Saturday, February 27, 2016

Decomposing the tasks

When a task cannot be improved in speed, it is usually due to the fact that it is too complex. We have to decompose these tasks in more simple subtasks, and see if we can make exercises for them.

In order to improve in FAC (Find All Checks), I trained the subtasks for which we already have exercises. I managed to improve those to 35 correct clicks per minute, which I consider good enough, for the moment. One exercise causes me trouble though: FAC with No Queens. I improved from 20 to 25 cpm, but since then I'm plateauing for two days. In the last session I wrote down what costs me time:

  • Switching to the next position
  • Switching to the other colour
  • Switching from one piece to another
  • Checks with pawns are overlooked
  • Checking with a piece that turns out to be pinned
  • One check with the bishop is overlooked while the other is not
  • Checks by taking a piece are overlooked
  • Checks by pawn takes pawn are overlooked
  • Discovered checks are overlooked when the front piece is a pawn
  • The wrong king is checked
  • Can't find the king
I made another list  while doing M1-e:
  • Give check on a protected square
  • Abandon a check because you look at the wrong piece of two pieces who can give check on the same square.
  • Choose the wrong check, the one that isn't mate.
  • Abandon a check because you overlook that you cover the checking square
  • Overlooking an escape square
  • Overlooking the simple mate because being focussed on the complicated non-mate
  • It takes time to check all escape squares
  • Trying to mate the black queen
How can we categorize these lists with time consuming issues?

When we switch from position, colour or attacking piece, we need time to orientate and to adapt.
We could take the following measures:
  • Multi clicks in one position.
  • No switch of colours.
  • Subsets with the same attacker.
When something has gone wrong in the orientation phase, we might find ourself trying to mate our own king, or the opponents queen. Luckily this doesn't happen too often.

There is a whole host of things that we tend to overlook.

Can you find more issues that eat up time during FAC or M1-e?



  1. I will make a comment based on my experience (and trials and attempts to overcome it):

    Switching to the next position - most often I solve the puzzles WITHOUT the look (sight) of the table. I increase the screen as NOT to see the numbers. This way I force myself to "solve as best as possible to go next puzzle". Sometimes it fails and then I check out how many solution is needed. However it is just 3-5% of puzzles.

    Switching to the other colour - this task is not so difficult to me. It takes me about 2-3 minutes to adjust to new condition.

    Switching from one piece to another - I try to make some kind of organized order, but there is a room for improvement. Anyway I try NOT to click random squares, but "all squares for one piece and then the next". I start with the simplest (Bishops and Rooks) and finish with the most difficult (Queen).

    Checks with pawns are overlooked - it happens to me too. The worst scenario is when the pawn is very far from the King... and the ONLY (lacking) move is to promote it. That problem is in my case present too.

    Checking with a piece that turns out to be pinned - it is much more difficult problem. Anyway some problems requires to mate with the piece (relatively) pinned! The worst scenario is to mate with fully pinned piece... as it is illegal move. When I solve mate in 1 HARD - it is very often the case :(. It harms the score (and performance) quite much.

    One check with the bishop is overlooked while the other is not. I feel the pain, but I try to see the board wider - most often the Bishops make check in a very long route and/or at the edge of the board. These "long and edge" moves should be another post (article) as this topic is very interesting :).

    Checks by taking a piece are overlooked. I was sick of this, but it turns out I overcame it (at least to the degree I am not frustrated with that anymore). Hard to say what caused this effect, but probably a concious decision to overcome it and about 3-5K puzzles solved this way.

    Checks by pawn takes pawn are overlooked. This problem was not that big as previous one, but important. Nowadays it looks like it has gone. Sometimes I hit it when I find all the rest moves, but simply FORGETS to check out pawn checks.

    Discovered checks are overlooked when the front piece is a pawn. I had this problem, but in a tiny extend. Why? Because I am a double (discover) check lover - I analysed this element to the core and whenever I see this possibility I ALWAYS check (try) it out! It is like an obsession - especially when it is a double check (with promotion). Then I cannot hold on myself on cold hands ;) :)

    The wrong king is checked. It happens to me when I am really tired (exhausted) or when I think very deeply about other things - not about the puzzles I am solving at the moment. Anyway it is rather "a song from the past" in my case.

    Can't find the king - it is an extremally seldom case. It happens to me when the King is hidden at a very untypical hoard (group) of pieces. Most often I can find the proper King within 1,5-2 seconds!

    I will make another comments related to M1-e. I can just say, I had to overcome the same type of problems as you mention. The only way is to describe, analyse and explain the nature of mistakes... and find any way to overcome it. I have to admit your articles and (new) tools help me VERY MUCH at thinking over the methods and ways to eradicate these nasty problems and errors (mistakes). However it takes time, effort and dedication. The most important is to "make a concious effort to get rid of the problem" and practice it until it has gone (or minimalize to the level you can "go through").

    Let me know if this helps and if you need my next comments related to the problems you listed above.

    1. 2-3 minutes to adjust to another colour? I hope not! What I mean is that the check exercises require to check with both colours. We must simplify subtasks as much as possible. That means that we have to eliminate these colour changes within the same exercise. By splitting the exercise in two, we have still the same amount of positions, but we get rid of complexity that makes the exercise a source for plateauing.

      The exercise that helped me a lot, is the FAC with only white the white queen and the black king. It has the right simplicity that makes it easy to improve, and the subtask has a high frequency of occurrence, in all sorts of tactics. BTW, this exercise is very close related to the microdrills of MDLM.

      I'm looking for idea's for new exercises, and the first stage is to identify all eaters of time during an exercise (no matter which). So please comment, when you find these little buggers.

    2. If we want to go to the simplicity - yeah, we should definitely make "one colour" puzzles (only white vs only black)... unless we master this.

      And what about the time consumers? Here are some suggestions:

      1) JUST one solution - this way you could have hide the table and do as fast as possible without wondering why the hell the puzzle will not changed (to the new one) even if you marked/clicked the correct solution. It is only (especially) valid to the position with mate in 1 (easy and hard).

      2) Less numbers of pieces - if you have 30 or 40 pieces on the board, the probability you miss some themes (pins, illegal moves, captures from the other edge of the board, etc.). That's why the puzzles should be limited up to 15-20 pieces (and we can increase the number of pieces when we start scoring really well).

      3) Mate training with the use of blocking/interfering. Most puzzles I have great difficulty are based on the "mate with the blocking the open lines". In real games it is unproductive and impractical. It is the same as if you decided to undeveloped your pieces to the original squares (1st or 8th line/rank). It does not make any sense to average chessplayer. That's why we would not be able to perform well at such puzzles... unless we would divide it at the simpler puzzles. What could it be? Show attacked squares around the King (only King box squares) which are attacked MORE than once. I really hope it should (could) help - at least in my case.

      4) Find the unprotected (not defended) piece around the King (King box area). Most often I make a "mating move"... only to notice that I forgot to protect (guard) the unprotected piece under King's attack. The result? The King just take that piece and it is NOT mate!

      5) Make a check with one piece - the same like "Knigh MATES" (Bishop mates). There are ONLY two pieces - the King and specific piece (N, B, R and Q). It looks extremally easy, but unless we perform a very good score - this may slow down the other (a bit more complex and difficult) elements.

      I think it may be enough for now. I am so frustrated that I want to shout loud! It is ridiculous we cannot break (crush into the pieces) such easy puzzle! And these puzzles with a big degree of complexity should be broken down into less pieces. Mating the opponent's King when the board if full of pieces and you have to 20-25 checks... it really a VERY hard task! Probably even some masters would have hard time to solve it quite fast!

    3. Lain is busy with four new exercises that address additional subtasks for M1-h:
      Find protector of the checking square
      Find the blocker of the check
      Find the pinner that prevents a check
      Find the piece that releases a new escape square while giving check

      That should address a lot of the issues you mention. If the training subsets contain enough positions, it is relative easy to make additional subsets with less material.

      @Lain: is there something you are struggling with? The subsets you showed looked well enough already.

    4. The FEN list is ready. I was AFK to "make" the UI for the problems (copy, change titles, load positions, etc.).

  2. What struck me the most? There are SO MANY problems with such "trivial" element as mate in 1! However we can see how many elements/components it is built on. The better we are at such "chess atoms", the better performance (and score that goes with it) we can show. It is like the excercise "make a check only with the Knight" - it should be practice until everyone reaches at leasr 60 CPM. I have score 75 CPM so far, but it looks like my system (method) is not good enough to break 80. And if we cannot solve such a chess atom excercise - we would be (precious) wasting time when the REALLY hard puzzle come! And such hard puzzle is "Find all checks (FAC)" and "Mate in 1 HARD". They are real killers to our lack of skills. Sad, but true. I fully agree with Aox - some microelements (I call these "chess atoms") have to be solved with a lightning speed - otherwise more complex puzzles (tasks) will kill our "super-slow and shredded" skills!

    FAC NO QUEENS 20 25 37 46

  3. Just to show you how hard work is needed: I have solved about 15K NEW (type of) puzzles to score quite solid performance (12K is the sum of the numbers counted/prooved to this moment at my posts - at the thread: Wednesday, February 24, 2016 - Salt mining in progress). After another 35-40K all of the puzzles I should obtain an optimal (maximum) performance. The biggest challenge is to use the best method (algoritm) to build a high quality skill at these puzzles/tools. IF Tempo writes another posts/articles - I will try to comment some important tricks I use to fasten the process. And some of these tricks help to achieve 60-70 clicks (mates) per minute - it means they are really solid one, aren't they?

    1. This blog used to have a subtitle "Chess improvement by effort. Achoo!!". The sneeze at the end symbolizing the allergic reaction to the word "effort". My strains are aiming at reaching the same results as you do, but without the same amount of effort, by developing new exercises that should lessen the time and energy needed. I want to prevent Robert Coble from jumping out of the window when he succumbs under the idea that he will be banished to the salt mines some day.

    2. Glass houses don't have windows: on a clear day, you can "see" forever. ;-)

  4. @Tempo
    reading your post i wonder if the "find all moves" of Lucas chess may help, did you ever test this?

  5. @Lain: I really love FAES-h. It addresses my aura vision. Good job!

    1. Nice! :D I'm glad that the tool works and can be used to make progress.

  6. I like the FAES-e more, they might be improvable. Is it possible to get a few thousand positions here?

    1. Tomasz has a high initial score at FAES-h. Without any previous training. Or more correct: with preliminary M1-e training. Suggesting that FAES-h and M1-h are closely related. Not so strange, if you realize that mating the king = cutting of all escape squares + check.

      FAES-h is ideally suited for the slow approach. That means, that you can imagine the aura of the white pieces forming the box around the black king every position. I'm going to give this slow approach a serious try, in order to find out if it works. I.e., that I become fast in it without ever training fast. I hope I'm disciplined enough and don't forget it and unintentionally speed up.

    2. I will upload more positions to FAES-e as soon as I can.

      But I would like to know before if you disagree with the following. Most of the positions have a lot of solution squares, and sometimes the positions are very similar, with exactly -or almost- the same solution.

      Here is an example of some random positions:

      r1bQkb1r/ppp2ppp/2n2n2/4p3/2B1P3/5N2/PPP2PPP/RNB1K2R b KQkq - 1 6 d8e1d1d2f1e2g1
      8/1R4pp/p2k1p2/P7/7r/2r2K2/7P/4R3 w - - 2 33 e2g2f2d6c6c5d5
      8/2k3p1/p1R2p2/P6p/6r1/8/7P/3R3K b - - 1 37 b7b8c6h1
      8/6p1/6R1/P1k2p1p/6r1/8/7P/7K w - - 3 42 h1c5b5b4d5d4c4
      8/6p1/P4R2/2k4p/6r1/5p2/7P/7K w - - 1 44 h1c5b5b4d5d4c4
      8/r5p1/8/5R2/8/8/5k1P/7K b - - 5 49 e2e3e1h1
      8/8/6p1/8/5k1P/5R1K/8/8 b - - 1 56 e4e5f3h3g2h2

      Should I have a filter criteria to remove "undesirable" positions? or they are fine?

    3. They look ok, but there should be some % where one king is in check

    4. I uploaded the new positions.

      It's a bit hard to control the check percentage, because there are a lot less check positions per game. I could remove non-check positions, but I don't like that idea.

  7. As an interesting Footnote
    In the foreward of Mauricio Flores Rios : Chess Structures , GM Axel Bachmann wrote:
    "In my carreer I have seen close to 100 000 chess games, including most of the grandmaster-level games played over the last decade. The cummulative experience from spending a minute or two on each of this games has allowed me to gain an exellent positional understanding. Staring at a position for a few seconds is often enough for me to see who is better, which plans will work, which pieces should be traded, etc."

    ( Bold by me )

    1. Meaning: that IF the salt mines work, we will have to develop salt mines for positional play too. "Identify all weak pawns" etcetera. ;(

    2. Yes,yes: IF
      But then the other tactics first
      I have already some exercises in the pipeline

    3. Great Comment. I think a case can be made for extensive game study and an understanding as to how to get to a positional advantage where tactical positions are created. The context of moves that went before is of value, imo.

  8. New exercises/to be tested:

    PD: I hate the weebly file manager, it has weird bugs.

    1. Great effort! Thanks! I put them on my sidebar. Tomorrow I will try them.