30 April 2009

Unclear Positions

Chess.com features a regular Q&A column by IM Andrew Martin, appropriately called 'Your Questions Answered By Andrew Martin'. Martin's column for 26 April 2009 had a good question on the meaning of 'unclear position'.

Jess Patrick: Dear Andrew, With regard to the evaluation of chess positions, what, in your opinion, is meant by "unclear position"? Does it mean "One side is better, but I can't prove which one"? Maybe it means "equal but not drawish". What are your thoughts?

Andrew Martin: Dear Jess, Unclear is a word or evaluation often used by an annotator when they don't know what's going on in a position! It's used less these days now that computers have turned everyone into armchair grandmasters. I think your own evaluations above of 'unclear' are among the best I have seen.

I liked Martin's definition ('don't know what's going on in a position') even better than his interlocutor's propositions and it reminded me of a strategy I used when I first started playing correspondence chess. In those days, ECO had just been introduced, and its pages were full of positions terminating with the infinity symbol (''), meaning 'the position is unclear'. I used to steer my correspondence games toward those positions, hoping for the chance to tackle them in the game. The positions were always extremely interesting and took considerable time to sort through the various positional and tactical nuances.

One position I remember well is shown in the following diagram. It arises after the moves 1.e4 e5 2.Nf3 Nc6 3.Bb5 a6 4.Ba4 d6 5.c3 f5 6.exf5 Bxf5 7.O-O Bd3 8.Re1 Be7 9.Qb3 Rb8 10.Qd5 e4 11.Bb3 Nh6 12.Ng5 Ne5 13.Ne6 Qd7 14.Nxg7+ Kd8 15.Ne6+ Kc8 (ECO C74), a line known as the Siesta Variation. Why it's called that is a mystery, because the line is anything but quiet. ECO's analysis, by Radcenko, stopped at '15...Kc8∞'.

1975 Golden Knights Nq-55
Owen, L.B.

Weeks, M.
(After 15...Kd8-c8)
[FEN "1rk4r/1ppqb2p/p2pN2n/3Qn3/4p3/1BPb4/PP1P1PPP/RNB1R1K1 w - - 0 16"]

At the time of the game I was already rated master or near-master and my opponent was in the same category. As we approached the diagrammed position, I spent hours analyzing it and finally had a 'Bingo!' moment when I realized that 16.Nc5! was very strong, and possibly a refutation of Black's opening strategy. The game finished 16.Nc5 c6 (16...Qg4, seeking counterplay, is better and was the response I expected) 17.Nxd7 cxd5 18.Nxe5 dxe5 19.Bxd5 Rf8 20.b4 Kc7 21.Bxe4 Rbd8 22.Na3 1-0.

I don't think I've looked at the game since it finished, but curious about what the computer would say, I plugged the diagrammed position into a chess engine. It found 16.Nc5 almost immediately, at around 7-ply in its calculations. I imagine that the entire variation has long been abandoned.

I have a tall stack of old, pre-Chessbase Informants standing in the corner of my study. It might be interesting to go back and look for positions tagged by the annotator as '∞'. Depending on what I discover, it might even be the start of a new series of posts on this blog.


Tom Chivers said...

A related question is, are there any positions computers find unclear? In the sense that either their evaluations don't settle during in analysis, but bobble either side of 0; or, whereby some computers prefer white and some black and this disagreement cannot be easily resolved.

(I seem to recall that an older version of Fritz in some lines of the poisoned pawn (the e5 ones?) used to yo-yo its evaluations indefinitely.)

Mark Weeks said...

I configure my engines to display the first 10 moves of their calculations. I frequently see a very small (0.10-0.15, i.e. 1/10th of a Pawn) difference between the 1st choice and the 10th. These positions are invariably nontactical, where a few moves might improve the position (e.g. Rfe1), while the others are just shifting the pieces randomly to similar squares (Rfd1, Rff1, etc.). It appears that the engine doesn't 'know' enough about chess to rank the moves sensibly.