Logo de Sousa, J. D. (2003). A Reply to Dowd's Commentary on My Paper.
Journal of Memetics - Evolutionary Models of Information Transmission, 7.
http://cfpm.org/jom-emit/2003/vol7/de_sousa_jd_reply.html

A Reply to Dowd's Commentary on My Paper

João Dinis de Sousa
Rua Actor Vale 49 4º Esq
1900-024 Lisboa  Portugal
j.d.sousa@oninet.pt, j.d.sousa@netc.pt

I will start my response by focusing on Dowd's first minor critique (section 1.1 of his commentary). He remarks that I put too much emphasis on moves rather than on positions in the evolutionary process of chess openings. He says: "First of all, all opening sequences result in positions (...). Thus positions (opening or middlegame) are as evolutionary as are openings".  Of course, I agree that positions are intertwined with moves in every chess game. I also agree that, during the evolutionary process of a given opening, both its underlying moves and its underlying positions evolve too. Finally, both positions and moves are memetic replicators (they spread from one player's mind to another, are written in books, etc). I agree with all that. If I considered moves, rather than positions, to be the main replicators in my framework, it was because moves are informationally simpler than positions. Like genes, they are discrete, particulate entities, and easy to represent. Moves are instructive replicators that entirely codify the phenotypic properties of chess games. The recipe-cake analogy (Dawkins 1986) applies to the relationship between moves and games (see section 1.2 of my article; in the literature, a game is described by the sequence of its moves, exactly like a biological organism is described by its genes). In each position, several moves can be played, so positions function as loci. This makes possible the framework that I depicted in section 5 of my article: moves as replicators, positions as loci, games as interactors, openings as memomes, branches of the theory of openings as species. My framework doesn't parallel biology exactly (I grant this in section 5.1) and it is not the only framework possible, but it enables one to visualise what happens in the evolution of chess openings in the light of replicator-based selectionism.

Dowd could reply that positions, like moves, are also replicators and can also be used to describe games entirely (for example, one could depict a game in written form, by drawing diagrams of all positions composing it). However, that would be an inefficient way to describe a game, since the representation of each position requires much more information than the representation of each move. That's why everywhere games are represented by their underlying moves, not by their positions. As they are simpler, moves approach the ideal of particulate, discrete entities better than positions.

Summarizing my response to Dowd's first minor critique, I agree with him that both positions and moves evolve in the process of evolution of openings, but I consider a memetic framework based on moves as their instructive replicators more illuminating and elegant than a framework based on positions as instructive replicators.

Next, I will comment on some minor points raised by Dowd in section 1.1 of his commentary.

In the third paragraph of section 1.1, Dowd says (taking about me) "he intermixes the use of position and combination" in the article. But positions and combinations are quite different things: a position is the representation of the situation of the chessboard at a given moment, and a combination is a sequence of moves. No possible confusion between the two. So, when the title of figure 2 in my article is "Figure 2. A brilliant combination", and the figure itself shows a chess position, obviously what I meant was not that the shown position was a combination but instead that the shown position was the starting point for the combination that followed. This should be obvious to Dowd and to any reader acquainted with chess, so it should not be interpreted, as Dowd did, as an instance of 'intermixing the use of position and combination'.

Later in section 1.1, Dowd considers my description of positions as both memes and loci to be inconsistent.  Certainly, when one speaks of genes and genetic loci in biology, the two things are quite distinct. However, memetics doesn't parallel genetics in every detail. In particular, memetics deals with ideas, and ideas may exist on a wide range of things, including loci where other ideas compete (for example, a philosophical question is itself a meme and an operational locus to other memes - the possible answers to the question). Returning to chess positions, insofar as they can be remembered, transmitted, and written by humans (for example chess diagrams in books represent positions), they are clearly memes. And, insofar as, in each position, several moves can be played, and hence compete memetically in player's minds, positions may be conceptualized as loci. So, I don't agree that there is an inconsistency in considering positions as both memes and loci. They display both aspects.

In paragraph 6 of section 1.1, Dowd writes: "Also, the choice of the Evergreen game was a poor one, even thought 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." Yes, XIX century attacks were often unsound, and this may be the case of the Evergreen game too, but how is this relevant to the point I was making? I was just giving an example of a highly prolific chess meme not associated with an opening move and, for this purpose, the combination played in the Evergreen game certainly qualifies.

Now turning to another minor critique from Dowd, still in section 1.1. In paragraph 8, he goes on to criticize my emphasis on the fecundity of opening moves, and my downplaying of the fecundity of middlegame moves: "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" (in this context, I assume that he is using the expression "middlegames" to mean "moves and positions in the middlegame"). I would just drop the "perhaps": it is clear that opening moves exhibit much higher fecundity than middlegame moves (some brilliant middlegame moves, like the ones of the Evergreen combination I presented, achieve high proliferation in chess literature and in players' minds, but they are never repeated in new games; and proliferation in new games is the most interesting type of proliferation from the evolutionary point of view).

The earlier a move occurs in the opening the more often it is played in real games. Some of the earlier middlegame moves and positions probably manage to find themselves replicated in other games, but the more one advances into the middlegame the more unlikely this replication becomes. For example, of the millions of moves played at the 40th move in a given year worldwide, almost all are unique. Dowd discusses two similar combinations (figures 1 and 2, referring to the Kavalek-Marovic and Marovic-Petrossian games respectively) to show that middlegame moves and positions also have fecundity in the sense of proliferating to new games. Now, what these two diagrams show is somewhat similar combinations (the move Nxe4) being played in two different games. The two moves are not equal (despite a white knight being sacrificed in both against two black pawns in e4, and some additional similarities, the starting positions are different). So, if there exists any replication here (and I'm ready to admit that it does at the memetic level) it is not like the perfect replication enjoyed by popular opening moves and positions. Each popular opening move (for example, the move 6.Re1 after the opening sequence1.e4 e5 2.Nf3 Nc6 3.Bb5 a6 4.Ba4 Nf6 5.0-0 Be7 (any example would do) replicates with perfect fidelity to other games. The same cannot be said about the move Nxe4 played in the two combinations referred by Dowd because they belong to different positions. Still, I admit that replication of a pattern occurred, in an imprecise way, via Marovic's mind, as Dowd says (in the second game, Marovic played Nxe4 inspired by the similar move played by Kavalek against him some years before, figures 1 and 2). The problem is when Dowd says: "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...". I strongly disagree. Where is the evolutionary process here? How often does this replication happen? Is the "quality" exhibited by combinations the outcome of a recursive evolutionary process? Certainly not. Even though there is some replication of patterns, the patterns lack copying fidelity and, moreover, their proliferation to new games is too low, to allow for complex evolution. Patterns of combinations in the middlegame, even though they can replicate to other games, do not fit into what Dawkins (1982) called the active germ-line replicator, that is, a replicator that is potentially the ancestor of an indefinitely long line of descendant replicators and whose nature can influence its own replication. Unlike middlegame combination moves, opening moves are clearly active germ-line replicators: they last long in evolutionary time. Moreover, they sometimes have very high fecundity (and they proliferate, not only in the literature, as happens usually with combinations, but in thousands of new real games). That's why, as I argue in my article (sections 7 and 8), there are complex evolutionary processes going on (including arms races, and the onset of adaptive quality) in chess openings, and that's the reason for my focus on opening moves.

Now turning to Dowd's second minor critique (section 1.2 of his commentary), what he says is that quasi-extinction of opening variations is often not definitive, because new analyses and novelties resurrect forgotten variations, making the underlying memes proliferate again. This is, of course, true. However, it doesn't apply to all cases of quasi-extinction, but only to those in which the evaluations (in Encyclopaedias of Chess Openings (ECOs) and other chess literature) that classified the variation as being too bad to one side were faulty. In many instances, these evaluations are correct, the disadvantage of one side is considerable, and the evolutionary decay of the variation is definitive (chess analysis has "scientific" features: objective, demonstrable, conclusions often arise). In these cases, where an objective analysis can prove the illness of a position to one side, quasi-extinction usually follows. Even in such cases, it is better to use the word "quasi-extinction" rather than "extinction", because the variation doesn't become totally extinct: many players who are unaware of bad evaluations in the ECOs continue playing it.

Finally, Dowd's major critique (his section 2) is that in my article I ignored recipemes (Langrish 1999) about how to think in chess and how to play the middlegame. Dowd discusses in his section 2 the works of de Groot, Aagard, and Watson about the relative importance of rules, pattern recognition, and calculus. If I didn't include a discussion of these themes, it was not because I don't respect them. On the contrary, I find them of the utmost importance to the study of chess, especially for players wanting to perfect their middlegame techniques. But these matters, interesting as they are, are beyond the scope of my article, the main goal of which was to apply a replicator selectionist approach to the explanation of the evolutionary processes occurring in the theory of chess openings. I would like also to add that, other than Aagard, and Watson, hundreds if not thousands of other prominent chessplayers and writers have contributed to the theory of chess with books full of recipemes about how to play the middlegame. For example, Karpov published many of his games in books annotated with analyses and suggestions of middlegame plans (for example, Karpov 1979,1988). The same applies to hundreds of other grandmasters. Capablanca, in his classic Chess Fundamentals (1952) explores many middlegame techniques, including, for example, the description of many kinds of weaknesses in the pawn structure, along with techniques to exploit them. Kotov's books Think Like a Grandmaster (1976) and Train Like a Grandmaster (1981) are also full of specific middlegame-related recipemes, including recommendations about the relative reflection time allocation between strategic planning and calculation.  The world of chess ideas is very rich indeed, is abundantly expressed in books, and is worth writing about, but, as I said, the scope of my article was not to discuss a wide range of these ideas, but rather to investigate the evolution of chess openings in the light of the selective processes operating on their elementary instructive replicators (moves).

References

Capablanca, J. R. (1952) Ajedrez Fundamental, Madrid: Marsiega.

Dawkins, R. (1982) The Extended Phenotype, Oxford: Oxford University Press.

Dawkins, R. (1986) The Blind Watchmaker, New York: W. W. Norton.

Karpov, A. (1979) Partidas Selectas, Madrid: Editorial Aguilera.

Karpov, A. (1988) The Semi-Open Game in Action, London: B. T. Batsford.

Kotov, A. (1976) Think Like a Grandmaster, London: B. T. Batsford.

Kotov, A. (1981) Train Like a Grandmaster, London: B. T. Batsford.

Langrish, J. Z. (1999) Different Types of Memes: Recipemes, Selectemes and Explanemes, Journal of Memetics - Evolutionary Models of Information Transmission 3, <http://cfpm.org/jom-emit/1999/vol3/langrish_jz.html>

© JoM-EMIT 2003

Back to Issue 1 Volume 7