The other day I was on Instant Chess and played a G15 game where I was Black and played the Ruy Lopez Schliemann Defense (aka Jaenish Gambit). This is a sharp line where Black is willing to sacrifice a Pawn or two for a K-side attack. The game opened: 1.e4 e5 2.Nf3 Nc6 3.Bb5 f5 4.exf5 e4 5.Qe2 Qe7 6.Bxc6 dxc6 7.Nd4
My engine opening book shows four moves for Black:
7…Qe5 (1 game)
7…c5 (0 games)
7…Nh6 (0 games)
7…Nf6 (0 games)
I played 7…Qe5 and there followed 8.Ne6 Bxe6 9.fxe6 Qxe6 strictly speaking Black does not have to play this immediately because the P isn’t going anywhere. He can also play 9…Nf6 or 9…Bd6.
When analyzing the game afterwards I was curious as to what the engines might consider the best move for Black was at move 7 since for all practical purposes were we out of the book.
Of course engine moves can’t always be relied on in situations like this because the number of non-forcing moves are endless.
For example, according to Stockfish 5 in this position Black has 5 moves that do not incur a disadvantage: …Nh6, …Qe5, …Nf6 and …g6 all of which are evaluated at equal or less than a quarter P advantage for White. …Bd7 comes in at slightly less than half a P, then …c5 is evaluated at ¾ Pawn advantage for White, so Black could play any of the first 5 moves with about an equal position.
Switching over to Houdini 2 the evaluations were nearly the same. Critter spit out about the same evaluations but thought …c5 was also favoring White but only by about ¼ of a P. Critter also thought …a6 was a reasonable move, evaluating it at about a half P in White’s favor. Its 8th choice (7…Kd8) was nearly one Pawn in White’s favor. Then I tried Deep Rybka 4 and found pretty much the same evaluations.
This told me what I already knew: engines are not of much practical help in evaluating the best moves in non-tactical positions. What this means is that to find opening innovations that are really worth something you have to evaluate alternative ideas that aren’t in the books or go really deep into the early middlegame in order to find a line that the engines have not considered or misevaluated. This is how GM’s and top level CC players find theoretical novelties that are worth something. For the rest of us playing a move suggested by an engine can be dangerous if you don't understand why it’s good or bad. You can end up in a middlegame or ending that is difficult to play if not lost.
I never trust engine evaluations in unbalanced material situations because they often over value material. I’ve found the best way out of this dilemma is to use Houdini because of its speed and run some Shootouts just to see what the results are likely to be.
Back to the game…after 7…Qe5 I was expecting 8.Nb6 but the guy played 8.Nd6 which kind of surprised me, but thinking about it after the game I concluded it’s probably the best because it forces Black to part with one of his B’s which is something you normally would try to avoid when you play an open game. After 8...Bxe6 9.fxe6 Qxe6 10.d3 Nf6 11.Nd2 we have this position and are now out of my opening books.
I castled Q-side and he traded on e4: 11…O-O-O 12.Nxe4 Nxe4 13.Qxe4 and now I blundered by trading Q’s without even thinkig. It was a blunder because after 13…Qxe4 14.dxe4 we have the following position:
So much for my plan to sacrifice a P or two for a K-side attack, but that aside, by trading Q’s I have, as you can see, left him a passed P in the ending which he exploited without much difficulty. The engines recognized this because they show a winning advantage for White. I am irked that I didn't recognize it immediately. It was one of those moves you play without thinking then realize you have just blundered...or am I the only one that does that?