Thursday, August 09, 2007

An attempt to write a narrative.

White to play and win.

Likesforest has done a courageous attempt to analyze this position. Since it looks very much alike how I started with my own analysis, I assume he doesn't mind I repeat his analysis here:

I see the knight on a4 and bishop on b7 are hanging. I also see the queen is overworked defending the back rank and the f6 bishop.

If I could play d7 immediately, Qxd7 is almost forced and then Qf4 win a bishop or knight. But, I have to play d6 first and that gives Black a tempo to protect them or trade pieces.

Black protects his bishop -
1.d6 Bg7 2.Bxg7 Kxg7 3. Qd4+ (winning a knight)
1.d6 Bh8 2.Bxf7+ Kxf7 (forced) 3.Qb3 (mate follows)

Black trades pieces -
1.d6 Bxf3 2.Bxf7+! (2...Kxf7? 3.Qb3 mate follows) 2...Kh8 3.Qxf3 (up a pawn and bishop pair, Black's king is cornered, his knight/bishop still hang, and his rook/queen are lined up on a diagonal--more material will be won in the next couple moves).

Black protects his knight -
1.d6 Nc5 2. ... - I'm stuck. That knight move is strong, defending b3/d7/e6.

Wow, my tactical analysis is uh-gly! While these lines win the material that I expected, White has much better moves he can play.

That is how it goes when you analyze in a trial and error way. You recognize a few unprotected pieces, see the backrank mate and that is about it. I dabbled around for two evenings with this position, assisted by Rybka, and I could annotate a lot of the winning variations. Still I had the feeling that I missed the essence of the position. Why exactly is black so hopelessly lost in this position?

Then I started to construct narratives for every winning line. During this process the main point of the position was revealed: f7 can't be defended. At first sight that is far from evident. In earlier posts I stressed the importance of piece activity in the middlegame. In a later post I added that that has to go one step further: the ultimate goal of piece activity is invasion into the enemy camp. f7 is one of the places where the white pieces converge. Especially the bishop, the knight and the queen. What are the pieces to defend f7? The king, the queen, the bishop and the rook. But the king is vulnerable, and the queen and the rook have other obligations to look after (e8).

So white starts to activate his bishop by 1.d6
Let's assume that black isn't aware of the danger and plays 1. . . . Bxf3
Then 2.Qb3 follows and mate is unavoidable. Let's look at a line:
2. . . . Qd7 3.Re7 Bxe7 4. Qxf7 Kh8 5. Qg7#

This came pretty much as a shock to me. At first this line looks to fall out of the blue skies. But the bishop on h6 takes away the important defending square f8 and d6 introduces a new invasion point at e7.

If black tries to defend f7 with his only free piece, the bishop, white has another killer move.
1.d6 Bc6 2. Ne5!
All of a sudden you see that the knight is untouchable at e5. Black can't take it without being mated. Now for instance 2. . . . Be8 3.d7 Bxd7 4. Nxf7 Qc7 5.Nd8+ Kh8 6.Qe8+ Bxe8 7.Rxe8#

Again it is amazing where this one comes from!
So the only serious try for black is 1.d6 Nc3
This keeps the white queen from b3 and gives the knight no time to step up the pressure on f7 with Ne5. But here you see the third killer move based on the weakness of f7.
2. Bxc7+ Kh8 3.d7 Qxd7 4.Be6 and black can't avoid to lose a full rook.

Black has a few other tries, but almost all strand on one of those three killer moves. You see that the superficial characteristics of the unprotected pieces plays no role at all!

This process may seem to look not very effective. But this is what I try to accomplish:
First I want to find the essence of the position by constructing narratives. That may take a few evenings. In this position I was obviously able to find the essence.
Second I ask myself, what do I need to see to find this essence at once? If I can find the essential characteristic how I can recognize this weakness, then I trust pattern recognition will do the rest and help me to recognize this pattern in thousands of different positions with equal characteristics. This part of the theory isn't proved yet.


  1. Doesn't the unprotected knight play a role in the line 1. d6 Bg7 2. Bxg7 Kxg7 3. Qd4+ and Qxa4 next. Or am I missing some other way to win?

  2. Loomis,
    no, the unprotected pieces play no role.
    For instance:

    1.d6 Bg7
    2.Ne5! Bxh6
    3.Qxh6 Qf8
    4.Bxf7+ Kh8


    2. ... Nc3
    3.Bxf7! Kh8
    4.Be6 Qf8
    5.Nf7+ Qxf7

    Right now, thanks to the narratives, I understand all the lines. But the problem is that I don't see them. Which gives me no guarantee that when I look again at this problem over a year or so that I will remember what to do. I'm convinced that it must be possible to actually see why those killermoves work. So one piece of the puzzle is still missing. I'm trying to find out.

  3. So perhaps we should ask ourselves: what is the Big Idea in each position ? Look at structure first with no or minimal calculations. The attack of f7 is the Straw that stirs the drink.

  4. I don't mind being used as a (bad!) example at all. It's an interesting position and dilemma.