26 February 2011

Rowson's 'Three Types of Theory'

A while back I started a series on opening principles in chess960 by drawing on known principles from traditional chess. The first two posts in the series were

Lately I've been reading 'Chess for Zebras' by Jonathon Rowson, an excellent book in the genre 'How can we know what we know about chess?' Divided into three parts, the book's third part is titled 'Thinking Colorfully about Black and White'. A little thought should convince you that the color difference applies specifically to the opening.

The first chapter of part three is 'Ch.11 - Three Types of Theory and What They Mean in Practice'. I've isolated a single paragraph by Rowson in each of his discussions of 'Three Types of Theory' that I think explains the gist of that particular discussion. This categorization of theory leads to the question, 'How does chess960 relate to that particular view of chess theory?'.

Hypertheory: [Is chess a win for White, a draw, or a win for Black?] Hypertheory is the omniscient view of chess, the view of chess 'under the aspect of eternity'. [...] No person could ever grasp chess from a hypertheoretical perspective, but in principle it should be possible for a machine to reach this fundamental perspective and develop 32-piece tablebases. This may take decades or even centuries, but [...] I think it will eventually happen.

Since the idea of a 32-piece tablebase might bump into some fundamental limit of material nature, I'll leave that to the mathematicians and physicists to debate. Current technology allows for a 6-piece tablebase and I have no idea when we can expect a 7-piece monster to be available. (*)

If a 32-piece tablebase is available some day, it will solve all starting positions in chess960 as easily as it solves traditional chess. The last chapter for chess960 will then be to determine for each specific start position whether it is a win for one side or a draw. This is an extension of the frequently asked question whether there exist any chess960 start positions that are too drawish or too lopsided for one side or the other. So far, no one has been able to give any specific examples although it is a frequently used argument against chess960.

Elite Theory: [Rowson says 2670+ players represent the elite, but doesn't explain why he picked this unusual number.] The main reason openings matter so much more at the elite level is that during the game the margin for error is so much smaller, and the level of concentration and technique is so much higher. This means if you gain a serious advantage out of the opening there is a very real chance you will win the game. Moreover, if you get nothing out of the opening it is much harder to outplay your opponent because they will normally know how to play rather well!

Since there is no opening theory in chess960, despite the fruitless efforts of countless converts to catalog 'best moves' in each of the 960 positions, there is no notion of 'elite theory'. Some critics go as far as saying that chess960 levels the playing field for all players, whether grandmaster or club amateur, but I'm convinced this is a serious underestimation of how much grandmasters know about the game. It might be a leveller across all grandmasters, e.g. between a super GM and a weak GM (a chess oxymoron if there ever was one), but the results from Mainz don't bear this out. The super grandmasters still manage to win.

Our Theory: [Below 2670?] In addition to the desire to copy the superstars, another reason we spend a disproportionate amount of time learning opening theory is that unlike other kinds of chess work, the fruits of the labor are very tangible. If we learn a new opening idea we can usually imagine putting it into practice much more readily than if we learn any other kind of chess idea. However, this is a limitation of our imaginations rather than a true reflection of the kinds of work that will help us improve.

Another frequent argument against chess960 is that it is primarily for elite players, because that is the group who searches for opening novelties at move 20 and beyond. I submit that this argument is also 'a limitation of our imaginations'. What better way to force improving players to think for themselves than by forcing them to solve new problems starting from the very first move?

In part three of Rowson's book -- highly recommended, if I need to say so in Black and White -- are chapters on 'White's Advantage' (think 'initiative') and 'Black's Advantage' (think Adorjan). No matter what we might think of it as an alternative to traditional chess, chess960 forces us to consider in greater depth the many subtleties of the chess opening.

***

(*) On my main blog I've collected examples where tablebases have shown famous players to be wrong in their evaluation of certain endgames: see Posts with label Endgame TB. The 7-piece version will provide many more examples.

No comments: