Tuesday, July 05, 2005

The jogging has begun.

The last days I am doing Bishop-jogging.
That works as follows:
I let the computer play against itself.
I can adjust the time between two moves.
With every move of white I scan the rakes that are coming out of whites Bishops.
With every move of black I scan the rakes that are coming out of blacks Bishops.
The scan is a combination of an X-ray scan and a real scan. I will post later to explain this.

I have jogged for about 3 hours now.
The time I need for the scan of 1 Bishop is diminished from 30 seconds per ply to 10 seconds per ply, and is still dropping.
What is even important, is that it starts to consume less energy.
So this is a hopeful development.

Basically you can look at the scans as extended microdrills on a crowded board.
A blunderscan and a scan for tactical shots is broken down into its compounding parts.
These parts are trained seperate. In the hope that a total scan can be done at a glance or two in the end.
The idea is that this total scan will trigger my pattern recognition, when there are possibilities for tactics in the position. In this way I hope to integrate my pattern recognition into my OTB play.

What scans do I have to train?
1. Bishops.
2. Rooks. Out of the rooks come the same sort of rakes as the Bishops.
3. Queen. This is more difficult than I thought first.
It is compounded of:
  • A Rook scan (rake).
  • A Bishop scan (rake under 45 degrees).
  • First a Rook-move then a Bishop-move.
  • First a Bishop-move then a Rook-move.
4. Knights. The Knight-scan is still under construction.
5. A targets scan. See Mousetrappers blog for an explanation.

My tounaments start july 16th, so I have some time to excersise still.
After the last tourney we will know if we have found the missing link or that another experiment has failed.


  1. You know even if this does not work on your next tournament - it maybe better if you dont drop it too easily.. Maybe it will take some time before your brain fully grasps what you are trying to teach it.. so maybe give it a few months(?)

  2. Hm, my jogging speed seems to be much faster than yours. For K, R and B I need roughly 1 second, for N 2 seconds and for Q 3 seconds. As far as I understand you do 2 move scans, whereas I only do scans for 1 move. The reason of this is to avoid being lost in a huge number of squares that have no value. If a target is found then it is very useful to look for Knight, Bishop, Rook or Queen paths to attack it. I fear that path finding without a target is too much time consuming. So my strategy is as follows: On opponent's clock do a careful target scan, this may take one or more minutes. Keep book of the targets. When the opponent has moved, do a very quick 1-move scan from the new position of the piece. If this interferes with one of the targets, then you have got a threat. This should avoid most of the stupid blunders that are the main cause of losses on our level.

  3. Mouse, that's correct I do a two move scan. In doing so I noticed that there is only a limited amount of rakes. So I'm confident when all the rakes are stored in my system, the scantime will drop to < 2 seconds for a bishop. I'm already on 7 seconds now.

    I'm going to implement the target scan too, so my approach will be from two sides.
    Since it's an experiment, the approach can differ overtime.

    Your idea for jogging is brilliant.