The most helpful book I have is CC GM Robin Smith’s “Modern Chess Analysis” which, though badly outdated, still gives good useful advice on conducting engine analysis. While browsing the book the other day I came across the following interesting game and Smith’s analysis.
In the book, Smith played back in the days of Fritz 5, so much of his commentary is no longer valid because of the great increase in strength of today’s engines. Still, if you have ever played over GM games with an engine you will occasionally find annotations where their evaluation does not agree with what the engine is telling you. When that happens, my advice is: believe the GM.
In the book Smith describes how in those days using multiple engines, he let them run for hours, or sometimes overnight, and played engine vs. engine tournaments to help him in search for hidden ideas, flaws and such like in his games. He also pointed out many cases in which engine evaluations were just plain wrong. That’s still the case today even with much stronger engines. At the upper levels of CC play, those who rely solely on engine generated moves will lose nearly all their games against opponents who are of at least IM strength and who are also using an engine.
In this game, it was interesting to compare Smith’s notes to the output of today’s strong engines. For this game I used Houdini, Critter and Spike, but was unable to find any flaws in Smith’s analysis. One thing I did notice was that his claim that Black was winning could not be substantiated because, just like in Smith’s case, the engines showed an advantage for Black but White was still able to hold the draw.