Engine Evaluation

A couple of paths converged recently to give me a fresh perspective on an old topic: chess engine evaluation of specific positions. The first path was on using the engines to evaluate positions in various games. In several games the engines kept producing evaluations that didn't look right to me. The positions were all of the type having complicated material tradeoffs on the board -- Pawns vs. pieces, Rooks vs. minor pieces with Pawns on one side or the other, Queens vs. pieces -- that sort of thing. An engine would tell me the position was worth +1.0 for White or -2.0 for Black, where instinct told me that the evaluation might possibly be way off.

The second path was on engine evaluation of start positions in chess960. I won't go into detail here, because I know it's an esoteric subject that doesn't interest many people, so I'll just point to a post on my chess960 blog where I documented my findings: Twin Research. The upshot is that different engines play into the same variations, but the variations seem not to be the best. Is this because the engines are flawed or because there is something wrong with the chess960 start positions?

I realized that I didn't know much about the subject of material evaluation, engine or otherwise. I did, however, remember an article from Chess Life by Larry Kaufman. I found it in the March 1999 issue: 'The Evaluation of Material Imbalances' by IM Larry Kaufman. The then-IM is now a GM and the article is available on the Web on Dan Heisman's site: 'reprinted with Larry's permission'. I have more to say about the subject, but right now I have to help the girls decorate the Christmas tree. I'll come back to this another time.