Wednesday, June 15, 2005

I've been thinking lately

about a thoughtprocess. . .
Quite a few knights are busy with that now, which inspired me to have a look myself.

Some time ago I came across a system which worked with rectangles to scan the board in a sytematic way for tactics. It's a pity I don't know where I found it.
I believe it was from an American grandmaster, but I forgot the name.

I had a look at Dan Heismans list.
And at the list of de la Maza.
And at the lists of the other Knights.

So I decided to make a list of my own. Did I already mention I am self-opinionated?
I have two constraints:

1. It has to be simple. This means that everything that is self-evident is pruned.
If I am in check, I have to relief it. If I forget that, my opponent will help me to remember. So there's no necessity to record that in a list.
Time trouble is already bothering me, so long checklists are not going to be of help.

2. It has to help me to strengthen the spots where I am really weak.
From the analysis of my last 8 games emerged a certain picture of the problems I have.
The list has to be of help in this specific area.

To get some ideas I did an experiment.
I played blitz against the computer.
I used Arena with the SOS-engine, which is quite good.
I got pounded everytime, so I decided to lower the amount of plies the computer was thinking. To my big surprise I had to lower it to only 1 ply deep before I started to win the most of my games. I played a tempo, just the move that came up. without checking.
So that seems to be my level of chessvision when I play on automatic pilot:)

The idea I had was two write down the mistakes I make and to distill a checklist out of that.
If the use of the checklist is effective, I have to increase the searchdepth of the program.
Right now I have the searchdepth at 5 ply deep, and I score even.
I have a checklist of 7 points where I fail from time to time.
I will not publish it yet, because I am still experimenting.
In a typical game I use about 15 - 20 minutes now.
At Fics my rating is about 1750-1820, mostly gained at G15 - games.
So a Fics-rating of 1800 is equivalent with thinking 5 ply deep at average.
I have my program configured at 6 ply depth now.
I am working on my thoughtproces now, until it helps me to beat the program at 6 ply (max G15). If I manage to do that with a checklist, I know it's a good list, exactly aggravated to me.

SOPE circle 1: 66 down, 352 to go
Chapter 3 done.


  1. Great experiment TS! Let us know how it goes.

    BTW, has anyone collected all of the handles of the knights? I am starting to get confused with my notify list.


  2. Long checklist: My idea is to burn it into my brain so that I can go through the points in fractions of seconds or even simultaneously. I am not there yet, by far not. But I want to get there. And yes, Tempo, let us know the outcome of your experiments!