Saturday, January 30, 2016

Aura vision

Aox has, a bit by accident I presume, developed a nifty program to train aura vision: mate in 1 - hard.
As usually I'm overly critical about everything and take nothing for granted. That makes me quite annoying, not in the least place for myself. But there is nothing mean, or personal, or stubborn or nagging meant. I just doubt everything, that's a habit I had all my life. Twelve years of fruitless chess improvement has made me an expert in being wrong. Kinda. So I hope you will bear with me.

A bit about automatic training.
Although this post is about aura vision, I can't resist to talk about speed.
In the past we talked a lot about speed and automatic training. I came to the conclusion that training has to be conscious to yield results. There remained one clear exception though that contradicted this statement. I became better at playing Troyis by just playing it without any thinking whatsoever. In fact that even made me perform like a 2600 rated grandmaster at this exercise. The skills acquired by Troyis transferred to the chess board.

So I must admit that there are areas where unconscious automatic training will yield results. But you have to be careful. In the first week you make the biggest progress. After that you will plateau soon. But hey, who cares when you are already performing like a grandmaster?

I think that when you stall at mate in one, you already might have a very reasonable level of performance. It doesn't make much sense to try to improve beyond that level.

With Troyis you soon reach a plateau beyond which further improvement is next to impossible. There is a method though. When you give up automatic training, and try to figure out a strategy consciously. You can figure out a few rules, and integrate that into your play. The same is true for mate in one.

Well this are just a few thoughts to hopefully give you new ideas. If not, consider it not written.

Aura vision.
Let's talk about aura vision.

White to mate in 1 (hard)
The aura that radiates from the white pieces forms a cage around the black king. I take my time to visualize-II that cage. Since visualizing-II is the only way I know to communicate with my unconscious mind. After five days of training, it is hard to say if it works. It feels like it does, but usually this period of training is subject to euphoric bias, so I'm cautious. It certainly does not show it is working in a measurable way yet. For that a few weeks more training is needed. I intend to walk this path and to postpone the development of a thought process a little. I want to focus my energy in order to find out if this path is a dead end or not.

Thought process.
In fact, I can use these aura vision exercises perfectly well to investigate the ins and outs of a thought process. In each position there might be quite a few things going on:
  • Are all squares around the black king under attack or blockaded?
  • Is my king in check?
  • Is the piece I want to move not pinned against my king?
  • Is the square I want to move to not covered by the enemy?
  • Can the opponent interpose the check I intend to give?
As you see, there are quite a few subtasks to perform. All subtasks are closely related to the four actions of a piece that Radovic talks about:
  • Attack
  • Restrict
  • Block (shielding the line of attack)
  • Protect
So the training might yield results that transfer to other areas of the game.
I must be careful to train slow enough though, otherwise these subtasks may soon disappear from the training. If I try to improve the speed of the exercise in stead, it is easy to adopt a few strategies that are counter productive. For instance, I know that the black king will be mated on the square it stands on. So I can most of the time skip the check if all squares around him are covered. That changes the exercise from finding the mate to finding the right check. Since I'm not punished for a mistake, I might as well replace thinking with trying. I even found myself playing Qxg7 without even checking if the queen was protected there. It had to be, otherwise it was no mate. That way, speed prevents me from training the right subtasks, and leads easy to bad habits. So slow training is paramount.

Subtasks.
Even the "simple" task of mate in one proves to be pretty complex. There are rumours that it is a good idea to split mate in 1 in several subtasks. There might be a way to simplify the exercise by dividing the database in different categories. Now there are "hard" en "easy" mates. I can imagine different categories of mates:
  • No redundant pieces.
  • No pinned pieced.
  • Only white pieces.
  • No black pieces that can interpose.
  • No checks possible on a covered square.
  • Etcetera.
Every category has a complement. Pinned pieces being the complement of no pinned pieces, for instance. Just philosophising.

@Aox, thanks for this beautiful program!

13 comments:

  1. There are skills which can be improved to masterlevel by just doing the special chess-related exercises without any special thinking process, excample: Troys, Attackers,Defenders, Skill-trainer of Ashley...
    And there are exercises which are seemingly impossible to improve by just doing these over and over again examples: tactic puzzles, Mate in one easy ( at least for me ).
    I call these different exercises or skills "improvable" or "not improvable". The obvious difference between these both problem classes is the complexity, improvable exercises are easy, a decision can be done typically in less than a second; not improvable tasks are complex, several pieces are involved at the same time, a decision needs typically way more than a second.
    The goal has to be (IMO): to make not improvable tasks somehow improvable. The only possible way to do so is seemingly to find improvable subtasks and improve these until the original task got that easy that it gets improvable by itself.
    Your thought proces is trying to do exactly that. The complex decision that a move is a solution to the problem mate in 1 can be split in subtasks like the useful question "are all squares around the king blocked AT THE SAME TIME" or even further "is every single square arround the king blocked CHECKING THAT ONE AFTER THE OTHER" and so forth. Each of this subtasks is simple, based on a decision which takes less than a second. So we may hope they are improvable. And if we improve them enough the whole mate in 1 might become that easy that we can improve the whole mate in 1 task without any further thought process in one step or at least without further dissection of tasks we did not think about.
    Your thought process does make a complex problem to a sequence of simple sub problems.

    ReplyDelete
    Replies
    1. So we seem to agree on most important things here. Bummer ;)

      Delete
    2. How did you split the hard from the easy mates? Do they stem from the same database? What criteria did you use?

      Delete
  2. The hard ones where extracted from mate in 2 compositions, the easy ones are taken from 6 million real human games. I am not shure if i did transform the black mates to white mates? or did i just skip the black mates..?

    ReplyDelete
    Replies
    1. Aha, that's why I get a difficult position every now and then with M1-easy. Which is devastating for my average.

      Delete
  3. @Aox: How did you extract the mates? Did you do it manually, or did you have an extraction program (a commercial program or a program that you created)?

    ReplyDelete
  4. Dear Aox

    Is it possible to share this database? I have one guy request - who run a project - and he needs a lot of checkmate puzzles. Here is an exceprt:

    "I need hundred of thousands of tactical problems to feed my dataset. The purpose of tactics is that I can find out how people behave to new problems and how they persists in reproducing the very same problem.

    If we persist for some problems, I suppose that it's compulsive and I can see it with my EEG (in a brain region called the mesolimbic system). When we have a compulsion the cortex doesn't activate and chess problem are a great tool because if we do the same mistake we see that the cortex doesn't activate as it should. So this could explain why we persist to smoke cigarette even if we want to stop.

    I'd like to help people who want to stop smoking to overcome their addiction (the problem lies in the hedonic value lower than the incentive salience)."

    Let me know Aox if you would like to share your database with me (I can send him via private message). I would be very grateful (and I could test the puzzles at the program - skipping all mates as a training and measurime time used).

    ReplyDelete
  5. I think the idea of "box around the King" is quite useful - especially when the mates are not typical (standard ones). And your thought process looks very simple and powerful! I really like it and I am going to give it a try! Some of the elements are really hard to me (controlling the squares and pinned pieces).

    ReplyDelete
  6. in what format does this person need the "database" as pgn?

    ReplyDelete
  7. My friend has already replied:

    "PGN would be fine. Please also thank your friend too for me!

    I'll use CQL (http://www.rbnn.com/cql/) to sort the positions according to motifs/themes and an engine to check and add a weighted value using my algorithm".

    Let me know if you could help him anyhow. And if you need my email address for further correspondence - just let me know (I have your email - as I asked you for one chess thing related to M1 puzzle training) and I can send you an email. Thanks in advance!

    ReplyDelete
    Replies
    1. Tempo has a link to these "programs" at his blog as "BV mate in 1 easy" in the section "Links of interest"
      Anyone can save the webpage on his own computer and look for the file
      "fen.js" ( will be estimatingly in a folder )
      Every line of this file contains the "shortend" fen of one problem. With a good editor with makro abilitys it should be "easy" to create a pgn from it.

      Funny program this CQL
      I did not test it but it might be able to generate such a "database" of m1 problems all by itself, extracting them from game-databases which can be found in the net or bought at chessbase or chessok
      On the other hand i dont know if CQL will be able to handle the FEN-tag
      To test that your friend may check is with the png at this site : https://sites.google.com/site/edusadier/theartofdirectmateintwomoves
      where he can find 145 000 Mate in 2 problems as pgn

      Delete