Speaking of Engine Evaluations…

     This post is sort of related to the previous one and hopefully it will be helpful to anyone who uses an engine in an effort to improve their OTB play. Some people (Jeremy Silman, for example) suggest that engines are only good for tactics and that engines can't suggest plans, which is true. In a non-tactical situation strategy matters; you need a plan, even if it's nothing more than two or three moves that reposition a poorly placed piece. Also, if the position calls for you to attack on the K-side and you are fiddling around trying to attack on the other side, you are probably not going to be very successful. 
     This is a point I tried to make with some players who thought that by playing the K-Indian Attack they could just play the same moves with the same ideas against anything black played and thereby avoid learning any theory. My point was that it is necessary for white to vary his strategy depending on what setup black employs. Black can play various types of setups against the KIA: K-Indian, French, Sicilian, etc, etc and white will have to adjust his plan accordingly. So even when playing the KIA, some opening knowledge is required. Naturally, they didn't believe me...not even when I pointed out I wasn't the one saying it...it was a couple of different IM's and GM's in their KIA books. 
     The thing is, sometimes engines can't seem to suggest anything definite. They offer up a half a dozen moves with the evaluations all within a hair's breadth of each other, so what do you do then? Is the top choice always the best? This is where some knowledge of strategy helps. Increase an engine's number of lines to four or five, sometimes more, and you'll often see lines with very small (insignificant) differences in their evaluations. This is an indication that more than one line is playable in the position. 
     You will often see this if you play over GM games and notice that they do not have a high match up rate with engines, but are playing moves that are further down the list. Their moves are not bad, they are often just alternative ways to play and sometimes they may be even better that the engine's top moves. 
     The reason why even GM's can't beat engines OTB is because of tactical mistakes and humans DO miss a lot of tactics. Just play over any old game annotated in the pre-computer era and you will quite often see authors, even great ones, missing a tactic here and there. Take a gander at this position from one of my recent games where black has just played 15...Qc8-a6. 

     Komodo 8 was showing an evaluation under a quarter P and suggesting various R-moves which tells me it can't really come up with anything that actually favors either side. As a human, I want to attack on the K-side and to that end 16.f5 looks like a really swell move, but it has a tactical flaw based on black's last move. Put the Q on another square, say c7, and 16.f5 is then OK; black has adequate defenses, but white is at least doing something constructive. Of course, there's no doubt that OTB his chances would be much better than playing against a computer. 
     Another move that suggested itself was 16.g4, a move that wasn't in the top 5 choices, but when I made it and let the engine alone for 20 minutes, it didn't find anything wrong and the difference in its evaluations were negligible… the top move was evaluated at 0.06 and the next three at 0.00! A Shootout with Stockfish 6 resulted in white scoring +2 -0 =3. Whereas, after the rather pointless looking 16.Rfb1 white only scored +0 -2 =3 in a Shootout! So, I played 16.g4 and eventually won! 
    The point is, when an engine is suggesting several moves and they are all evaluated nearly the same, choose the ones that look logical to you and then analyze them. It may very well be that a move that is NOT in the top two or three lines the engine is suggesting might actually offer better winning chances.