- Analysis vs. judgement
- Memory: the touchstone of talent?
- Towards a general model
- Further reading
The reason for doing this, apart from my own general interest, is because:
- chess research has had a big influence on psychology (particularly in skill research: see, for example, Nicolas, Charness, Vicente and Brewer)
- chess research may have important implications for players
- many players have misconceptions about what chess research shows: until recently (at least!), this included myself
"All candidate moves should be identified at once and listed in one's head. This job cannot be done piecemeal, by first examining one move and then look at another." -- Kotov,A Think like a Grandmaster, tr. Cafferty.  (Batsford)Kotov also insists that each branch of what may be a complex tree should be examined once and once only: anything else shows lack of confidence and leads to a waste of precious minutes on the clock.
I know club players who feel guilty because they rarely or never use this technique. There are a number of important points that can be made in reassurance:
- Kotov was already a strong player before he adopted this approach, and while this may be what one needs to get from FM to IM, it may not be the most important thing for you;
- not all strong players adopt Kotov's candidate move approach
(would they be still stronger using it? - see position 6/78 - W to move candidate moves?);
and, most importantly,
- not all positions are suitable for this approach: it is a way of working through positions of a pronounced tactical nature.
"Those chess lovers who ask me how many moves I usually calculate in advance, when making a combination, are always astonished when I reply, quite truthfully, 'as a rule not a single one' "Moreover, if you read many primers on play: Chernev's Logical Chess, for example, or his The Most Instructive Games of Chess Ever Played or even Nimzovitch's My System, they all seem to emphasise positional play and judgement as being the most important aspect of chess.
There is something quite appealing about concluding that strategical judgement is more important than tactical analysis. It may confirm the supremacy of the world of ideas over the brute and grubby realm of tactics; in a sense, strategy is a more inclusive, higher-level way of talking about positions than tactics, just as the gas laws are a higher-level way of talking about the behaviour of gas molecules.
But as well as being appealing, it may be correct. Many people have argued that chess research shows that, whatever makes the difference between masters and the rest of us, it's not the ability to calculate. For example, the British GM David Norwood has published a booklet summarising his views on chess, which includes this statement:
" It is often supposed that, apart from their 'extraordinary powers of memory', expert players have phenomenal powers of calculation. The beginner believes that experts can calculate dozens of moves ahead and he will lose to them only because he cannot calculate ahead so far. Yet this is utter nonsense. From my own experience I can say that grandmasters do not do an inordinate amount of calculating. Tests (notably de Groot's experiments) supports me in this claim. If anything, grandmasters often consider fewer alternatives; they tend not to look at as many possible moves as weaker players do. And so, perversely, chess skill often seems to reflect the ability to avoid calculations. It is, in truth, not clear that chess is a game of calculation. Of course there are times when intense calculation is called for, and often the master is better at dealing with these situations than the amateur. No wonder, he has had more practise than the amateur, but all the same his innate calculating ability need not be any greater. Most of the time it is something quite different that is required in chess, something more akin to 'understanding' or 'insight'." -- David NORWOOD, Chess and Education [1995, Gresham College, London].Paul Powell's Chess Page includes a polemic to much the same effect, and there are a number of off-line sources (what are they called again? Oh, yes, "books") which concur. I'd like to look at de Groot's research, but first, review the history of chess research that de Groot had to go on before his own studies. simultaneous display, but it is one of the most impressive demonstrations of chess skill. Masters take on dozens of opponents with little apparent mental effort, and achieve huge winning percentages. There seems no time to calculate, as the moves are made almost at a glance. More amazing is blindfold chess, and yet masters can even give blindfold simultaneous displays.
What enables masters to do this?
The most obvious explanation is that they are cleverer than the rest of us. But while there are some links between Chess and IQ, masters are not do not score much more highly than other players - certainly not enough to explain the dramatic differences in performance you can see at a simul. And although chess masters' memory for chess, as Binet showed, is striking, it seems no better for other spheres.
In fact, for a long while there seemed not much more to add than that masters are better than other players at chess - full stop. What was needed was some task that was not the game of chess itself, but which showed important differences between masters and the rest of us. One was found in the recall of chess positions.
The paradigm was established in an experiment by Djakow, Rudik and Petrovsky . They presented a chess problem to masters and non-players, and found that after a minute's exposure the masters could recall the position very much better than the non-players, although this is perhaps a difference that one would expect.
In his book Thought and Choice in Chess, Dutch master and psychologist Adrian de Groot describes how he extended and refined this experiment. He also showed a series of chess positions to a panel composed of four chess players for a few seconds only. Grandmaster Euwe and Master de Groot were very good at this - almost perfect (93%) - whereas other players (an expert and a club player) were much worse (68% and 51%), getting no position entirely right. de Groot talks about a "gulf that separates the master from the non-master".
This is a difference that perhaps is more of a surprise.
de Groot had another surprise for us too: he showed the following
position "A" (from one of his own games) to five Grandmasters and
five experts, and recorded and analysed their thoughts:
He then drew up a table of the number of depth of lines analysed. [ If you would like to compare your thoughts, there is some light commentary here: Comment on (de Groot).] He concluded:
"It is unequivocal that depth of calculation cannot be the prime distinguishing characteristic between the grandmaster and the expert player."This hits one with rather a thump. But of course:
"Substantial differences in thinking do exist. The gap between levels of performance of the G- and E-group is enormous: four of the five subjects would almost certainly have won the game; even the fifth would have had a better chance than the E-players who without exception let their opportunity go down the drain."de Groot's research thus led to two things:
- a widespread belief that GMs do not calculate much better than other strong players
- a very fruitful series of studies on the recall of chess positions.
- recall tends to be in bite-sized chunks, e.g. fianchetto/castled king [Kg1,Rf1,Bg2,Pf2,Pg3,Ph2], or an inter-related set of pieces in attack and defence
- errors tend to be of the same "chunked" nature: for example, including typical elements that are normally missing [say, a Bg2], or excluding elements that are not normally present [say, a Pf4]
- recall by strong players is no better than other players for random positions (Chase and Simon, 1973)
There's a nice real-life example of chunked memory, and playing by analogy, in John Nunn's game against Korchnoi in the Blindfold round of the 1994 Melody Amber tournament.
So we seem to have a paradigm for research which is not itself chess, but seems able to tell us something about the way chessplayers perform during games - even suggesting how they think. There was developed the 'recognition-association' model, which suggested that strong players play essentially by analogy: recalling positions and elements of positions, and making judgements about the current position based on this recall. This surely makes a great deal of sense. How often have we seen players flash out the 'stock' Bxh7 sacrifice? Is this skill or merely recognition? Gerald Abrahams, in The Chess Mind, argues strongly for the primacy of vision, and says:
"Seeing the idea precedes the logical argument."Similarly, Hartston and Wason give an imaginary post mortem conversation which I seem to have heard before:
[Stronger player rejects a move suggested by the weaker player]So, this "vision thing" seems more important than memory or calculation. I have two immediate caveats to the direction that chess research took, following de Groot:
Weaker player: What's wrong with it?
Stronger player: It's not good.
Weaker player: Why not?
Stronger player: It's not the sort of move you play in this sort of position
[end of conversation]
- It should be noted immediately that even if GMs do not calculate much better than other strong players, if you want to be a strong player, then I think you do need to be able to calculate. de Groot presents no data for club-class players compared to "experts", but I suspect they would be pretty telling. It may not be the thing that is most important for you to improve, and it may be the thing that is most difficult for you to improve, but I have a feeling that tactical alertness and reliability in analysis are important to develop for we club players.
- Even if we accept that Grandmasters' judgements and technique are much better, and that many positions do not require analysis, it still may be that GMs do calculate rather better than other strong players when the need arises.
"... one other thing is the GMs superiority in tactics. For example Christiansen can find tactics in any position. If you're a GM you should be able to overpower the IM tactically. The GM will often blow out the IM in this area. " -- Nick de FIRMIAN, in How to get Better at Chess Chess Masters On Their Art by GM Larry Evans, IM Jeremy B Silman and Betty RobertsAnd in fact, if you look at the the de Groot material again, the GM/Expert measures were completed once for each player, for only one position. It may be that de Groot chose a position with a number of features that made it particularly hard for the experts to grasp. The key move 1. Bxd5 is a little anti-positional and the Black position does not look vulnerable at first blush. It is interesting to speculate what results he might have got with this one: Colle - O'Hanlon, Nice Ol, 1930: 1. d4 d5 2. Nf3 Nf6 3. e3 c5 4. c3 e6 5. Bd3 Bd6 6. Nbd2 Nbd7 7. O-O O-O 8. Re1 Re8 9. e4 dxe4 10. Nxe4 Nxe4 11. Bxe4 cxd4
Now, Dennis Holding in his 1985 book The Psychology of Chess Skill, collates a number of strands of evidence which suggest that in fact, as we might have all suspected at the outset, strong players do analyse deeper and faster than weaker ones, even in random positions. [But they also are more efficient, homing in on good lines and abandoning duff ones. So it's not just quantity but also the quality of their analysis which is better.] Norwood's claims above look more shaky.
The points made above are not meant to be unkind to de Groot: his work has been so widely cited because it was exciting and important. [I do feel a bit cheated by Norwood, who after all could have found Holding's work as easily as I could.] But it is in the nature of science to improve on past work. There are lots of other good things in de Groot's book: for example, when do players abandon a line of thought and look for something better? (Answer seems to be: when the assessment of a position at the end of the variation being considered is out of keeping with the initial evaluation of the position.)chess is 99% tactics", and CJS Purdy added "Chess is the art of analysis"; on the other hand David Norwood and Reti suggest that analysis is relatively unimportant.
Surely both positions are partly true. (This "one the one hand..., on the other" style makes one yearn for genetic engineers to develop the one-armed academic.) I might offer that memory, judgement and analysis are all important, that we have different abilities and experiences for each aspect, and that different positions make different demands on each skill. Without some pains being taken to discover the balance of skills amongst players of different strengths (style notwithstanding), and the balance of demands of different positions, we are not yet in a position to say what (if anything) is the key to chess mastery.
In support of this general position, I quote from the abstract of a paper by Dennis Holding:
"The paper reviews the evidence for and against the recognition-association theory and a forward-search (SEEK) theory of chess skill. The recognition-association theory appears to be founded on indirect evidence concerning visual short-term memory, together with supplementary assumptions that may be questioned, and provides no role for verbal processes. There is no direct support for the theory, which omits forward search for reasons that are reexamined. In contrast, the SEEK theory maintains that move choice is based on search and evaluation processes supplemented (or else supplanted) by a knowledge base. These processes are directly evidenced by experimental findings. The objection that search theories cannot account for speed chess is met by a review of the available evidence. It is concluded that chess skill relies on thinking ahead rather than on pattern recognition." -- Holding, PSYCHOLOGICAL RESEARCH, 1992Now, we needn't agree with everything Holding says, but it is clear that we may have been hasty in dismissing the importance of analysis (thinking ahead).
The whole of this issue of Psychological Research is devoted to chess, and Holding has written a whole book on The Psychology of Chess Skill (1985). There is clearly much research of interest going on, much of which suggests that chess is more complicated than the 'recognition-association' model. And however resounding the pronouncements made about the importance of chess skills, chess remains more complex, and more interesting, than any pat statement.
One last point. It may be that de Groot had something right all along, namely, his method. I think part of the attraction of the later recall paradigm is that there is no messiness: subjects are given a definite task and the languageless patterns of results are scored according to a definite scheme. But de Groot saw no virtue in this style of work: in fact his book is in part a long plea in favour of "introspective" methods in psychology, where subjects are encouraged to talk about their thinking, and the whole messy, unreliable collection of their words is analysed. I have a suspicion that rather than seek the key(s) to chess through some neat experimental technique, the closer we come to studying chess itself, the more likely we are to find an answer.
In addition to books and papers cited above, I have found interesting and useful Krogius' Psychology of chess and Pfleger and Treppner's Chess: the mechanics of mind. Hartston and Wason have published a readable book on the Psychology of Chess: it has had favourable reviews and covers the psychoanalytic approach rarely referred to elsewhere.
If the psychology side interests you, you may also be interested to read San Francisco Exploratory's Exhibit on Memory