Tuesday, September 18, 2012

There are no categories of grand schemes

Quite some time ago I was working on sacrifices at f7 from a chapter in one of Papa Polgar's bricks.
I wanted to define a "grand scheme of sacrifices on f7". I had figured out that in order to make such sacrifice work, you need three pieces that work in a tandem: a knight (on g5 for instance), a bishop on the diagonal a2-g8 and a queen ready to jump on that diagonal.
Further e6 should be underprotected, for instance because there is a black knight at d7.
The targets are:
  • Mate
  • Smothered queen
  • Ra8 via a knightfork at c7
At the time I was very proud of it, and over the years I have used the knowledge two or three times to judge a sacrifice. It might have even won me a game, I can't remember.

So when I had memorized 178 high rated games lately and decided to categorize them, I was very optimistic. You probably have seen that character trait of mine one or two times before :)
What I found is that the positions simply are too different to push them in a conceptual framework.
I found categories like "incomprehensible moves" or "kingposition invasion I underestimate". But in essence that are categories of my mind. It describes the troubles I have with chess.

When I compare the positions within such category, they are simply too diverse to press them in a grand scheme. If I look back at my grand scheme of sacrifices at f7, I found that I had deceived myself. Maybe only 15% of the sacrifices in that chapter fitted in the grand scheme, after some proper pressure. And the reason for that I ├╝berhaupt managed to do so was that a lot of the pieces where still on their standard positions that belonged to certain openings. Bias and oversimplification obscured my view.

If you can't build an outer framework where tactical positions fit in, it is logical that we have a transfer problem. In the beginning one can come away with finding identical geometrical patterns like backrank skewers and the like. But when you have done some serious tactical training, you will find that you soon run out of usable patterns. Most of us already know those these patterns well.
Tactical positions are just too diverse to be mastered by geometrical patterns alone.

This means that we have to find a new approach.
Geometrical patterns are of finite usability.
An outer framework isn't possible.
The positions are just way too different. In fact that is the reason why chess never bores.
We have to look in a whole new direction.

Pieces fulfil functions. Pieces have relationships. I like to call these the inner structure of a position. In contrast to the outer framework we talked about.
If we learn skills that help us to see the inner structure of a position, we have something that transfers from one position to another.
Let me try to explain this by example.

Black to move.
You can find the solution here.
This is a very simple yet high rated problem.
There are only three reasonable moves, two of them are wrong. Or at least not good enough.

The pattern is well known, the white queen stands in the line of fire of a battery. The only thing you have to do is to find a suitable target for the bishop.

After an average thinking time of 4 min and 18 seconds
  • 116 out of 960 people played 1.... Bc6. Their average rating was 1926.
  • 393 out of 960 played 1.... Bb5. Their average rating was 1962.
  • 451 people played the right move  1.... Ba4. Average rating 2005.
If you take some time to contemplate those figures and compare them to how simple the solution actually is, an alarming picture emerges. From what kind of collective braindamage must we suffer to perform so poorly? How is it possible that we look at Bc6 while the answer Bxc6+ remains in our blind spot? How is it possible that we don't see the difference between Rf1 and Rd1 as a target?

The reason is that we only look at the pieces and the moves, while we are blind to the functions and the relationships of the pieces.

We have seen certain characteristics of the pieces, but fairly random so. We have seen that the bishop on e4 is unprotected. We have seen a mate with Qxg2. We have seen that b3 interrupt a bishop on a4.
But he, come on, that's no excuse! We collectively used more than 4 minutes at average!

If we want to learn something from this position we must learn to see the difference between Rf1 and Rd1. That is something that we can take with us to another position. No matter how different that position will be. So what's the difference?

The rook on d1 is at the focus of both Rd8 and Ba4, while the rook on f1 is only in the focus of Bb5. So in the first case you will win a whole rook while in the second case you will win only the exchange.

The method I use is to explicitly identify these relationships and to represent them in a diagram with arrows and colored squares. See how that works.


  1. The sac at f7 is "only" interesting for Bc4 player, less interesting for Bg2 or Bb5 player. At playchess.com i play from time to time against a player who play from the very beginning of his game against f7. We need to create situations where we can use such a knowledge

    I wonder if you still know/remember "the middlegame" of Polgar. I would expect that a player who "knows" polgars middlegame should be at least a 2000.

    "Most of us already know those these patterns well"

    How good is good enough? Susan Polgar is using parts of her brain which are responisble for face recognition to recognise chess pattern (this part is working in "hyperspeed"), Classplayer not.

    At school at break between classes some schoolmates and i where combating at collecting 4 leaf clover. Thats a very well known pattern. The average pupild did find non during a 20 min break, our little group did find 10+ each member. I did see 4 leave clover out of a corner of my eye when i was looking without focus at any specific clover.

  2. Yes, but your pet subject has nothing to do with the essence of this post.

    I gave up the middlegame brick soon once I discovered that 25% of the analysis of the book was flawed by Rybka.

  3. @BDK,
    I might agree. If I knew what you are talking about.

  4. I dont think its absolutly necessary to learn only 100% best moves. You could see Polgars Middlegame as a collection of methods. "Which method is possible", or "which method is used by a master" might be a different question then "which method is the ultimate best". So Polgars Middlegame could be used as a collection of "Candidate moves/ideas". I agree, it seems to be better to learn "only" the ( estimatingly ) best. Still you could have simply learnd "only" those 25% = #1000 (!) and learn from the errors of the master at the other 75%. But i remember how frustrating it is to find so many "faults". At Chess-Mentor there had been so many "errors" that i did end my diamond membership at chess.com.