Due to some close observations of what exactly happens during calculation everything falls into place. All 3 discoveries which I made in the past play a role plus, I hate to say it, visualization.
Those 3 discoveries were:
- backwards thinking.
- drills that enlighten the burden of the STM by transferring subtasks from conscious thinking to the procedural memory (think of playing Troyis).
- Identify the characteristics of the position.
- Evaluate the characteristics.
- Visualize the future position.
White to move.
Identify the characteristics.
To identify the characteristics of the position you have to formule little narratives. Those little narratives consist typically of only one sentence. Like (see diagram 1):
- Black threatens mate on g2
- Bishop on e7 is unprotected
- Knightphork a6
- Queen on b7 has little space
- Both rooks are on the same diagonal (skewer)
- Whites B and Q converge at g7
- Pawnphork on d6
- Knightphork on c6
- Pawnphork c6
- Backrank mate threatens
- b6 controls a5, where you would like to put your knight
A lot of the narratives above evaluate to"utter nonsense". Others look more promising. This conscious evaluation has to take place to direct the process of calculation. The very fact that you start with narratives and evaluation in stead of random trial and error of moves makes this backwards thinking.
After doing lots of visualization exercises I used to dismiss the necessity of these exercises with the following argument: I can't even solve these complex problems with perfect visualization (=eyes wide open plus using an analysis board) so visualisation is not paramount. So why bother? Of course you must make your visualisation as perfect as possible, otherwise it lowers your rating. But don't expect too much from it.
If you have a look at the position above, then a few characteristics of the position are missing. Those narratives can't be formulated at this very moment since they haven't emerged yet.
Take for instance the following diagram after 1.Nxb6 Nxb6 2.Na5 Qa7 3.c5
Black to move.
All of a sudden a discovered attack cxd6 emerged. That characteristic simply wasn't visible in the first diagram. This means you have to repeat the same process of characteristic recognition every few moves. At every junction of branches of the tree of analysis, that is. This nicely fits in with Aagaards stepping stones, by the way.
To improve at calculation it is necessary to make both characteristic recognition and visualisation as automatic and perfect as possible. At the moment I miss typically about 30% of the characteristics in a complex position, and I need a lot of brainpower for it. Which means that it is taxing for the STM.
The same is true for visualisation. Which means I have to repeat the same moves over and over again in order to keep them fresh in my STM.
No wonder that I'm bad at calculation!