Friday, October 26, 2007

How a grandmaster sees tactics

In my previous post I claimed that a grandmaster would recognize the right move in the following diagram à tempo.

Of course I can't backup this claim. But that doesn't mean I can't speculate about it and formulate a hypothesis. The claim is based on the positions that GM Danielsen played correct à tempo in his blitz games. This concrete position stems from PCT though.
It took me 6 hours with the aid of Rybka to form an opinion about this position. There are many tiny little details that play a role, and there are some red herrings as Loomis called it. This brings me to the first part of my hypothesis:

There is so much going on in this position, that even a grandmaster will need much time to see all the relevant details.

If this part of the thesis is true AND my claim is true THEN

A grandmaster must have found a way to simplify this position in his mind in a way that doesn't change the outcome.

And that is the second part of my hypothesis. If you look at the comments at my previous post, you will find several indications of methods how to simplify the position. The one of Christian is very strong for instance. On the other hand, different commentators found different things difficult. A few hadn't seen that the bishop is hanging.
The fact that I brought up the position was because I found it difficult as a whole. How on earth can you be sure so fast that you can leave your bishop hanging without losing a piece at the end of the combination?

For your convenience, this was the position:

White to move.

The question I try to answer in this post is if it is possible to simplify the position in the diagram in such way that you can be sure what the outcome will be the same? I mean, if my hypothesis is true, then I have to learn to simplify positions logically.

I try to break down the position in it's constituent elements. Let me first think a few pieces away:

White to move.

The pawn at d4 is attacked 4 times and defended 4 times. Usually that should be enough. And it would be, if it wasn't for the black queen that is badly placed. If the queen was on d8 and that rook on d6, there was no problem. But since the queen stands in front of the rooks, white can take the pawn on d4 without problems. That leads to an equal position where two rook are traded for a queen and a pawn. The position above shouldn't cause you too much trouble. Now let's add a bishop on both sides:

White to move.

The white bishop adds a pin to the equation. This means that effectively there is one defender of d4 less. This only means that the pawn on d4 is lost. I think we have a critical situation here where the difference between a grandmaster and me shows. Because here it is that I see all kind of phantom threats and ditto moves. I have to check every candidate move here before I dare to say that d4 is lost. But a GM simply says, based on his experience, "due to the pin the pawn on d4 is lost". If I'm able to formulate a narrative why such simplification doesn't change the outcome, no matter how much candidate moves are investigated, I would be able to see the simplicity of the position myself. But I can't. Yet.

Since the black knight is pinned, Nxd4 means the build up of a discovered attack. So black MUST either move his queen out of the way or he must counterattack a white piece (with a6), otherwise he loses a piece.

The black bishop adds an attack against c2.

If we add another white bishop and a black knight, we get the original position:

White to move.

White can play Nxd4 safely right away, because the battery guarantees that white gets his piece back if black takes on g5. If black doesn't take on g5, the situation is as above, black must save his queen or play a6.
So it is my take that the grandmaster makes another simplification here, which goes as follows "I can take safely on d4 without losing a piece. The only risk is that black can win a pawn back in the end." Since it is blitz, he leaves it to the opponent to find out if he can get his pawn back.

This must be how it works. Learn to simplify!


  1. Cool breakdown! The question is how to get to not having to think about things like that, but just seeing the answer in the fully complex position without any jiu jitsu.

  2. Very interesting. Just a point of clarification, though, is this concrete tactical tactics, vague tactical tactics or positional tactics in your mind? :)

  3. The question is how to get to not having to think about things like that, but just seeing the answer in the fully complex position without any jiu jitsu.

    Could it be as easy as continuing to increase the difficulty of the tactics you train on a daily basis?

    On the other hand, it's like you said Tempo, there are a lot of things going on in the position. Maybe instead of taking in all the details and candidate moves in the blink of an eye, one needs to learn what details can be ignored.

  4. Glenn,
    I hoped to give you a subtle hint that your definition of tactics is overstretching usefulness by putting "tactics" behind everything. In vain, of course:)

  5. SP,

    Could it be as easy as continuing to increase the difficulty of the tactics you train on a daily basis?

    No, that is not sufficient. I will post about that later.

  6. In vain, of course. :) Of course. :)

    Ok, but what you are talking about in this and the previous post is that the GM can see these simple (for a GM) tactics very, very quickly.

    And, this is just tactics. And it is concrete. That is what we are talking about not some vague "the GM can find good moves that have nothing to do with tactics".