov raises is that in game 1, the computer showed little regard for king safety, but in game two, things had changed.
Talk of "king safety" as an evaluation function seems to me far too coarse-grained. There are thousands of factors that go into the computer's evaluation of a given position, none of which translates directly into our general concept of king safety. The computer will handle some positions in a way we'd deem overly cautious with respect to its king, and on other occasions strike us as foolhardy. The types of positions in games 1 and 2 were just too different to draw reasonable comparisons, in my opinion.
And speaking of king safety, how are to account for White's blundering into a potential perpetual check? I'm afraid it's just our old friend, the horizon effect, an example of which can be seen in this recent post. As in that study, White can drag out the perpetual through a number of permutations before any position gets repeated a third time. The computer can't just "see" that a repetition will be forced; it has to calculate that the position will occur. Altogether, I believe it takes around 40 ply after 43...Qb6+ to force a three-time repetition, and that's a huge search depth even for Deep Blue.
The aftermath of this game was quite strange, and not adequately dealt with at all by the film. One gets the sense that the match was essentially lost at this point, and that Kasparov's terrible game 6 loss, after three intervening draws, was the inevitable result of IBM's psychological warfare. Yet Kasparov had very good winning chances in each of games 3-5, especially in game 4. Had he won any of those games, the match would have almost surely ended in at least a draw, and recent chess history might have been very different.
Is the DVD worth getting? That depends. It's certainly no documentary, it's short on hard facts, and the repeated cut-away shots of the Turk leave me a bit unsure as to whether the director has a cornball approach to movie-making, thinks the audience is really too stupid to get Kasparov's charge, or is mocking Kasparov by showing images of skullduggery everywhere.
If you're interested in seeing some historical footage (including some clips of Kasparov from the 1980s), listening to the principals reminisce about a very high-profile moment in chess history, and want to encourage filmmakers to believe that there's an audience for chess movies, then buy or rent it.
The DVD comes with a few extras: a director's commentary, a very brief "biography" of Kasparov, a painfully slow replay of the match's games with what I assume is ChessMaster's voiced commentary, and 8 hours' access to a restricted version of ChessMaster 10000.
(Long aside: It's a very strong engine, but if you're in the market for a chess engine, my recommendation is for Shredder 9, for three reasons.
First, Shredder has dominated the SSDF rating lists for years - see here.
Second, while both CM 10000 and Shredder connect you to chess servers, there's no comparison between the two on that score. The ChessBase (the parent company of Shredder) playchess.com server is very well-attended and titled players are there just about 24 hours a day - even Anand and Kasparov have shown up, and I've played Mickey Adams, Yasser Seirawan, Hikaru Nakamura and other strong GMs there. They also broadcast live events and feature delightfully entertaining instructional shows (like mine!).
My trip to the CM 10000 online server, on the other hand, was like whistling in a ghost town, but, to be fair, it might be that those on the 8-hour plan are restricted from seeing all the real action. (If that's the case, please let me know, as while I hope to make the best case I can for ChessBase over ChessMaster, I don't want to do so at the cost of the truth.)
Third, there's a compatibility issue. ChessBase is the leader in chess software, and by purchasing Shredder, you have the ability to handle and interact with any other ChessBase databases you may purchase. I assume (but don't know) that CM can handle .pgn files, but I'd be shocked if it could handle .cbv and other ChessBase files.
BIG DISCLAIMER: I work for ChessBase! That said, I purchased lots of ChessBase software for many years prior to joining them, so the foregoing is much more a case of my putting my mouth where my money is, not shameless corporate shilling. Nevertheless, I believe in full disclosure, so take that into account.)
The film takes a dramatic look back at Kasparov's infamous 3.5-2.5 1997 match loss to IBM's "Deep Blue" computer from a largely pro-Kasparov perspective. The two culprits of the match loss, from Kasparov's point of view, were (alleged) human intervention and the IBM team's psychological warfare at his expense. I'll discuss each in turn.
1. Human Intervention?
In 1996, Kasparov lost game 1 of his match with an earlier version of Deep Blue, won game 2, and after a pair of draws, won two more games against the computer, and very easily at that. Having won the match 4-2, and having then won the first game of the 1997 match in good style, things seemed to be completely under control.
Then came game 2. The computer outplayed Kasparov on the White side of a main line Ruy Lopez, and after move 26 the following position arose:
Black is in trouble here, as he is getting squished in a position devoid of counterplay. Kasparov has a moment of hope here, as White can now win a couple of pawns with 36.Qb6 Rd8 37.axb5 Rab8 38.Qxa6, but this will allow Black to gain some counterplay with 38...e4. It may or may not be enough, but it gives Black a chance to save the game. Note: there's nothing clearly wrong with 36.Qb6. There aren't any hidden tactical points at the end of a combination that win for Black or even force a draw. White really does win material and maintains an overall advantage, at least as far as the computer can tell; it's just that Black gets some counterplay as a result.
Instead, Deep Blue played 36.axb5 axb5 37.Be4!, immediately snuffing out Black's hopes for counterplay by preventing the ...e4 advance and thereby sealing in the Black pieces. It's an uncontroversially strong move, but what makes it remarkable here is that the computer prefers it to the safe win of two pawns!
That was disturbing enough, but things got worse.
In this position, Deep Blue played the seemingly safer 44.Kf1, preventing the Black queen from coming in to f2, and after 44...Rb8 45.Ra6 Kasparov resigned.
Unfortunately, as Kasparov found out in the evening between games two and three, Black has a draw in the final position: 45...Qe3! 46.Qxd6 Re8! 47.h4 h5! 48.Bf3 Qc1+ 49.Kf2 Qd2+ 50.Be2 Qf4+ with perpetual. Paradoxically, 44.Kh1 would have won easily, as 44...Rb8 45.Ra6 Qe3 46.Qxd6 Re8 47.Ra1! Qxe4 48.Ra7+ Kg8 49.Qd7 is terminal, but after 44.Kf1 it doesn't work: 47.Ra1 Qxe4 48.Ra7+ Kg8 49.Qd7 Qd3+ with a perpetual.
Between these two surprises - the computer's playing a deeply human move on move 37 and then missing a "simple" perpetual check a few moves later - Kasparov concluded that something was rotten in IBM land. Likening Deep Blue's play in game two to Maradona's famous "hand of God" goal in the 1986 World Cup - a blatant accusation of cheating - Kasparov for the first time requested Deep Blue's game two "logs" - printouts of the computer's thinking - in order to make sense of these two shocking moments.
Psychological Warfare
The IBM team was understandably angered by Kasparov's accusation (assuming, of course, that they're innocent!), and was a bit less than forthcoming with those logs. One reason was a concern that it might give Kasparov sufficient insight into the way the computer evaluated different positions, but it also became a way of further unbalancing Kasparov's psyche, as they actually admit. What Kasparov had thought was a friendly match, done in the interest of science (as well as mutual pocket-lining, no doubt), had turned into a mega-corporation's assault on him, the victim, in IBM's efforts to win at any price.
This was Kasparov's perception, at least, and the filmmakers seem happy to encourage this view as well. One silly technique, repeated ad nauseaum, are various shots of a replica of the "Turk", a 19th century gizmo that purported to be a chess-playing machine, but was in fact operated by a human scrunched up in a small compartment under the board. (Just the way Deep Blue seemed to be the one playing 37.Be4, but "really" wasn't. Get it? Subtle, right?)
Film-making techniques aside, do we have any reason to accept this perspective on what happened as the objective truth of the matter? To take the easier part first, I think the psychological warfare charge can be accepted straight away, as even the IBM team acknowledges that they were trying to psych him out. It's a bit funny to me that Kasparov is complaining about such a thing, however, as he was known throughout almost his entire career for both his psychological toughness and for himself psyching out many of his opponents.
(An example: I once saw him give a simul to a group of 30 or 40 players with ratings under 1800 - most of them young children - and even in that setting he worked himself up into a competitive fervor beforehand, pacing like a caged lion. For the first go-round, he'd shake hands and flash his smile on cue for the pictures, but after that his expression was deadly serious. Sometimes he'd even glare, as if he were playing not kids with ratings in the 1400-1600 range, but a team of young Azeri GMs.)
About the cheating charge, I find it very unlikely, and certainly not well-substantiated by the limited evidence of the two iffy moves. One complaint Kasparov raises is that in game 1, the computer showed little regard for king safety, but in game two, things had changed.
Talk of "king safety" as an evaluation function seems to me far too coarse-grained. There are thousands of factors that go into the computer's evaluation of a given position, none of which translates directly into our general concept of king safety. The computer will handle some positions in a way we'd deem overly cautious with respect to its king, and on other occasions strike us as foolhardy. The types of positions in games 1 and 2 were just too different to draw reasonable comparisons, in my opinion.
And speaking of king safety, how are to account for White's blundering into a potential perpetual check? I'm afraid it's just our old friend, the horizon effect, an example of which can be seen in this recent post. As in that study, White can drag out the perpetual through a number of permutations before any position gets repeated a third time. The computer can't just "see" that a repetition will be forced; it has to calculate that the position will occur. Altogether, I believe it takes around 40 ply after 43...Qb6+ to force a three-time repetition, and that's a huge search depth even for Deep Blue.
The aftermath of this game was quite strange, and not adequately dealt with at all by the film. One gets the sense that the match was essentially lost at this point, and that Kasparov's terrible game 6 loss, after three intervening draws, was the inevitable result of IBM's psychological warfare. Yet Kasparov had very good winning chances in each of games 3-5, especially in game 4. Had he won any of those games, the match would have almost surely ended in at least a draw, and recent chess history might have been very different.
Is the DVD worth getting? That depends. It's certainly no documentary, it's short on hard facts, and the repeated cut-away shots of the Turk leave me a bit unsure as to whether the director has a cornball approach to movie-making, thinks the audience is really too stupid to get Kasparov's charge, or is mocking Kasparov by showing images of skullduggery everywhere.
If you're interested in seeing some historical footage (including some clips of Kasparov from the 1980s), listening to the principals reminisce about a very high-profile moment in chess history, and want to encourage filmmakers to believe that there's an audience for chess movies, then buy or rent it.
The DVD comes with a few extras: a director's commentary, a very brief "biography" of Kasparov, a painfully slow replay of the match's games with what I assume is ChessMaster's voiced commentary, and 8 hours' access to a restricted version of ChessMaster 10000.
(Long aside: It's a very strong engine, but if you're in the market for a chess engine, my recommendation is for Shredder 9, for three reasons.
First, Shredder has dominated the SSDF rating lists for years - see here.
Second, while both CM 10000 and Shredder connect you to chess servers, there's no comparison between the two on that score. The ChessBase (the parent company of Shredder) playchess.com server is very well-attended and titled players are there just about 24 hours a day - even Anand and Kasparov have shown up, and I've played Mickey Adams, Yasser Seirawan, Hikaru Nakamura and other strong GMs there. They also broadcast live events and feature delightfully entertaining instructional shows (like mine!).
My trip to the CM 10000 online server, on the other hand, was like whistling in a ghost town, but, to be fair, it might be that those on the 8-hour plan are restricted from seeing all the real action. (If that's the case, please let me know, as while I hope to make the best case I can for ChessBase over ChessMaster, I don't want to do so at the cost of the truth.)
Third, there's a compatibility issue. ChessBase is the leader in chess software, and by purchasing Shredder, you have the ability to handle and interact with any other ChessBase databases you may purchase. I assume (but don't know) that CM can handle .pgn files, but I'd be shocked if it could handle .cbv and other ChessBase files.
BIG DISCLAIMER: I work for ChessBase! That said, I purchased lots of ChessBase software for many years prior to joining them, so the foregoing is much more a case of my putting my mouth where my money is, not shameless corporate shilling. Nevertheless, I believe in full disclosure, so take that into account.)
