Sunday, November 20, 2011

It's not rocket science














.
.
.
.
After a long journey of reasoning everything seems to fall in the right place. Guess what, it is much simpler than I thought. Why didn't I see it before? It is the only logical way. Patterns, concepts, visualization, speed, it is all irrelevant.

I have been analysing about 100 problems rated 2000 - 2300 at Chess Tempo. For 90% of the problems I found that my pattern recognition, my knowledge of tactical concepts and my visualization were more than sufficient to solve the problems. By far.
Yet I made mistakes often. What is the nature of those mistakes?

Take the following problem from Chess Tempo (diagram below, rated 2216).
I suggest you first try to solve it at Chess Tempo here so that my ideas will come across better. Before reading further.


















.
.
.
Black to move.
There are only 4 moves in a row which you must make correctly. None of these are hard to find. None of them is illogical. None of them requires more than a mediocre ability to visualize. Yet it is easy to go astray. Since every move has reasonable looking alternatives. People at CT found 14 different suboptimal moves. Moves that change the outcome of the game to the worse.

So the real problem is to learn to choose between two or more reasonable looking moves. I apply the following trainingmethod now:

  • I work on a subset of problems at CT with a rating between 2000 and 2300.
  • I don't use a clock.
  • I try to find the solution for some time. For no other reason then to find the problems I can't solve.
  • I use a checklist with questions I ask myself about the position.
  • I look up the solution.
  • I formulate why I didn't find the correct solution myself.
  • I formulate a question I should ask myself in order to guide my thoughtprocess the next time.
  • I put the question on the list.
What I expect is that at first my list will grow. At a certain moment that should stabilize. Then I expect the process to become automated and I will no longer need a list.

10 comments:

  1. Me Munich: I didnt find the solution, too.
    The puzzle has a 2708 CT Blitz rating, with average solving time of 75 seconds. Usually only master players will get this puzzle served. Only around 10% of all masters manage to solve this puzzle.
    And it does not make me wonder.
    What is so hard about this puzzle?
    I is pattern recognition (recognising a concept) that actually hurts in finding the solution. This puzzle is an exception of what is knowledge.

    Those people who moved 1...Bd4 (which is wrong) had in mind probably a concept similar to this puzzle:
    http://chesstempo.com/chess-problems/49765

    Those who played 1...Qg3 saw that they are going to win a piece (which is only temporarily).

    Those (almost all of them I guess) who did see Qh6+ as a possible candidate move did not see the later attraction motive Bg3+, but only considered Rf8+ which does not win. "So it must be Qg3+... it is winning a piece".

    It is a mean tricky puzzle, that puts upside down everything a master has learned.

    However, an exception does not make a general rule.

    Tempo, your method of training - I believe it could work. The main draw back is, that you cant cover many puzzles.
    The difficult puzzles where you were able to find the solution - you wont learn much from them.

    If you lower the difficulty of the puzzles, then you will still find many puzzle you do wrong - ask there the questions which all can be summed up into one big question:
    Why didnt you find the solution?
    And in case of the easy range: Many much weaker players than me did find the solution and needed less time. So there must be a lack of knowledge I have, whereas others (weaker players) know more about the concept.

    If you dont like easy puzzles, I believe there is a different method:
    Look at a profile of a high rated player such as "roman oganisian" in Chess Tempo.
    Look at his puzzle history.
    Where did he solve a puzzle in 10 seconds, where the average solver needed 60 or more seconds?
    I looked at a few puzzles of this kind. I remember a 2-mover mate threat but silent move (= no check, no capture).
    How could roman oganisian find the solution so quick?
    He simply knew the concept/pattern, while others did not.
    Not all puzzles may have a concept that is easy to formulate. But many have. Looking at puzzles Roman did find very fast despite others didnt often show you examples of concepts, that are actually easy, but probably not known by us.
    Other possibility: Roman saw the puzzle before an remembered it.

    I believe the knowledge about concepts/pattern/typical tactics is individual. There are concepts I dont know that is known by many other chess players. I am especially interested in puzzles, that have easy concepts, are often to happen, and which I dont know well.
    Easy for others but not easy for me. Yesterday I covered 1400++ puzzles and found 17 puzzles I did wrong, and many more, that took me way longer than the average.
    So that makes maybe 50-60 puzzles that are worth to be reviewed (the rest (= 1350) are probably wasted time). And the good thing about them: many of them are very likely to happen in real games, and next time I dont miss them.

    But if you draw many conclusions from difficult puzzles - I believe it can work, too. I only have the feeling, that the time invested into 1 difficult puzzles you cant do is less good spend than the same time spend searching for easy puzzles you also cant do. You will simply find more to be learned. (at least I think so).

    ReplyDelete
  2. I could not find the solution to this one simply because I was blind to the backward-moving Qh6+. Still, while I was trying to see how it leads to a win, I made another error at Qd2+ Qe2 Rf8+, failing to see that Bf3 blocks mate and that instead Bg3+ distracts. I attribute almost all errors to errors in insufficient breadth - "I didn't see Qh6+ in my possibilities" - or insufficient depth - "I couldn't clearly calculate the nine half-moves from Qh6+ until Qxe2." Breadth is trained by thoroughness and perception enhancement. Depth is solved by flexing the calculation muscle. That being easily said, I was frustrated that often I spent 20 minutes on a problem trying to see it to the end and still felt inadequate when I failed at similar rates due to breadth or depth.

    ReplyDelete
  3. @Munich,

    What is so hard about this puzzle?
    I is pattern recognition (recognising a concept) that actually hurts in finding the solution. This puzzle is an exception of what is knowledge.


    90% of the 2000+ puzzles at CT have exactly this theme: choose between two or more reasonable alternatives of which only one is good. But isn't the same true about our games? How often do we say:"I thought about that but I rejected it because..."?

    Tempo, your method of training - I believe it could work. The main draw back is, that you cant cover many puzzles.

    That's why I do not take too much time to find the solution myself. If I can't find it within say 3 minutes, I know this problem is for me. I take more time for formulating "which question must I ask myself the next time when I see a similar position?".

    I played Rf8+ in stead of Qe2+ in reaction on a question on my list: When the follow up is not clear, can a new piece enter the scene? Now I added: Can you prevent that a hunted king flees away? It's all very logical actually.

    ReplyDelete
  4. @Soap,
    Depth is solved by flexing the calculation muscle.

    Nope. What arose was: if this position is going somewhere, then it must start with Qh6+. On g3 the Queen stands in the way of the bishop. For the second move: my Queen on h2+ is too expensive if a bishop can do the same.

    What I want too say is that when you choose the right move, the line is very straightforward. In fact you prune the rest. When the branch has no sidebranches, there is no problem at all to visualize the main branch. But you need a narrative to guide you.

    It is all about pruning, not about calculation muscle. It's about creating a convincing narrative which makes you feel sure which branches to prune.

    ReplyDelete
  5. We all know the main tactical motives. It doesn't make sense to improve along that line any further. We must learn to decide between two moves by means of logical reasoning. With hindsight you can't persist that this is a difficult position.

    It's not rocket science. But we reason in a chaotic way.

    ReplyDelete
  6. Me Munich: Do we know them all?
    I doubt that. I can find them all. At the same time my STM is very occupied with calculation.
    It would be good if I am not only able to solve them, but to "know" them. Like you can see a back rank check mate within 3 seconds.

    I am sure I discovered many new ideas when solving my tag themed sets. Also, if I watch Roman Oganisian, he is doing puzzles in about 10 seconds where the average guy needed a minute or so. I was looking through his solving history, and he is sometimes fast the first time he sees a puzzle, so he could not know the solution before (Unless he cheats and has a second user. But I doubt this.)

    Here some examples of similar concepts. I am sure I do not memorize these puzzles, but rather the concept of them.
    I believe it is no coincidence that they all have almost the same rating. They have the same concept (=idea/pattern). It makes sense.
    I admit I did not know this concept well, and this is why my solving time is not so good in this concept. But it will be for the future. It is worth looking at these puzzles, all very close to 1350 CT Blitz elo (all in my Pin-set):
    http://chesstempo.com/chess-problems/49543
    http://chesstempo.com/chess-problems/62163
    http://chesstempo.com/chess-problems/62935

    The last one took me only 11 seconds. I bet I will still improve in solving time if I see this kind of puzzle more often or if I repeat the set.

    I want to point to the hypothesis of aoxomoxoa, who formulated some points about learning, taking into consideration lots of science papers.
    Point 5, 8 & 9 speak against doing high rated puzzles.
    Also keep in mind point 4 when you do high rated puzzles. You need to go back and review your puzzles.

    Then again, aoxomoxoa does not need to be right. I beblieve his hypothesis is quite correct, though.

    ReplyDelete
  7. P.S. sorry, here is the link to aox hypothesis:
    http://aoxomoxoa-wondering.blogspot.com/2011/07/learning-improving-without-repetition.html

    My texts are a bit lenghty and repetitive. Sorry for that, but shorter texts would require more time for me to do... :-)

    ReplyDelete
  8. @Munich,
    We are familiar with the 27 main tactical themes. We are not going to gain anything on that road. That simply doesn't make sense.

    What we are not familiar with, are the topics "in between". What is causing us trouble? I insist that the problems with 2000 - 2300 are not difficult AT ALL. It is not rocket science. But we MUST focus on what we are doing wrong only.

    We are not talking about hundreds of mysterious ways of failing here. My list with questions is now 17 topics long. But I notice already some overlap and organisation. It will not become longer than 100, I'm sure.

    I found an even more effective way to study: I look at the faulthy moves other people make. That are usually the mistakes I forgot to make. Since it is in the same position I already am familiar with, that is very effective. At this tempo my list will be complete soon.

    Repetition is irrelevant here.
    Why worry about the level of the problems? It is about the 100 main types of mistakes you make. Fix that and forget the rest.

    ReplyDelete
  9. me Munich: If I look at the thinking process of masters, then their is evidence, that they rely heavily on their long term memory.
    If I look at chesstempo profiles who do a lot of blitz or standard puzzles (most of them doing "hard" mode), then I cant see any improvement. Especially on CT Blitz there seem to be no improvement (I prefer to look at Blitz rating, because in Standard the rating improves the longer you take your time).
    No improvement after about 2000 puzzles, in some case no improvement after about 3000 puzzles.
    O.k., they dont write questions down, and writing down questions is probably a good thing to do. It helps memorizing things better. After he failed a pretty tough puzzle at around 1100 CT Blitz rating (It was a "backward" move), my son Joey (6 years old) wanted me to write down "Next time I shall not forget to take that rook with the bishop". Well, of course this is naive, because this notice will only help him if he sees that particular puzzle again. Nevertheless it might have somehow helped him to look more for backward moves.
    And there we are probably again. It has to do with memorizing stuff.
    Masters take advantage of their LTM, my son puts puzzles into his LTM.

    Aoxomoxoa said in my diary (you can read it in Chesstempo), that speed is important. Because your short term memory is about 20 second long. If you think further, than the "beginning" gets forgotten. You have a scope of 20 seconds. You need to press all information into this 20 seconds. Or other scientist say, that you can hold about 7 items at once ("Magic 7").
    If you have more information, then you "zip" informations together so that are left with 7 zip files in your brain (STM). It is an advantage to have the information ready in quick time.

    Yes, somehow your questions will be repetitive, and there wont be more than 100 questions at the end. In that way you repeat over and over again (=memorize) these questions. Maybe you will find out, that you can zip some questions together. And at that time you will find out that it is CCT again. (Joking: But by that time, you will know very quickly, that it is CCT again).

    I am curious about the outcome. It could help to have the important questions ready in a game.
    On the other hand, others found them, too. You could skip your 100 question catalog and go straight to the three questions that zipped it allready together: CCT. After a while you will ask them subconsiously.

    Speed is important. You might hope that your 100 questions will speed up the process. But they wont. At best you decode difficult concepts into several easy concepts. Like someone who is learning to read and prefers to learn words such as desoxyrybonuclein acid.
    At the end it is a simple word, isnt it?
    Look - it is easy: "de" or "des", then "oxy" which is simply the air, "ry" well, this is quite special, we simply memorize it. "bo" like boat, or boot, or boy. "nu" is probably best memorized with "nuc" and "lein" is a more or less frequent ending in some other words.
    See? it is easy. You can decode it, and next time it will be easy to read "rhythm" as well as "ingenuity" or "Quiescense" or "hieroglyph(ic)"...

    ReplyDelete
  10. @Munich,

    glad to see your didn't lose your sense of humor:)

    Of course we are only considering CCT moves. By definition, otherwise there would be no tactical sequence. But CCT is a container which is full of garbage moves too. We must get rid of the garbage.

    What I am talking about is the choice between two reasonable looking CCT moves. The process of logical reasoning to make a choice. There is a logical hierarchy in those moves. But you have to be familiar with that process of reasoning.

    With the correct reasoning, you can prune. And after pruning, there is no STM overload anymore. Since you only consider the right moves and forget the rest.

    My list of questions is meant to help me to remember the right method of reasoning. Only until it has become second nature.

    Look at this vid where Rowson describes how he suffered from cognitive overload with only 4 pieces and how a process of looking what was really going on in the position saved him after training this.

    ReplyDelete