Tuesday, November 06, 2007

Provisional conclusion (Part V)

continued from part I, part II, part III, part IV and the Intermezzo. . .

The problem with investigating while blogging is that you don't know beforehand where your investigation is going to lead you. After a little detour along the beanfields we are back at duplo-moves again. Beancounting works fine as long as there are no duplo-moves around. The beancounting method reliefs the short term memory when both players pile up on one piece. So that is a little success. But it is a small area, yet it works well within the given constraints. Since one of the constraints is "as long as there are no duplo-moves around", the question arises "how do you know if there are duplo-moves around?".

In the past we have done a lot of work on the recognition of duplo-moves, for instance by the identification of the potential targets of a duplo move. See attempts like my rake-scanning and Christian's Target Feature Count. So to me it is an old question.

I have partly found a work-around in the past months. By focussing on concrete positional plans it proves often to be possible to avoid the complex tactical positions alltogether, so that I get around the question of duplo-moves.

I found that emulating a grandmaster meant "knowing what can be neglected without changing the outcome". Beancounting fell into this scheme because you can neglect the actual order of the captures. Everytime you find something to neglect, you have found a method to keep your short term memory free from loading.
Concrete positional plans fall into this category since they neglect an enormous amount of moves without being punished for it.

So if you want to improve, concrete positional play is the way to go. Since that is about moves that usually are counted in 0.3 pawn-points or less, you have to be good in tactics first! Otherwise you cannot achieve what you want without drawbacks. After 4 years of intensive tactics training, I'm ready for this step.

No matter how good these idea's are, you can always be bitten from behind by a duplo-move that you didn't notice in time. Let's see if we can find something new about duplo-moves now we have given a lot of topics a new place. You never know. It is a problem that we have to solve anyhow sooner or later.

Duplo-moves are the main technique to gain wood in a forced way. There are other methods, like traps and pawnpromotion. But these I like to see as niches. I find it convenient to ignore them for now. The "real meat" consists of duplo-moves.

A duplo-move attacks two targets at the same time. A target can be a piece or a square. The targets can be inline or not. The duplo move can make use of one or two attacking pieces. You find more details here.

I take one of the positions Glenn came up with.

White to move.

3 duplo-moves jump into your eye:
  • Discovered attack Ng5 or Ne5
  • Knight fork e5
  • Pin Q vs Nf6-Be7
1.Rxe7 is a typical prelimanary move which puts the targets in place.With ordinary beancounting the N on d7 was well protected by 3 pieces. But:
  • Due to a duplo-move (the pin), you can't count the N on f6 as defender of d7.
  • Bxd7 puts the black queen and rook in the line of fire of another duplo-move, the discovered attack
  • Qxd7 puts both the queen and the bishop in a knightfork (another duplo move)
All defenders of d7 were compromised by a duplo move. So you can use the beancounting, but you have to subtract the defenders that are compromised. In case d7 had an extra defender in the form of Rd8 or so, d7 would be defended enough. The same is true for h7, the Nf6 is compromised as defender by the pin.

Maybe I should have a look at the other examples of Glenn, to see if we have a tendency here.

1 comment:

  1. You mentioned my Target Feature Count of old days which was kind of a formula approach, and because we are humans and not computers, such an approach will never work in a practical game. Bean counting, feature counting, all this stuff is far too time-consuming.

    My own approach these days is much simpler and may be summarized as looking for weaknesses (targets) on both sides.

    There are two kinds of these: positional and tactical. So you always must be aware of the «tacticality» of a position which may be somewhere between zero and a hundred percent.

    In positional play, the key question is to decide if a weakness is dangerous or not. If you attack, you create weaknesses in your own position but you must be sure that the opponent cannot attack them. So the crucial point is to weigh the weaknesses on both sides and to keep the better end.

    In tactical play, just the same. Look at the GMs who play at the verge of being checkmated, coolly going for attack and winning.

    Don't say we are no GMs. We can do just the same, only on a lower level. But the recipe of success should work for us, too.