Saturday, July 19, 2014

Engines and Books

     I was recently looking at the below position (using ChessOK Aquarium 2012) from an online game I played a few years ago and it was my (White’s) turn when I noticed some differences in the evaluations of Stockfish 5 and Houdini 2. The brief explanation of what I saw shows the danger of just zipping through a game and relying strictly on computer generated analysis. We all know you won’t learn anything except where you made a tactical mistake by looking at engine analysis and we also know that when it comes to top level correspondence chess with its heavy engine use, just playing the recommended engine move won’t get you anywhere near the top levels.
     Then I found a must read article by GM Kevin Spraggett on his Blog where he said, “BUT worse still, many reputable authors have produced best-selling products that are deliberately dumbed-down with absolutely useless and often redundant computer-generated analysis of variations. Computer analysis trivializes the basic skills that are needed to become a master level player…I REFUSE on principle to fall for the commercialized or chess-engine processed information put out by those who want me to later buy their products…” You must read the whole article!

White to move
 



     As you can see from the above analysis panels using Stockfish 5 and Houdini 2 the best move is 18.O-O, but the evaluations, especially for the second choice (18.Bd3) are quite different. In the game I actually played 18.e4 to which Black replied 18…Bxe4. Retreating the B to c8 was evaluated 0.00 by SF5 while H2 recommended retreating it to d7 and evaluated the position at -0.09. Black played 18…Bxe4 and the question is, “How should it be evaluated?”
     Interestingly, while analyzing the position after 18.e4, SF5 listed 18…Bxe4 as its third choice and evaluated it at +5.22 which is an easy win. H2 on the other hand also listed 18…Bxe4 as its third choice but evaluated it at only a half Pawn in White’s favor! That’s a major difference. When the move 18…Bxe4 was actually made SF5’s evaluation jumped to 6-plus P’s and H2’s immediately jumped to nearly 4 P’s! In the game I won quickly after 18…Bxe4 19.Qxg4+
     After making the better move 18…Bc8 both engines evaluated the position at nearly equal while 18…Bd7 weighed in at 1/3 of a P advantage to White by SF5 and H2 put it at dead equal. After 18…Bd7 both engines recommended 19.Be5, but it took Stockfish several minutes to find it.
     Clearly, analyzing your games with an engine and expecting that’s all it takes to find the best moves and at the same time improve your chess won’t work. And, as GM Spraggett points out, neither will spending a ton of money on crappy chess books, especially opening books, help.

No comments:

Post a Comment