Wednesday, December 28, 2011

Transfer


















.
.
.
.
AoxomoxoA said it beautifully well:

I did start at CT with standard tactics training = "focus on high rated problems. Just to minimize the effect of learning new patterns.". My standard rating did get up higher and higher... but my Fide estimate based on standard did go down and my Blitz rating did not change. The CT-Blitz-Rating is better correlated to the OTB-Elo then CT-Standard-Rating. My problem was: I was just getting slower and slower.

This is exactly the problem we have to solve. We know how to solve complex puzzles in the studyroom when we are given enough time. But it doesn't transfer to OTB play. It is exactly the problem of cc players who play well in cc but not OTB. This must be a problem of automation, of ability, of skill.

I have done lots and lots of difficult puzzles in the past with the same result as Aox. And I followed the same route to overcome it. But by focussing on patterns you are slaughtered by the law of diminishing returns, sooner or later. No, we have to dive deep in the problem as described above.

I'm taking  my time not so much for finding the solution of a puzzle but for interpreting it. Asking myself: how can I find this easier in the future, how must I interrogate a position to find the solution. In an attempt to find a better way than just probing randomly generated moves. And the method has showed itself: logical reasoning. The mind only needs a pair of crutches to help. That pair of crutches is the thoughtprocess I'm building. In the hope of course to be able to throw them away one day.

The thoughtprocess is build around purposes, goals, targets.
  • What must be the goal of this position?
  • What is the problem to reach this goal? 
  • How can I overcome this problem? 
  • Can I falsify the remedy?
These kind of questions should help you to overcome the habit of random probing. The habit of random probing is fuelled by pattern recognition. The questions of the thoughtprocess though are gearing the mind towards a goal. Thus pruning the tree of analysis from branches of moves that don't lead towards a goal. Until the tree becomes so bald that it becomes managable for the mind.

11 comments:

  1. The method of reasoning can be automated, too. Going over your guidance list over and over again and test it against the puzzles to solve should have a positive effect.
    But having patterns quickly ready to help calculating should help, too.
    The processes inside the brain of a GM is different of the process of an A-class player. A GM is heavily relying on his LTM.
    The LTM is his database. The trouble is - I dont know what a GM has stored there. What is he all the time comparing? Is it some sort of a guidance list which he goes through and compares it with the current game he is playing? Possibly yes. But not only.
    For sure he has lots of patterns in his database, too.
    I had this experience: I started remembering CT puzzles in my real games. I am pretty sure I used my LTM more than ever before in my games.

    So I dont know what GMs store in their "Database" (=LTM). But I know what I have stored there: I stored CT puzzles, and it helped me.

    I pretty much it behave a bit like learning a language. You start with knowing a lot of words. But if you have to do a conversation, then you cant actively use them. You want to say something, but by the time you put your words together, the conversation is already at a different point.

    Chess is in many ways like learning a language. But it is not exactly the same. How do I get my knowledge of CT puzzles more active? I dont know. I guess by repeating the puzzles over and over again?
    It could be, that discussing the tags in the CT puzzles is very helpful.
    Having a guidance list to check the puzzles is helpful. And memorizing the puzzle is helpful, too. Do all three, and maybe you manage to actively use your new gained knowledge later in your games?

    ReplyDelete
  2. I did write a questionnaire while doing the puzzles. I was facinated by getting deeper and deeper in positions understanding them better. At last >I< had no success with this method.
    It was big fun though, like reading a criminal story. I did the analysis first, then i was able to understand my faults right away after i saw the solution.
    I think i said it here some posts ago: Masters spots tactics like i do see a bright red spot at a white wall ( i was watching that at a game of a high master with a member of my club where the masted did give explanations to the audience, breathtaking!!). They do it, without any "thinking". The thought process might be better for Masters but they can do better chess then i without any thinking. For many many years research was thinking, that the thougth process of masters would not be better than the one of Class players at all ( i read that as: the difference is not decisiv ). By years of experience, a master can tell what type of moves is possible in a position. He can tell the main features of a position in a second, memorise the position in no time, and be able to reproduce the position. He knows what moves have to be considderd, if a tactic is possible and of what type it might be, without thinking. Masters show even changes in the visual cortex, they see differences of chess positions ( instead only differences in human faces ).
    Thes changes cant be done with logical thinking ( my believe ).
    see: http://video.google.com/videoplay?docid=-6378985927858479238

    My conclusion: its necessary to memorise "tons" of positions ( = be able to name the "best" move in no time ) to create the chunks wich helps to be aware of other position too and improve the memory access speed to chess positions, to be able to differ positions quick and so on.
    The right thoughtprocess might give me 100? elopoints but i need several hundreds more.
    I dont see how your type of traninig will give you the skills named in the video but without these there seems to be no way to mastership...

    ReplyDelete
  3. The guidance training could help. At the end it is a list in your LTM you check against the current position. It is the automation that does the trick.
    This automation might help to memorize puzzles better.

    I give you an example of memorizing words. Try to memorize these words:
    -Thumb
    -Shoe
    -thing
    -chair
    -hand
    -beauty
    -mine
    -cake
    -cat
    -religion

    Read it once or twice, and then look away and try to write it down in exactly the same order.
    Difficult? Maybe.

    Now take this guidance:
    - thumbs up (you show 1 thumb)
    - a pair of shoe = 2 shoes
    - all good things are 3
    - a table has 4 legs.
    - your hand has 5 fingers.
    - a beauty you want to have s...x with!
    - the 7 dwarfs of snow-white worked in a mine
    - who eight my cake? (eight sounds like "ate")
    - a cat has 9 lives
    - the 10 comandments of the holy bible is about religion.

    so the numbers from 1-10 not only gives you the right order, but actually also helps you to memorize these 10 words better.
    Why should that be much different if you have a guidance list you go through over and over. It is some sort of mnemotechnic to have a good guidance list in your head.
    Also, while thinking over the puzzle it makes it stay longer in your LTM.
    By explaining to yourself how and why the puzzle works, you retain it better in your LTM. With a normal quick puzzle you need to repeat it after 8 days. But difficult puzzles I discussed with tempo here, I think I will still know the solution immediatly. Even after 8 days have past. Probably I will still be able to solve them in a month.

    ReplyDelete
  4. I wanted to add something for the statistical relevance: 42% of all CT puzzles are check mate puzzles (this is a context for your guidance list!). To achieve a high rating in CT, it will help very much if you are good in check mate puzzles.

    ReplyDelete
  5. Yes,
    it becomes more and more clear that in order to solve the problem as explained in the post,we need to make two efforts:

    Add intelligence.
    Automate it.

    I'm busy with the first step now, to find a more intelligent thoughtprocess. Once found, we must find a way to automate it.

    Step two is difficult to imagine when step one hasn't cristallized, but we already know that repetition and speed probably will play a role.

    Step one and two are of quite a different nature, so it is not likely to follow this path by accident.

    What we usually call exercises of pattern recognition, is doing step two without step one. I use to call this crude "the automation of stupidity".

    For step two we probably have to discover patterns in logic reasoning. Like standard scenario's, for instance. Yes, I'm pretty sure we have to find patterns there. But not geometrical patterns.

    Maybe:
    step two, translate logic into patterns.
    step 3 memorize the patterns.

    May be.

    ReplyDelete
  6. Aox: how many thousands could be necessary? If I cut away every puzzle below a rating of 1200 CT Blitz rating, then there are "only" 30.000 puzzles left.

    Tempo: automating stupidy. It depends really at what range you are looking at.
    I looked at your easy tactic training and need to admit that I too feel your range is too easy.
    The trouble is, that if your success rate in solving them is to high (say 97% as a starting value), that means you added only 3% intelligence on it. This cant be effective.
    But doing an easy level where your success rate is 80%, and repeat the 20% fails, plus 20% below-average-very-slowly-solved puzzles ("yellow´s") will wast around 60% (because 60% you knew very well). But you get compensation:
    The 20% wrong solved puzzles, plus the 20% very slow solved puzzles are puzzles that are very rewarding if you learn them. Why? Because statistically, these kind of patterns happen to appear very often in your games.
    And after you wasted the initial 60% --> you then never come across them again. You only concentrate on the other 40%. You automate these 40%.

    ReplyDelete
  7. @Munich: If anyone can "solve" 30000 CT-Puzzles in ~10 sec each... ehhh, it has to be a Super-Super-GM (in Tactics)
    There is still a need/possibility for to learn non tactics positions: Opening ( 10-thousands positions anyway!!), Endgame, Positional, Strategic...

    ReplyDelete
  8. I wonder what is so difficult about memorizing 30.000 puzzles. Should be a possible task:

    10.000 of them I can do very fast --> no need to learn them.
    10.000 I can solve, but need to repeat them for a while till I memorize them. Maybe 7 repetitions?
    10.000 Puzzles I did wrong. I need 2-3 repetitions more of them. 10 repetitions?

    That makes:
    10.000
    70.000
    100.000
    -------
    180.000

    500 per day, and you should be getting there in about a year?
    Unrealistic goal?

    ReplyDelete
  9. @Munich
    "500 per day, and you should be getting there in about a year?
    Unrealistic goal?"

    500 a day is way to much, you get depressed ( at least ).
    7 repetitions is not enough, especially if they are complex. I think the dosage and speed of empirical rabbit is ok: ~ 100 a day and i would say ~15(?) repetitions (in the whole! life) ... A few years...I am working on it ;-)

    Munich life is not a sprint, life is maraton, better to start slow and then increase the speed then having breakdowns and heart-attacks

    http://www.kumon.de/Die-Anwendung.23.0.html
    Read there : Jeden Tag ein kleines Stück
    and
    Wiederholungen festigen das Gelernte

    ReplyDelete
  10. How long do you need to solve this?
    CT Blitz rated at 1371:
    http://chesstempo.com/chess-problems/21582


    Well, I guess it can happen every now and again, that you fall back. We are not always at our peak in our concentration.
    My history of solving times for the above puzzle:
    Repetition Nr. 2, 21.11.11 = 27 seconds
    Repetition Nr. 3, 26.11.11 = 10 seconds
    Repetition Nr. 4, 29.11.11 = 12 seconds
    Repetition Nr. 5, 04.12.11 = 18 seconds
    Repetition Nr. 6, 29.12.11 = 1:36!

    Today I was repeating it a 6th time.
    I needed 1:36 (in words "one minute and 36 seconds").

    The strange thing about the history is, that after my 3rd repetition, I needed longer and longer to solve it.
    I guess it is an exception.
    I continue doing this puzzle until I can do it in 12 seconds. My former goal had been to keep repeating a puzzle until I can do it in 9 seconds. If my former goal had been 12 seconds, then I would not have repeated it after my 3rd repetition.

    Due to this I am a bit wondering if I should really relax my 9 second goal to 12 seconds. The more relaxed my goal is, the less puzzles I need to repeat.
    If a youth can improve by not repeating a puzzle, then an adult should be able to improve if he sees each puzzle 3 times in row within a few days. Adults have a memory that is worse to a youth. But it isnt THAT worse, is it?

    ReplyDelete
  11. But aox, if I do just 100 a day, it would take "forever". I might not live long enough. I am afraid I dont have the patience for that.

    I need to see a result soon. (so far I am content with my progress).

    ReplyDelete