Training the mini skills

The list of mini skills in the previous post gives a clear direction which way to go. I have a much clearer picture of how the training of mini skills should look like, nowadays.

Observations from practice showed that we are way less familiar with the basics of tactics than we are inclined to think. The basics are known, but only partly. And what is known from the basics, is only partly automated. Both issues must be addressed by a proper training:

• We must supplement the lacking knowledge (system II)
• We must automate the knowledge (system I)
• The training must provide sufficient feedback

The margins of error
System II
The margins of error are very small. When we use a problem set with problems that are too complex, then there is too little repetition of the important issues. Important issues drown in the flood of unimportant issues. It is mainly a system II exercise. But since the conclusions drown in the flood of information, we tend to draw the same conclusions over and over again, without ever realizing that we forget these conclusions in the time period between repetitions.

System I
When the problem set is too simple, we tend to work on autopilot. When we work on autopilot, our consciousness is insufficiently activated. When our consciousness isn't awake, there is no sufficient feedback from our training. No feedback, no automation. We have a name for this type of training: the salt mines.

Constraints of the training
We need a separate training set for each mini skill. "Recognition of the targets of a duplo attack" was on my list of mini skills in the previous post. Since "duplo attack" is a collective name for different tactical elements, we can segregate the list further:

• Fork/Double attack
• Discovered attack
• Pin
• Skewer
• Röntgen attack

In fact, it seems right to devote a problem set to every tactical label in the list at Chess Tempo.

• The training set must not be too big, since we need repetition. 200 problems, at max. Maybe smaller is even better. 
• The training set should not be too complex. I use problems with a rating between 1500 and 1550. 
• The set must be specialized to improve the repetition rate. I intend to make a set for every tactical label at Chess Tempo. Right now I'm working on fork/double attacks. 
• I use only problems with a maximum of two moves. That makes every problem relevant in accordance to its label, and it guarantees that problems are not too complex. The conclusions do not drown in the mass.

First experiences
The first impression of training with such set seems promising. Despite the fact that the problems are low rated and restricted to maximal two moves, a lot of new knowledge is discovered. It is quite surprising that there still is so much to know. The new knowledge is accompanied with new geometrical patterns. Whether that knowledge is going to be automated remains to be seen, though. I must probably play a bit with the margins.

Experiences from the past
I worked with problem sets that were labeled as double attacks in the past. But that were sets which were rated above 2000. And the sets contained a few thousand problems. So there was no relevant repetition. Solving took typically a few hours per problem. It was a typical exercise for system II.

I have worked with problem sets that were labeled as double attacks in the past which were rated below 1500. But since I focused on pace, and the problem sets were very big, there was little room for feedback, there was little relevant repetition and there was no (time for) developing of new knowledge.

Source: wikipedia