Three critiques, two minor and one major, regarding de Sousa’s (2002) framework for the evolutionary processes of chess openings, are provided in terms not of disproving his framework, but instead expanding on it, especially in relationship to his consideration that positions are memes, but do not contribute to this evolutionary process. One minor critique focuses on his insistence on separating opening moves from positions; opening moves lead to positions, and the human mind, in playing and learning chess, focuses on positions, rather than moves. The second minor critique is of his description of quasi-extinction, making variations disappear from the literature; this does indeed happen, but variations also sometimes spring back into the literature because a hard-working master has re-evaluated a position resulting from opening moves. The final and major critique is that de Sousa does not give adequate thought to the development of chess thought, which is itself a recipeme of major importance in chess evolution.
In an otherwise excellent article expanding on Dawkins’ (1976) concept of memes as units of cultural transmission, postulating a "specific class of memes that populate the minds of players in the international chess community: the ideas related to the chess moves themselves," de Sousa (2002) makes an error in his evaluation of chess positions. One can start with his description of patterns of good and bad health in the king’s fortress, or castled position. He correctly notes there that, "Knowledge of these patterns is very important to a player’s strength; they are further examples of chess memes," but there begins to err in ending the statement with "which do not directly correspond to moves."
This error is further compounded in his description immediately preceeding the position known as "The Evergreen Game" (Horowitz 1961), where he states that, "Whilst combinations are also successful memes, and knowledge of them contributes to a player’s expertise in the middlegame, they will not be emphasized in this article because they, unlike opening moves, cannot be repeated in new games, and so there are no evolutionary processes behind them." He later notes, "Putting it simply, combinations don’t evolve, openings do."
First, some minor critiques; the intial one based on these statements. First of all, all opening sequences result in positions (he intermixes the use of position and combination in the article, which I feel is a mistake). Thus positions (opening or middlegame) are as evolutionary as are openings; in fact, it is from a specific opening position that the evolutionary process of chess openings begins. He cites the Sveshnikov opening specifically and the great number of novelties that have resulted here; each novelty of course corresponds to a specific position, which is also a meme capable of evolution. He himself falls into the pit of his own descriptions at times, for example:
If say, White’s 14th move is a winning move, then Black’s 13thmove, played immediately beforehand, was a losing move, because it led to a losing position, Therefore, in future games, players who have understood what happened will avoid playing the same position with Black.
He thus falls into his own trap, it is indeed true that positions, which are the results of moves, are as evolutionary as the moves themselves. Several possible moves and several possible positions will both be now explored by players for the possibility of avoiding that losing position. He also describes positions as both memes and loci, which strikes me as inconsistent as well. Simply viewing positions and openings as evolutionary memes that can occupy various loci in the chessplayer’s mind seems to me much more consistent. Also, in chess, it has been noted that the opening and middlegame have been growing much closer; in some variations, novelties such as discussed by de Sousa don't occur until the 25th to 30thmove, well into the middlegame. And understanding the middlegame is understanding chess, where players will show much more success if they understand that portion of the game.
Also, the choice of the Evergreen game was a poor one, even though it may be a striking position. The game was played in an era when chess play was best described as "swashbuckling," and razor-sharp, although often unsound, attacks predominated. The Evergreen game has little relevance to the modern player as the game was played in much a different way than games played today (although authors still consider various possibilities in the Evergreen game even today, which makes de Sousa’s comment that positions don’t evolve somewhat suspect. It is not unusual to find, in chess magazines, players writing in with analysis they find to refute the previous analysis of a classic game).
It must be noted that de Sousa may have been proceeding from an assumption that middlegames do not meet the criteria set forth by Dawkins (1976) of any successful replicator: copying-fidelity, fecundity, and longevity. That is, middlegames have more to do with pattern recognition (which may be imprecise and not copied with the same fidelity in the chessplayer’s mind).
Openings certainly are followed closely by chessplayers and thus exhibit fecundity perhaps to a greater extent than middlegames; but this does not mean that middlegames lack fecundity. Positions and combinations certainly exhibit longevity in that they are often repeated in the chess literature, and are passed on through games no less so than openings. However, if middlegame positions or combinations had no means of replicating themselves, we would have to view them as Sewall Wright (1932) did in his theory of adaptive landscapes; they would become extinct. However, as Gavrilets (1999) has more recently noted, this model, along with Wright’s shifting balance theory, lacks the rigor to explain evolution of one adaptive population to another when selection should discourage any changes away from the current adaptive peak. One certainly hears chessplayers, at tournaments, in the post-mortem analysis session, make comments such as "This game reminds me of Botvinnik’s famous win over Capablanca," meaning that the pattern has been replicated, even if not precisely so.
For an example of this type of imprecise replication of middlegame positions, consider first the following position from Kavalek-Marovic (Kavalek, 1973):
Kavalek went on to win after the hammer-blow 24. Nxe4! fxe4 25 Rxf6 Rxf6 26. Bxe4. In his report on the strong IBM International Tournament at Amsterdam in 1973, Kavalek is sure "that this moment was fixed in Marovic's mind and that it reappeared during his game with Petrosian":
With this, Marovic was able to beat the former world champion and eventual
tournament winner. The positions were not identical, but similar, but there
was a replication of a combination from an early game. This would seem to provide some proof that
combinations also evolve, albeit in an imprecise way with the human host
as a carrier of the meme, or as Hull (1988) has described
any whole that carries a meme, an interactor.
But perhaps this dilemma will be resolved by viewing openings and positions (middlegames; combinations) as meme-complexes, "a set of mutually-assisting memes which have co-evolved a symbiotic relationship" (Grant, 1990; Hofstadter, 1985). A synonymous term for meme-complexes is scheme, which relates well to the chessplayer’s preparation in knowing both opening variations and middlegame themes or positions.
My second minor critique comes from his discussion of quasi-extinction, and how variations thought to bring great advantage to one side will eventually disappear from the literature, citing this as an "example of how, in the theory of openings, variations that give great advantage to one side bring evolutionary deadlocks." One cannot fault these statements; they are true. But they stop short of the point, surely known to de Sousa as a strong chess player, that sometimes these are based on faulty evaluations and games, and that in time, another player may resurrect the line, finding it is not so bad after all. There is sometimes superficiality in the chess literature, especially in annotated games, where the annotator stops short, saying, for example, "and white wins." But later analysis often finds that this is not true, and a move becomes fashionable again, or to use the language of memetics, re-enters the evolutionary pool. For example, consider the following from Dutch Grandmaster John van der Wiel’s (2000) article on the ressurection of a line in the Queen’s Gambit Accepted, after the moves 1. d4 d5 2. c4 dxc4 3. e4 c5 4. d5 Nf6 5. Nc3:
Linares 1993 saw a fashion hype of 5. ... b5. A survey by Shirov soon followed (Yearbook 28), but not long after that it transpired that in the crucial line with 6. Bf4 Qa5 both 7. a4 Ne4 8. Ne2 and Shirov's 7. Bd2 were quite promising for White. Then the variation lay dormant again.
Figure 3. Arrow illustrating the move 6. … Ba6to be carried out, reviving the 5. … b5 line.
van der Wiel’s use of the term "dormant" is interesting, in that
memetics uses the term as well to mean "currently without human hosts,"
laying in wait to "re-activate themselves" (Grant,
1990). Thus 5. … b5is
not as bad a move as was thought by even strong grandmasters such as Shirov,
and it simply took someone who saw further into the position to resurrect
or re-activate the move. For some reason, de Sousa ignores this possibility,
although it happens time and time again, with some chess masters having commented
tongue-in-cheek that a novelty is simply a move that has been forgotten by
the current generation of players. This is of course, only a half-truth, and
less likely to occur today, as de Sousa notes, with the great proliferation
of databases such as ChessBase, but it must be noted that any database will
only be as good as its operator.
These critiques are minor and should only be considered as extensions to De Sousa’s seminal work, as should the one major critique I propose to bring to his work, which relates back both to his description of positions as non-evolutionary (he also discusses how moves occupy ecological niches in their environment - the mind of players. I would argue that positions occupy that niche with even greater strength than individual moves; when playing a game, one often forgets the specific move or move order called for, although the general position is well-remembered) and the evolution of chess thinking. By way of introduction, it should be noted that de Groot (1978) brought forth the seminal study of chess thinking, in which in was found that chess masters were better players because they were able to remember patterns better than poor players, and of course, knew what to do with those patterns. In some ways, a computer chess player without an opening book is much like a poor chess player - every position is new to the machine. It will beat the poor player, and often even the master, due to its ability to calculate far deeper than a human can. Most human thinking proceeds in this matter; if one picks up a textbook teaching medical students how to read radiographs, for example (Squire and Novelline, 1995), pattern recognition is stressed. The radiologist, in reading a chest radiograph, for example, will first concentrate on bony structures and then move on to soft tissue to diagnose whether abnormalities such as emphysema or pleural effusion exist.
Chess thinking has evolved over time, building on de Groot’s work, no less than openings have evolved. In fact, one could contend that for the great masses of players noted in deSousa’s paper, that opening knowledge is nowhere near as important as it is to understand specific positions in which one can execute combinations to win the game. In most amateur games, the refinements of the professional’s opening is lost, and one proceeds through many mistakes. Most books that profess to teach the amateur player to become a better player use the "pattern recognition of positions" method (de la Maza, 2002).
Several books on thinking in chess have recently appeared, such as Aagard'sExcelling at Chess (2001). Most of them build on the work of de Groot in humans as first recognizers of patterns who then calculate specific variations. Aagard has published several good books in recent years, his book on the Panov-Botvinnik of the Caro-Kann (1998) taught me how to think in terms of similar patterns in that opening, a useful heuristic in winning games (and of course, is based on de Groot). And of course, as Tarrasch himself noted, after the opening the gods have placed the middlegame - the source of positions or combinations. Most opening books today stress understanding the resultant middlegame positions that arise much more than knowing the moves - de Sousa again stresses moves.
Of course, that method is not useful when the pattern does not match the actual position (similar positions can vary quite a bit, a point I will cover later), which is why John Watson' Secrets of Modern Chess Strategy (1999) was immediately hailed as a classic in chess thinking. It took me a long time to really understand Watson, in which the argument is that calculation must be more important than pattern recognition and set rules in chess. Thus, Tarrasch's old admonition that "A Knight on b6 always stands badly" must take a back seat to a specific position in which, due to calculation, a knight on b6 stands well.
Watson and Aagard can be seen as providing a meme of a certain type also mentioned by de Sousa, the recipeme, which are competing ideas of how to do things (Langrish, 1999). These types of debates have existed in chess since the time it was first recognized that chess was not merely a game, but could be studied scientifically. Probably the best-known debate occurred around the 1930s with the classical school (Tarrasch) competing with the so-called hypermoderns (Reti, Nimzowitch). Both stressed rules, although the rules were different - Tarrasch for example felt that space should be gained by occupying the center with pawns, whereas Reti and Nimzowitch advocated that piece control of the center, through the so-called "fianchettoed bishop" (a bishop on b2 or g2) was important.
What bothered me initially about Watson was that it appeared that he was saying that, for humans to be good chess players, they must be like computers, throw pattern recognition and rules out the window, and calculate each position on its own merit. Recently, the debate on chess thinking and its evolution has been between Aagard and Watson. However, I think Aagard’s reasoning is faulty, as he appears to share my earlier incorrect view of Watson’s work. Aagard’s chapter 1 is called "Think Like a Human - and Excel at Chess" and Chapter 3 is called "No Rules?, " implying, of course, that Watson is advocating first that humans must think like computers to win chess games and second that the rule books must thusly be thrown out. I think Aagard, in his own thinking, shows a common human characteristic - that of dichotomous thinking, or what is sometimes called either-or thinking. Watson is not saying that calculation is all-important, he is not saying there are "No Rules" - he is instead noting that modern chess must rely more on calculation than adhering to rules.When I read and re-read and again re-read Watson, I realized that he was not saying that humans need to think like computers and that rules were not important. Instead, Watson was arguing that pattern recognition and rules were a ground level of chess understanding, but to excel at chess, to steal Aagard's title, one must adopt certain thinking and critical thinking skills [note 1]. Although I think Watson’s theory and views will prove sounder in the end, both Watson and Aagard are providing us with therecipemesthat will cause chess thinking to evolve along with the openings.
This is from a line of the so-called 150 attack in the Modern/Pirc Complex -
1. e4 d6 2. d4 Nf6 3. Nc3 g6 4. Be3 c6 5. h3 Nbd7 6. f4 b5 7. a3 a6 8. Nf3 Nb6 9. Bd3 e6!? 10. 0-0 c5 11. f5 exf 12. exf c4 13. fxg cxd 14. exf7 Kxf7
Figure 4. – Pirc 150 Attack position, Line A.
1. e4 d6 2. d4 Nf6 3. Nc3 g6 4. Be3 c6 5. h3 Nbd7 6. f4 b5 7. a3 a6 8. Nf3 Nb6 9. Bd3 e6!? 10. 0-0 c5 11. f5 c4 12. fxe6 cxd 13. exf7+
Figure 5. Pirc 150 Attack, Line B.
Basically one move pair of difference between the two lines that would
not appear that different to the untrained eye. What is the difference, in
short? In line A, white's piece sacrifice has brought
him to a nearly won position. Black's opening of the d1-h5 diagonal with the pawn trade
will probably end up being fatal. In line B, white has
compensation for the piece, but no more. His attack will be much more difficult.
Either side still has good winning chances Thus Watson’s contention would
be that the patterns are similar; the rules are similar (white strives to
attack a piece down), but that specific calculation is necessary to solve
the problems of each. de Sousa would actually be in agreement with this, based
on his work; it is his insistence that the positions are not evolutionary
but only the moves that is tautological in nature and confusing: moves lead
to positions which lead to more moves.
In summary, I will again say I am not critiquing de Sousa’s work on
the ground that it is not sound. It is in fact quite sound and will prove
to be a seminal discussion of the evolutionary processes of chess openings.
But chess is not just an opening, or a middlegame, or an endgame - it is
also two people thinking, and the thinking process is as important in the
evolution of chess as is the opening. And the thinking process involves the
knowledge of positions, how they are the same, and how they differ from one
another. Instead what I have written should be seen as an expansion of his
work, as another "building block" in the area, an area that is ripe
for further exploration, especially given chess’ status as an intellectual
game and the fact that more books are written on chess each year than on
any other subject.
Aagard, J. (1998). Easy Guide to the Panov-Botvinnik Attack. Guilford, CT: Globe Pequot.
Aagard, J. (2001). Excelling at Chess. Gloucester: Everyman Chess.
Dawkins, R. (1976). The Selfish Gene. Oxford: Oxford University Press.
de Groot, A.D.(1978). Thought and Choice in Chess. The Hague: Mouton.
de la Maza, M. (2002). Rapid Chess Improvement. Gloucester: Everyman Chess.
de Sousa, J.D. (2002). Chess Moves and their Memomics: a Framework for the Evolutionary Processes of Chess Openings, Journal of Memetics - Evolutionary Models of Information Transmission 6, http://jom-emit.cpfm.org/2002/vol6/de_sousa_jd.html.
Gavrilets, S. (1999). A Dynamical Theory of Speciation on Holey Adaptive Landscapes. The American Naturalist154, 1, 1-22.
Grant, G. (1990). Memetic Lexicon. http://pespmc1.vub.ac.be/MEMLEX.html.
Hofstadter, D.R. (1985). Metamagical Themes: Questions for the Essence of Mind and Pattern. New York: Basic Books.
Horowitz, A. (1961) The Golden Treasury of Chess. New York: Barnes and Noble.
Hull, D.L. (1988). Interactors Versus Vehicles. In: Plotkin, H.C. (Ed)The Role of Behavior in Evolution. Cambridge: MIT Press.
Kavalek, L. (1973). Come to Holland! The IBM Tournament at Amsterdam Chess Life and Review 28, 558-566.
Langrish, J.Z. (1999). Different Types of Memes: Recipemes, Selectemes, and Explanemes, Journal of Memetics - Evolutionary Models of Information Transmission 3, http://www.com.mmu.ac.uk/jom-emit/1999/vol3/langrish_jz.html
Squire, L. & Novelline, R. (1995).Fundamentals of Radiology, 4th Ed. Cambridge: Harvard University Press.
van der Wiel, J. (2000). QGA Acceptable Now?, New in Chess Yearbook 57, 173.
Watson, J. (1999). Secrets of Modern Chess Strategy: Advances Since Nimzowitch. London: Gambit Publishing.
Wright, S. (1932). The Roles of Mutation, Inbreeding, Crossbreeding, and Selection in Evolution. Proceedings of the Sixth International Congress on Genetics, 1, 356-366.
Back to Issue 3 Volume 7