Some players complain that the finding of opening innovations by human players is a dying art because engines are coming up with theoretical novelties. Others counterclaim engines are useful to find traps and tactics but are not considered as good as strong GM’s because engines can't see strategic ideas until the position is well-established. For example, in many cases if you pick a line in a King’s Indian or a Benoni in which the GM’s claim is “even” you will find engines will evaluate it in White's favor. Even so, many opening novelties these days are engine generated with the GM giving the final evaluation.
In this position from one of my CC games the Fritz 12 opening book has only one move for Black: 20…Rae8 with Black winning by means of a R-lift (...Re6 and ...Rg6, using the R in an attack against the White K) . This was Aronin’s move against Anand at Linares, 2009 and since Aronin won the game, it would appear that the move is a good choice.
However, when analyzing the game with an engine it suggests that White’s position is slightly better. In fact, the engines show White keeping his slight advantage up until move 33 when Anand apparently missed the best move and the advantage abruptly swung over to his opponent. Therefore, this might be a good time to look for an improvement over Aronin’s 20…Rae8.
20…Rad8 Houdini 2 suggests this move. Stockfish 4 initially suggested 20...Rfe8 but H2 thought one of the R's, preferably the a8R should go to the d-file. Indeed, upon deeper investigation there was an improvement for White in SF’s main line and so H2’s move of 20…Rad8 seems the best choice. Also, further investigation of 20...Rad8 with SF seemed to confirm Houdini’s evaluation. It’s likely that blindly following the opening book in this case would lead to a loss.
Some time back I was White and was following an opening line played by two 2500 CC players in which White was given a slight plus and eventually won. I figured this was safe because I assumed both sides had checked the line with engines. That was the real problem: how many times have we been told never to assume anything? So, late in the opening I noticed what appeared to be an improvement for Black that left me in a difficult position. Of course my opponent found it too and I was left struggling for many moves; forget the engine; if a move looked like it might be good, I checked it out…all to no avail; I lost to an improvement suggested by a newer, stronger engine.
In the above position I had Black and it appears that White is winning. However…and it’s a BIG however, my engines at the time (which I think included Houdini 1.5) recommended something like 45…Rg5. Instead, I played 45…Rg6! Why? Because after 45…Rg6 even Houdini 2 wants to play 46.Rxg6?? and shows White has an advantage of between 1.5 – 2.0 Pawns. Stockfish 4 puts White’s advantage at over 3P’s; in either case, an easy win for White. But…and it’s a BIG but, those evaluations are wrong!
Eliminate the minor pieces and White wins. Eliminate the R's and it's a draw. To be fair though, if I had endgame tablebases installed on my computer, the evaluation might be different. I don’t know because I’ve never used them.
So, the point of my playing 45...Rg6 was that if White relied solely on his engine, he'd jump at the chance to trade R's and thereby be left with nothing more than a draw. Fortunately, that's exactly what he did and it took him another 32 moves when we reached the following position…
Sorting through all these kinds of things is what makes engine-assisted chess difficult and so labor intensive. I guess that’s why one top level CC player wrote that he couldn’t imagine playing a move with only three days’ reflection time. When the championship sections have 17 players and you need a week or more to do your research, it takes a LOT of time!