09 May 2013

The Language of Logic

I would like to take one more point from GM Rowson's discussion of the 'Black Is OK' concept, last seen on this blog in Not Lost at the Outset. I'm going to take it out of context, because the idea stands on its own merit irrespective of Black's initial outlook. Rowson wrote,

I noticed that in many chess positions White seems to be slightly better, doesn't make any mistakes from a human perspective, and then seems to be slightly worse. Those with a logical cast of mind rebel against this idea and suggest that all that this means is that the initial assessment was mistaken. [...] An assessment of 'slightly better' can change from one side to another without the implication that the initial assessment was 'wrong'. The role of the initial assessment was not to be right or wrong in any absolute sense, but to make a good guess. ['Chess for Zebras', p.229]

This reminded me of a beef I've often had when studying games that use the Informant style of annotation: codes rather than words. The assessment sometimes shifts from 'White is better' to 'Black is better' (or vice versa) without any indication of why the shift happened.

Nowadays we rely on a computer evaluation that reduces a complex position to a single number (or two numbers if you also consider the depth of the calculation). An evaluation of a 0.5 advantage for one side gives more information than the phrase 'stands better'. Going further, as I learned recently in A Pawn Equals 200 Rating Points, the 0.5 advantage translates to a 64% chance that the side with the advantage will win the game.

I don't know how much of the Informant code system predated the first publications of the ongoing series. The symbols '!', '?', etc., have been in use for a long time, but what about the rest? Here is an explanation of the system as it appeared in Informant no.1, covering the first half of 1966.

How accurate were those early, pre-engine assessments of 'has the upper hand', 'is slightly better', and so on? How well did those position assessments mesh with move assessments like '!?' and '?!'? I'll take a look at some of these questions in a future post.

No comments: