Remember Like a Grandmaster(?)
It's common to think that grandmasters have colossal, world-class memories. Although Adrian de Groot and others have to some extent debunked this*, it does seem true that GMs have outstanding domain-specific memories.** This is true, but shouldn't be exaggerated.
This thought was occasioned by two stories I came across today. The first came as I browsed the contents of ChessBase Magazine 132, in the section on tactics. IM Oliver Reeh recalls compiling the puzzles for the issue, when GM Leonid Kritz showed up. Reeh showed him an example starting 1.d4 d5 2.e3 Bf5 3.c4 c6 4.Qb3 Qb6 5.cxd5 Qxb3 6.axb3 Bxb1 7.dxc6 Be4, and now Kritz said "I know that motif!" Reeh replied "I should think so. After all, you were the one who played this game!" - which Kritz promptly denied! A quick browse of the database indicated that he had, albeit 11 years earlier.
The second story is buried in the notes to Lupulescu-Marin (here), and doesn't involve any such distant time span. In an earlier issue of ChessBase Magazine, he had offered a particular move as an improvement, but in his own game thought for a long time and failed to realize that he could immediately transpose into his own analysis.
This doesn't only happen to "ordinary" GMs either. The great Mikhail Tal tells one or two stories like this in his chess autobiography as well. And of course, it happens all the time to those of us further down the food chain. So while there are stories of Fischer remembering blitz games from 14 years earlier or Anand remembering ultra-complex ten year old analysis***, the truth is far more complex. There's no escaping it: errare humanum est.
* The standard experiment for this conclusion went roughly like this: they compared GMs, masters, and veritable novices on their ability to reconstruct a normal chess position they had seen for a few seconds. GMs and masters did very well, while novices had no real success. Then they had all three groups try to reconstruct nonsense positions, and this time there was no edge, or at least no appreciable edge, possessed by the stronger players. If GMs had some sort of "photographic" memory, this would not have been the case.
** In other words, they have developed the skill to remember chess positions, moves, ideas, etc. The stock explanation is that this occurs by "chunking"; that is, by compressing many distinct bits of chess information into a single concept. (A chess example: the formation of a white bishop on g2, king on g1, knight on f3 with pawns on h2, g3 and f2 might be seen as a single unit - one chunk - rather than as six distinct units. A universal example would be a word in a language. We process words as units, not as collections of letters.)
*** The Fischer story is that in 1971, at the time of his match with Taimanov, he showed Vasiukov (Taimanov's second) some speed games they had played in 1957 or 1958 when Fischer had visited Russia. The Anand reference is to his great win over Adams' Zaitsev Variation in the San Luis world championship; the analysis had been prepared for Kamsky a decade earlier.
Yet even these examples may not be quite as legendary as they may at first seem. For one thing, Fischer may have reflected on the games he played in Moscow, taking notes about the openings if nothing else. That process of study and overlearning makes memorization far less mysterious. I've also heard that the story has been exaggerated, that he didn't demonstrate all the moves of all the games, but bits of some of them. As for Anand's analysis, it's not given that he never looked at the analysis a second time from the moment he discovered it, and it's also quite likely that what he did during the game was a mixture of remembering and reconstructing. This doesn't mean that the examples aren't impressive - they are. But they are more explicable than it might at first seem.
This thought was occasioned by two stories I came across today. The first came as I browsed the contents of ChessBase Magazine 132, in the section on tactics. IM Oliver Reeh recalls compiling the puzzles for the issue, when GM Leonid Kritz showed up. Reeh showed him an example starting 1.d4 d5 2.e3 Bf5 3.c4 c6 4.Qb3 Qb6 5.cxd5 Qxb3 6.axb3 Bxb1 7.dxc6 Be4, and now Kritz said "I know that motif!" Reeh replied "I should think so. After all, you were the one who played this game!" - which Kritz promptly denied! A quick browse of the database indicated that he had, albeit 11 years earlier.
The second story is buried in the notes to Lupulescu-Marin (here), and doesn't involve any such distant time span. In an earlier issue of ChessBase Magazine, he had offered a particular move as an improvement, but in his own game thought for a long time and failed to realize that he could immediately transpose into his own analysis.
This doesn't only happen to "ordinary" GMs either. The great Mikhail Tal tells one or two stories like this in his chess autobiography as well. And of course, it happens all the time to those of us further down the food chain. So while there are stories of Fischer remembering blitz games from 14 years earlier or Anand remembering ultra-complex ten year old analysis***, the truth is far more complex. There's no escaping it: errare humanum est.
* The standard experiment for this conclusion went roughly like this: they compared GMs, masters, and veritable novices on their ability to reconstruct a normal chess position they had seen for a few seconds. GMs and masters did very well, while novices had no real success. Then they had all three groups try to reconstruct nonsense positions, and this time there was no edge, or at least no appreciable edge, possessed by the stronger players. If GMs had some sort of "photographic" memory, this would not have been the case.
** In other words, they have developed the skill to remember chess positions, moves, ideas, etc. The stock explanation is that this occurs by "chunking"; that is, by compressing many distinct bits of chess information into a single concept. (A chess example: the formation of a white bishop on g2, king on g1, knight on f3 with pawns on h2, g3 and f2 might be seen as a single unit - one chunk - rather than as six distinct units. A universal example would be a word in a language. We process words as units, not as collections of letters.)
*** The Fischer story is that in 1971, at the time of his match with Taimanov, he showed Vasiukov (Taimanov's second) some speed games they had played in 1957 or 1958 when Fischer had visited Russia. The Anand reference is to his great win over Adams' Zaitsev Variation in the San Luis world championship; the analysis had been prepared for Kamsky a decade earlier.
Yet even these examples may not be quite as legendary as they may at first seem. For one thing, Fischer may have reflected on the games he played in Moscow, taking notes about the openings if nothing else. That process of study and overlearning makes memorization far less mysterious. I've also heard that the story has been exaggerated, that he didn't demonstrate all the moves of all the games, but bits of some of them. As for Anand's analysis, it's not given that he never looked at the analysis a second time from the moment he discovered it, and it's also quite likely that what he did during the game was a mixture of remembering and reconstructing. This doesn't mean that the examples aren't impressive - they are. But they are more explicable than it might at first seem.
Posted by Dennis Monokroussos on
Wednesday October 7, 2009 at 11:35am
It turns out that the bit of the brain used normally for human facial recognition can get 'hi-jacked', so to speak, by the specific area of individual expertise. For experienced chessplayers, recognising &remembering positions/motifs is exactly the same mental process as recognising people.
So, if you're capable of remembering a person you spent one evening with 13 years ago- in detail- you too have the mental tools to recreate Fischer's feat. Maybe not the ability, but the tools...
J.A. Topfke
[DM: What stories?]