In the November issue of Chess Life, GM Alejandro Ramirez, who tied for 1st-7th, gave some interesting analysis of his game against GM Julio Sadorra. Look at the following position:
Of this position Ramirez wrote, in part, “A swift glance will reveal nothing more than a normal KID/Benoni setup with Black having a ton of dark squares and White having a nice space advantage. A deeper study reveals that Black doesn’t have a plethora of dark squares, he has one: d4. This square isn’t even heavily controlled by him; as soon as White is able to play b4 and put the N back on b3, he will have full control over d4. White will then have a space advantage on the Q-side, the K-side and the center. Black, unlike most Benonis, has no targets. The d7-N has no squares. He doesn’t want to trade on d5, and at the same time he doesn’t have a useful move. Houdini might think that this game is about equal. I think Black’s position is awful.”
This is a good example of GM thinking about squares and such, but did you get that part where he said, “Houdini might think that this game is about equal. I think Black’s position is awful.”
This illustrates a point I’ve made elsewhere in this Blog that when an engine evaluation differs from a GM’s, trust the GM! I put this position into my program and let Ivanhoe (we already know Houdini thinks the position is about equal) look at it for about 10-12 minutes just to see what it came up with.
Ivanhoe also thought the position was about equal, giving White a 0.11 advantage. But…and this is important…when you jump to the end of the variation, the advantage changes drastically to nearly 2 Pawn’s in White’s favor.
This has to do with the ‘principle variation’ which is described as the line the engine believes represents best play by both sides if it has been set up to display thinking. CCGM Robin Smith advised that when looking at the PV, don’t pay too much attention to the whole line that is displayed. Never believe the programs PV all the way to the end. The first move displayed will be the move the engine thinks is best, but after that, all bets are off.
His advice was, after letting the engine run a sufficient amount of time, paste the moves into the game then jump to the end of the evaluation and step backwards through the moves. Using this technique will often result in your seeing something quite different than just looking at the initial output.
Such is the case here. After 10-12 minutes, Ivanhoe displayed the main line to move 11.Rd1 and evaluated the position as nearly equal (0.11), but when I stopped the analysis and went directly to the position after 11.Rd1, the evaluation was almost a 2P advantage for White, which is actually more in line with Ramirez’ evaluation. Stepping backwards we see that Black could have possibly improved with 7…Nxb3 or later with 10…Nc3.
My goal here is not to provide a complete analysis of this position, which is something I’m not capable of doing anyway, but to show, or rather hint at, the way GM’s look at positions when evaluating them. Of particular interest in this position is Ramirez’ comments when he is discussing squares. Also, it is another good example of the proper way to use engines to analyze games. In some positions you still can’t, even these days, rely solely on their output and you can’t just plug in an engine’s principle variation and think you’ve got good analysis. It still takes some investigating on the part of the human and you can’t get top notch evaluation by spending 15 minutes on a position like I did in preparing this post. That’s why titled CC players even with engines need days to evaluate positions in order to arrive at the best move.
Still, even we average players can learn a lot if we are willing to put some effort into analyzing with engines and do not, like most of us, just let it analyze our game at a few seconds a move and think we have come up with the best lines.