The Moves You Don’t Even Look At

Sometime in my early twenties, I read Alexander Kotov’s famous book Think Like a Grandmaster, which revolutionized how chess players think about thinking. From Kotov I learned the idea of making a list of candidate moves, and analyzing each move once and only once (instead of bouncing back and forth between variations, which was my habit at the time).

Nevertheless, Kotov’s system had one big flaw, which many people have pointed out: Where do the candidate moves come from? There are actually two parts to this question: Where do you get the ideas for good moves, and how do you rule out the bad moves? In most chess positions, there are between 30 and 40 possible moves, and the great majority of them have to be ruled out instinctively. Otherwise we would never get anywhere in our analysis. Computers, of course, don’t rule any moves out, and that’s one reason they come up so often with ideas that shock us.

Here’s a game I played against the computer yesterday that really shocked me. Not the game itself (which I won), but the analysis afterwards, when I realized (with the computer’s help) that the move I played was wrong, and the move I discarded without even analyzing it was right.

Position after 35. … Rf8. White to move.

FEN: 5rk1/1p3p2/3qn3/pP1pNpPp/P1pP3P/2P5/KQ6/4R3 w – – 0 36

I was playing White and Shredder, with its rating set to 2220 so that it makes a mistake occasionally, was Black. I felt very good about my position. Although I am a pawn down, my knight on e5 is in a beautiful, dominating position, and my king is completely safe, so that I can concentrate all my forces on attacking Black’s weaknesses and attacking his king. And there are lots of weaknesses to attack: the pawns on d5, f5, and h5.

However, playing devil’s advocate for a second, we can also look at the position from Black’s point of view. Although his kingside pawns are kind of loose, after an eventual … Ng7 it’s not exactly clear how White will win any of them. I once lost a game against Jesse Kraai in somewhat similar circumstances. In that game, too, I had sacrificed a pawn to mess up his pawn structure, thinking, “Oh, I can win back that pawn any time.” But until you win it back, a pawn is a pawn. You can’t just dismiss Black’s material advantage. Furthermore, Black’s last move, 35. … Rf8, was almost surely made with the idea of (at some point) playing … f6, trading off a pair of pawns, and turning that weak extra pawn on f5 into a strong extra pawn.

I’ve given you this background so that you’ll understand why I never even analyzed the best move for White.

I took a time-out here, and I started out very well. The first thing I figured out was that Black’s “threat” to play … f6 is not a threat. For example, if 36. Rf1 f6 37. gf Rxf6 38. Qg2+ Ng7 39. Rg1!, an amusing situation has arisen where Black has no way to defend the knight on g7 without either losing the exchange or losing the pawn on d5 (after which his position would collapse). If instead Black moves his king instead of playing 38. … Ng7, for example 38. … Kf8, I was able to work out to my satisfaction that 39. Rg1 gave me a winning attack with ideas like Qg8+, Qh7, Rg8 or Re1, etc.

Now here’s what I should have thought: if the exchange of pawns is bad for Black, why should I wait for him to initiate it? Why don’t I initiate it? That would have led me to the candidate move 36. g6! If Black tries to keep the position closed with 36. … f6, then 37. Nf7 looks really good. It’s very hard for Black to keep White from invading to h6 with the queen and checkmating him. Probably Black would have to give up the exchange with 37. … Rxf7 38. gf+ Kxf7, but I think White is probably winning here because all of those weaknesses remain. On the other hand, if Black does trade pawns, with 36. … fg 37. Nxg6 Rf7 (or 37. … Rf6 38. Qg2 Ng7? 39. Qxd5! Qxd5 40. Ne7+ — a very thematic idea), White has tons of pressure not only on the three weak pawns but also on the g-file. The knight on g6 is a thorn in Black’s side that can go either to e5, or f4, or even to e7 or f8 with surprise checks, as in the blue variation I just showed you. The plan of Rg1, Rg5, Qf3, and Nf4 — winning the h-pawn with ongoing pressure against the king — is difficult for Black to stop. Rybka (the computer program I use for analysis) gives White about a 1.2-pawn advantage.

Why didn’t I see this? It’s so logical! Here’s why: Because I ruled out 36. g6 even before I began. Even with a 30-minute time-out, I never actually analyzed it. It was one of those moves that I “instinctively eliminated” from my candidate move list. Why? Because I thought it was playing right into Black’s hands. As I said before, it eliminates the doubled pawns and turns the (optically) weak pawn on f5 into an (optically) strong pawn that is defended by the rook. And even though I had just analyzed a line (the line in red above) where the combination of pressure on d5, h5, and along the g-file was decisive, I somehow thought of that as a peculiarity of that particular arrangement of pieces, rather than a fundamental truth about the position.

Now here’s what I thought instead. After realizing that 36. … f6 was not a threat, I felt that I had essentially a free move to improve my position. I looked at moves like 36. Qe2 and 36. Rf1, but I didn’t really see a way for White to break through Black’s fortress after … Ng7. (And indeed, there probably isn’t any way to do it without the extra punch of an open g-file!)

Then I drew on my vast wisdom accumulated by going to many of Mike Splane’s chess parties, and I asked myself, “What is the worst piece in White’s position?” And I came up with a stunning answer: the knight on e5! That’s right, when you think about it, the pride and joy of White’s position is all saddled up but has absolutely no place to go. Its only move is back to f3. Not only that, on e5 the knight is somewhat in the way of White’s rook, which also would love to come to e5.

So I asked myself, “How can I increase the power of my knight?” And I once again failed to consider the obvious answer: push my pawn to g6! If I do that, I will gain either g6 or f7 for my knight. But again, I couldn’t even bring myself to consider that possibility. And you can sort of understand why. After 36. g6 fg 37. Nxg6 the knight doesn’t look as if it has found a permanent outpost — it looks as if it’s just floating in air and will be driven away soon. You have to look at the position carefully to realize that it is in fact untouchable; for example, you have to see tricks like 37. … Rf7 38. Qe2 Nxd4? 39. Qe8+ Kg7 40. Qh8+ Kxg6 41. Rg1+ with mate to follow.

Instead I thought, “Aha! I can play 36. b6 followed by 37. Qb5! This not only gives me the possibility of winning back my pawn with Qxa5, but it also threatens Nd7, after which the knight will come to f6 and my game will win itself.”

So that’s what I played. And it worked! But it shouldn’t have.

36. b6? Nf4 37. Qb5? …

And now Shredder did me an unbelievable favor. According to Rybka, it should have played 37. … Nd3! 38. Nxd3 cd 39. Qxd3 f4! 40. Qb5 Rc8! Now all of a sudden it is White who has all the weaknesses, while Black’s previously weak f-pawn is turning into a menace! Rybka rates this position at +1 pawn for Black.

Instead, perhaps because it was set at 2220 strength instead of 2600 strength, Shredder played 37. … Ng2? 38. Re2 Nxh4? (diagram)

Position after 38. … Nxh4. White to move.

FEN: 5rk1/1p3p2/1P1q4/pQ1pNpPp/P1pP3n/2P5/K3R3/8 w – – 0 39

Shredder has won a second pawn, but allowed me to carry out my faulty plan.

39. Nd7! Nf3?

Probably 39. … Ng6 was best. Black has to give up the exchange. For instance, if 39. … Rd8 White has the wonderful finish (reminiscent of what actually happens in the game) 40. Qc5! Note that the threat is 41. Qxd6+ followed by 42. Re8+ followed by mate. So Black is basically forced to play 40. … Qxc5 dc. But because of the ongoing threat of the back rank mate, he is essentially powerless to prevent c6, cb, and b8Q.

40. Nxf8 Kxf8

And here it dawned on me that this was not a position where I was going to have to slog through a complicated exchange-versus-two-pawns endgame. It’s a position where I’m queening a pawn in five moves.

41. Qc5! Qxc5 42. dc …

Shredder can do whatever it wants on the kingside, but it can’t stop me from queening on b8.

42. … Nb1 43. c6! …

Sure, go ahead, take my rook.

43. … Nxe2 44. cb Ke7 45. b8Q Nxc3+ 46. Ka3 Black resigns.

You can see why I was super excited after this game! It looked as if I had won a beautiful game by asking the question, “What is my worst piece and how can I make it better?” But I was wrong! I only won because Shredder played too materialistically. In fact, I was too materialistic, too — allowing myself to be distracted by irrelevancies like winning the pawn on a5. And most unforgivably, my materialism prevented me from even considering the correct move, 36. g6!

What can we learn from this? How can we teach ourselves to think the thought that our brains refuse to think? This question may be beyond a poor mortal such as me. Or even Kotov.