What do you do with your analysis?

Patrick raised an interesting point.
It is better to use your time to solve problemsets than to build them.
As long as existing problemsets meet your demands that's true of course.
But take a look at this diagram from my game at the club against a 1900 player.

























FEN 8/p5pp/1pkpr3/5p2/1PPR1P2/8/P2R1KPP/4r3 w - - 4 25

White to move.
I played the black pieces and my last move was Re6?? blundering a pawn away. And hence the game.

Why?
Why do I have trouble with such a simple position while I fully analyzed a 10-fold more complex one in my post of june 17th with ease?
The reason is that you won't find much of such problems in problemsets.
Since I restarted playing in 1998 I have barely seen such positions on the board.
Because I avoided it.
But lately I changed my playing style. I want to develop a second weapon next to king attacks.
That weapon is to simplify to an ending and try to win that with endgame technique.

It is easy to analyze the above game.
But what am I going to do about it?
Note to self: don't blunder pawns away in the future.
That's not going to work of course.

Analysis will only be useful if I use it as feedback to correct my play.
That seems logical, but how many people do that?

When I make a problemset with exactly these kind of problems I can become familiar with the patterns and the next time my ears will fall off by the ringing of alarmbells in the above position.
Just as it is now at this very moment impossible to walk with my King into a pin without a red alert flashing.

You will probably have a hard time to find a problemset that covers the above kind of problem.
So why not make one of your own?
That's not as hard as it looks:

• Take a good gamebase.
• Select with SCID the games with the material balance as above.
• Let Fritz annotate the selection.
• Export the selection as PGN and filter only the games that meet certain conditions by processing Fritz' annotations.
• The problem starts one move before Fritz added a "!"

The only work is to write a filter for PGN files which is fairly simple.
The next time CTS is down I will do that (O no, thanks to Blue Devil I will work with PCT then:)
Speaking of which, even Margriet dropped her Sudoku addiction for PCT.

As Spacecowboy pointed out the problemset can be maintainted with Bookup.
Bookup has a training tool for exercising.

CTS at home

Since CTS proves to be an uncertain factor it becomes clear I have to find a way to become more independent of it. So I started to research a way to generate my own problem sets.
The basic idea is to automatically rip existing gamebases to create problemsets of my own.

This gives additional possibilities:

• I can create sets for area's where I'm weak at. For instance with endgame problems.
• I can reduce the amount of problems in a set so that spaced repetition becomes possible.
• I can experiment with the solving time.

Generating problemsets will not be so difficult with existing programs.
But CTS has 2 strong points which aren't so easy to copy without making a program myself (which I like to avoid!):

• The automatic rating of problems .
• The time system.

Maybe the first point can be solved by generating problems in a set with identical degree of difficulty. The timing system is another matter. Maybe I can make use of the timing system of chessbase. I have no idea if that is possible.

During my research I found a lot of usable problemsets on the web already.
For instance at Gunther Ossimitz' Chess page

The 8th circle

Since CTS is down I review George Renko's Intensive Course Tactics (I) again.
I did the CD already 7 times, more than a year ago.
While doing now the problems for the 8th time I estimate that in 25% of the cases I can reconstruct the answer fast, since I remember (parts of) the answer due to doing the 7 circles in the past. In 25% of the cases my new acquired skills at CTS helps me to get the answer (!). From the rest of the problems I remember vaguely that I have seen them before, but that doesn't help me with the answer.

So the results of the 7 circles after a year are somewhat meagre.
Of course I would do much better after 2 more repetitions now, but overall it isn't impressive.
I belief that has to do with the fact that the problems are rather complex. The combinations exist of ca. 6-13 parts each (see my post of past saturday and monday).

What WAS impressive though, was the amount of problems I could solve easier with the help of my new acquired skills at CTS. This is the first time I notice a cross-effect of doing problemset A influencing doing problemset B.
If the skills acquired at CTS help me in 25% of the cases of George Renko's Intensive Course Tactics (I), then it will help me in 25% of the cases of ANY given problemset.
The figures are estimated and rather subjective of course.
But no matter the exact figures, this is really an important fact!

It means that doing zillions of very basic problems helps to solve complex problems too.
I already hypothesized that, but now I have found the first shreds of evidence.

Man, I wished CTS was up again!

Panic!

I get an "invalid database" error at CTS. Does anybody get the same?

Gap closed, let the progress commence!

My OTB-vision has always been better by a 70-100 ratingpoints in comparison to my screen-vision. Which means that when I'm in difficulty with a problem at my screen, I often see the solution immediately when I set it up on the board.
Since I have always trained on a computer, except for papa Polgars brick, I always exercised at too low a level.
So my training with the computer couldn't have much effect on my OTB play.
Which it didn't indeed.

Last year I have done 47,000 problems at CTS and improved there with about 70 points, thus closing the gap between my screen-vision and my OTB-vision.
When the gap is closed, traning with the computer will effect immediately my OTB play.

I'm inclined to belief that my previous post on saturday shows that. That was about a combination last friday during an OTB game at my club.

In an earlier post a few months ago I formulated the hypothesis that complex combinations consist of easier parts (that's why it is called a "combination" in the first place, isn't it?) and that when you learn to solve those easy parts a tempo, the solution of the whole problem would be faster and easier.

That was the main reason I started with CTS, which is all about learning to solve simple problems a tempo.
The diagram in my last post on saturday seems to prove that hypothesis.
There are at least 10 parts of the combination in that diagram.

• Removal of the defender of Be2 by Rxc3
• The pin of e3 along the e-file
• The lack of space of the white queen
• The knightfork on e2
• The attack of Nf6 by Qh4 and Bc3
• The mate threat at g7 by Q and B
• The mate threat at f7 by Q and N (via Bxf6, Ng5 and Qxh7)
  
• The underprotection of d5
• The unprotected Bishop at d6
• The mate threat with Nxf2#

9 parts of the combination were spotted immediately.
The calculation was about how the parts influence each other.
I only didn't see the the mate threat with Nxf2#, which I saw only after he moved his King to h1.

Bellon gambit

For no apparent reason I always have a lot of success with the Bellon gambit against the English.
(1.c4 Nf6 2.Nc3 e5 3. Nf3 e4 4. Ng5 b5!?)
I played against 1862.

This position was reached at move 19























Black to move.

Thanks to CTS I spotted here the winning combination almost immediately.
Can you find it?

Partial malfunction

My estimated average rating at CTS improved another 10 points, according the expectation of doing another 6000 problems. So my total gain at CTS is now 70 points (1470-1540)

I'm experimenting with doing CTS problems early in the morning, when the brain is still thick and sleepy. In order to find out which part of the solving process is slowed down by sleep.

A typical solution goes this way:

• The eyes are drawn to a certain part of the board.
• Zzzz
• It costs an enormous amount of effort to see what's going on there.
• Zzzz
• To draw the conclusion that what you see here is not going to work is the following time consumer.
• Zzzz
• The moment you decide to look at another part of the board and you swap your eyes around the pattern recognition takes over at lightning speed and you see the solution instanteneous.

Trying to force a speed up of the process feels almost painful and doesn't work.
Time for coffee.

An inquiring mind needs to know

Yesterday I drunk quite a few beers.
I stayed to long in the sun.
Hence I got fever of 38.4 C (101.1 F)
My head was filled with cotton balls.
I felt that I couldn't perform any intellectual task.

That was a good moment to test the statement that your IQ is not influenced by how miserable you feel. That is a typical counter intuitive scientific finding.
So I did another 100 exercises at CTS and indeed, my performance was not degraded.
Actually I scored an all time high of 1564!

Update:
When I have a good night sleep, I perform badly at CTS during the first few hours after awaking.
Today I tried to see why. That was very difficult, since I felt rather musty.
Pattern recognition worked at its usual lightning speed though. So another part of the process is slowed down.

Usually we divide the solution process in two parts, pattern recognition and calculation.

Pattern recognition.
The speed of pattern recognition isn't influenced by alcohol and fever, nor by sleep.

Calculation.
That leads to the conclusion that alcohol and fever don't have an influence on the speed of calculation either but a good night sleep slows the calculation down (I hope you can follow this).
That's hard to imagine.
Are there other factors at play which I don't know of?

Another conclusion is that the brainprocess that is usually slowed down by alcohol, is not involved in solving chess problems.

So the question rises: "Which brainprocess which is involved in the solution of chessproblems is unaffected by alcohol and fever, but slows down after much sleep?"

Back on track, the experiment starts.

After 3,500 problems and 7 days I have an average rating of 1530 at CTS.
That´s exactly what I had when I left CTS a half year ago.
So efforts aren´t down the drain after all.

Now comes the interesting part at CTS.
I have done about 44,000 problems at CTS.
This means that I have repeated the problem window of 10,000 problems more than 4 times.
Usually after 4 circles is where calculating ends and pattern recognition starts.
And indeed I´m starting to recognize a lot of problems.
So let´s see where the 7 circles of madness leads me.

For the improvement from 1470 to 1530 I had to learn 2100 new patterns allready.
That took me 7 months and 44,000 problems.
So to master 1 new pattern you have to solve 21 problems at CTS.
This doesn´t look very effective.
The main reason for this is that the problemset is too big.
So it takes about 1.5 to 2 months to repeat the whole set of 10,000.
In the mean time my OTB rating improved from 1710 to 1750.

To improve further from 1530 to 1600 average rating at CTS I have calculated that I have to learn another 2300 new patterns.

Speeding up

The rating system of CTS has as result that the problems that are presented to you will cost you 10 seconds per problem at average.
That means that they will look fairly simple to you.
If you can do the problems that cost you 10 seconds in a shorter time, your rating at CTS will increase. Finally, when you do all problems in less than 3 seconds, you will have the maximum reward.

So doing problems in 3 seconds in stead of 10 seconds is what CTS is all about.
From this arise three questions:

1. How useful is this?
2. Is the improvent from 10 to 3 the best or are other figures more efficient? For instance from 2 minutes to 3 seconds or from 5 minutes to 1 minute.
3. Is CTS the best way to implement this?

Ad 1. Every master or grandmaster will score high at CTS. So it is a skill they have mastered in the past. Unknown is how much it contributes to their play. But it seems to be unavoidable to develop the skill.

Ad 2. About the 3 seconds part of the question: 3 seconds seems to be a reasonable time for the brainpart that is responsible for processing visual data and pattern recognition. If you ever want to give a simul this speed is necessary. Susan Polgar scored 2.4 seconds per move at her monster simul lately.
About the 10 seconds part of the question: I have no idea if this is best.

Ad 3. Using the theory of spaced repetition as ideal model, it is clear that the problemset at CTS is far to big and hence the rate of repetition is far too low.

After doing 2k+ problems last week my rating averages at 1510.
This is 40 points more than my start (was 1470) and 20 points less than my maximum (was 1530). So the 40k+ problems done in the past aren't totally down the drain after all.

Please move up a little

to create room around the table for our newest Knight Gregory.
He has allready done an incredible 3 circles.
Give him a warm welcome!