Sunday, December 18, 2011

What De La Maza forgot to tell us.


















.
.
.
.
I always had the feeling that there was something missing in what DLM told us about his method of chess improvement. Something he forgot to tell us. Not on purpose, but for the reason that he didn't realize that it was important. All the knights who aped his method over the years failed to get the same results as DLM did. Because of this little omission.

It took me 7 years of research to find out what that little something could be. Now I have found a reasoning that might explain everything. The pieces of the puzzle fall into place.

This should be the order of training:
  • Add intelligence.
  • Automate it.
With adding intelligence I mean that you must take your time to really grasp the concept behind a tactical puzzle. You must ask yourself: what is so special about this position that it works. You must ask yourself: how must I guide my mind so that I will see the solution the next time. This is by its very nature a slow process of conscious thinking.

With automation I mean you must make this concept so familiar that you can retrieve it without thinking. The recognition must become automatic. This is done by speeding up. There can be only one way to speed up, and that is by repetition. Since if something is new, you can't speed it up.

What most Knights Errant didn't do was taking the time to really grasp a puzzle. The focus was on patterns and speed, not on concepts and understanding. What we did was that we automated stupidity. (At which we became very fast, btw)

People are usually inclined to use only one of those two methods. CC players are focussing on understanding but forget to automate it, while others are focussed on speed but forget to add intelligence. I have tried both methods, but not at the same time and with the same material.

28 comments:

  1. But there is rule no 10 to keep in mind, too (s.th. that the bright knight overlooked. At least he didnt mention it.):
    rule no 10:
    "Problems already known, dont need to be repeated"

    That means, that if you can do a puzzle very fast after you have not seen it for quite some time, then you do not need to repeat it.
    How to find out if I can still do fast what I have learned? Well, from time to time I need to repeat everything. But most of the time I am only repeating puzzles I did slow or I did wrong (the "fails" are the puzzles that are most valuable).
    Richard from CT was so kind to give me an optional check box "use most recent time" when I filter for puzzles I did slow. Because it is important how fast I solved them the last time I have seen the puzzle.

    I am not sure if DLM repeated only the puzzles he did slow or still failed.

    The whole idea in

    -add intelligence
    -then automate it

    can be found in the chapter "deliberate practice".

    ReplyDelete
  2. @Munich,

    The whole idea in

    -add intelligence
    -then automate it

    can be found in the chapter "deliberate practice".

    And did you do it?

    ReplyDelete
  3. Munich, if Mazza would had known this, he'd mentioned it specifically to help his minions avoid this tremendous pitfall they fell into. We are talking about years worth of effort poured into his method by not so stupid people, to assert that he told them about this is not a bit harsh statement on their intelligence.

    z.

    ReplyDelete
  4. Before you can add intelligence you must have it.

    ReplyDelete
  5. The reason that De La Maza did improve that much is: De La maza did dramatic many serious games in these 2 years

    De La Mazas about his Method

    "In the 64 day circle spend no more than five minutes trying to find the first move
    and no more than an additional five minutes working out all of the variations. If
    you fail to solve the problem within the allotted time, simply read through the
    solution. When you are working through the initial problems in your 1000
    problem set, you should find the first move in much less than five minutes. If
    you are not, then the problems are too difficult. In my 64 day circle I did not use
    the ten minute time period with any frequency until the last four weeks when the
    number of problems per day was relatively small."

    http://www.masschess.org/Chess_Horizons/Articles/2001-01_Sample_400_Points_Part_1.pdf

    http://www.masschess.org/Chess_Horizons/Articles/2001-04_Sample_400_Points_Part_2.pdf

    As far as i see De La Maza did not add to much intelligence and no automatisation of intelligence at all

    ReplyDelete
  6. @Aox,
    The reason that De La Maza did improve that much is: De La maza did dramatic many serious games in these 2 years

    I have done that too. Others have done that too. Without effect.

    As far as i see De La Maza did not add to much intelligence and no automatisation of intelligence at all

    Yes, that is why the Knights Errant were put on the wrong foot.

    That he added intelligence is pure speculation. Maybe he was very intelligent and had a fast mind.

    It's my reasoning that leads to this speculation. It would fit all the known facts.

    Of course it is better to abandon speculation and to try to proof this hypothesis. Which is exactly what I'm doing.

    I train two hours a day and do some additional speed training when I get tired of "adding intelligence":)
    Overloaded piece. 4 down 21 to go.

    ReplyDelete
  7. Just to get the same understanding:
    "add intelligence" = "to learn s.th. I did not know (well) before"

    Or in other words: deliberate practice = to focus on the things I am not good at.

    So answering your question "did you do it?" --> Yes, I did do that. I focus on puzzles I fail. I go through a lot of easy puzzles, just to find out which of them I cant do, despite that they are easy for others. All those puzzles I could do (fast) - they dont need repetition.
    I consider puzzles that took me long as "not known well" too. That is why I repeat these slow solved "easy" puzzles, too.
    But I could agree, that it is not important to do easy puzzles (I have good reasons, though). The main point is finding puzzles I fail - and learn them.

    ReplyDelete
  8. @Munich,
    Adding intelligence = use your slow conscious intelligence to find out what the position is about, to discover the relationships, to define a strategy to solve the problem, maybe even to generalize the solution so that it can be applied in other positions as well.

    Automation = to convert the concepts of the solution into chuncks that can be remembered in <4 seconds. By means of speedtraining and repetition.

    So you do this. Did you get a statistically significant improvement after plateauing for some time?

    ReplyDelete
  9. If I am not mistaken, I read somewhere that learning chess equels hard work, some 10000 hours.

    So nothing to do with intelligence or talent, just working your butt off. :-)

    ReplyDelete
  10. I do not believe rule no 10: “Problems already known, don’t need to be repeated.”

    Just because you are fast on one occasion does not mean you will always be fast. Just because you are fast does not mean that you cannot be faster. If you just practice the non-obvious stuff you will train yourself not look for the obvious. The mental algorithm you want is to check out the obvious, and if that does not work, look for the non-obvious. The non-obvious has to be reasoned out, and practicing speed spotting does not help it very much. The obvious has to be spotted automatically, and speed spotting does help here. Rule number ten makes you spend your time on the training that is least productive!

    With regard to the main topic, sportsman say “practice does not make perfect - perfect practice makes perfect.”

    ReplyDelete
  11. Tempo: Yes! I gained around 200 elo in Chesscube within 2-3 months.
    Also my CT Blitz rating considerably, and it is currently at an ATH. Which isnt the end. More is very likely possible. But I dont want to do Blitz right now, because of the punishment-point-system for puzzles you have seen before. I can wait with that prove, but I believe I am currently maybe around 1900-1950?

    I will test it in January, when my internet conection is more reliable and faster. (I dont want to do blitz with the disadvantage of a very laggy internet connection.)

    Click on my name or take this link:
    http://chesstempo.com/chess-statistics/munich

    If you write me a personal forum message, I provide you with my login (name and password) for chesscube, so you can see my enormous success.

    For the easy puzzles I failed: I usually failed them after thinking 1 minute or longer. I did thoroughly thinking, but I never wrote anything down. Afterwards I could often clap an my forehead - similar like you can do after you see the solution of an difficult puzzle, which is actually not so difficult if you know the solution.
    Sometimes a simple 2-mover.
    For the slow-solved "easy" puzzles (which were not easy for me, but just for the average "others") - sometimes more thoroughly thinking (more than a minute), sometimes less intense thinking (solving the 2-mover in 20 seconds after the 5th time I have seen it = still slow).
    The barrier between not knowing and knowing is not so sharp, you know.
    But repeating puzzles I fail is the part which is valuable most (I think).

    ReplyDelete
  12. Books 'predator at the chessboard' is great for building intelligence wrt basic tactics. Best I have seen. I wish I had seen that instead of MDLM when I first started playing chess.
    http://www.chesstactics.org/

    ReplyDelete
  13. Tempo did ask Munich:
    "Did you get a statistically significant improvement after plateauing for some time?"

    Munich had a dramatic increase in CT rating after doing Easy-speed-tactics(EST), but the database is to small to be a prove. We need more Guinea pigs.
    I did leave my plateau in tactics to: http://aoxomoxoa-wondering.blogspot.com/2011/09/leaving-plateau-in-tactics.html
    Emirical Rabbit did some analysis to on this field.

    Munich, Emirical Rabbit and I did not work on the Thought Process.
    Smirnov like to talk about it and is talking about automatisation of the "right" thoughtprocess after ~6 weeks of training with it.


    @Emirical Rabbit
    I made a definition in my blog what "known" should mean: You can solve the problem ~ 5 times quicker that average solver and the problem is alread fixed in your LTM. The Thesis is THEN:

    Problems already known, dont need to be repeated ( at least for a "long" period of time ).

    ReplyDelete
  14. @Tempo
    I did skim De La Maza's book, I dont see any automatisation of intelligence ( thoughtprocesses ), at least not in tactics. His suggested thoughtprocess is interesting though.

    ReplyDelete
  15. I had an dramatic increase in strength with my training between 21st of April - 8th of July 2011. I did some training before 21st of April for approx. a month or so, but the method starting from April the 21st was new, and I consider this method (it is very much based on aox 10 point hypothesis) to be the training that finally helped. I have no doubt I improved because I could immediately "see" what was going on in my brain: during my games I remember easy puzzles from CT. Strange as it is it actually works the other way round, too: If I see new puzzles, I remember games I played in the past.
    Anyway, it is very clear to me, that I "learned" s.th. and put it into my LTM.
    I cant really say I learned "deep" understanding, because a puzzle which has a CT Blitz rating of just 1350 - it is usually plain simple. Not so much to explain about. On the other hand: If there is no intelligence needed - why did I previously fail them? So this tiny little bit of new information - I learned it. I would say this is adding intelligence, even if it is just a tiny little piece of new information. I often had a "aha"-experience. And an experience like "I am so stupid - it is so simple!". Or the feeling: "Actually I think I could have played this tactic many times before - if I had just known about it!"

    However, I need also to mention s.th., and it could be important, too: "What Munich forgot to tell":
    I tagged many CT puzzles. I tagged a lot puzzles with "fork", and I argued a lot with people who tagged puzzles as a fork while I consider them not to be a fork but to be discoveries.
    I invested enormous time in this arguing with other people, why a puzzle is a fork/discovery/capturing defender/x-ray/overloading/distraction/skewer.
    So I was really getting my thoughts deep around the tactical nature of the motive.
    I am not sure if actually this is the "trick" or if it was just wasted time to tag hundreds of puzzles.
    It might not be important. On the other hand, if you all talk about "adding intelligence" - well, how to call it, if it is not "adding intelligence"?

    There is more to mention:
    Despite my great success in the period 21st of April - 8th of July, I believe my whole training in November (plus a few days in (October & December) did not help me. At least not much. I cant feel any strength improvement. I cant see I play better.
    Why?
    I believe I did not improve, because I was only repeating what I covered in the period 21st of April - 8th of July.
    I improved in speed, but that should be natural, since I remembered a lot of these puzzles.
    I hardly learned anything new in November. Nevertheless I hope the November training was not wasted time. I just made sure I still keep my new gained knowledge. I repeated these puzzles, and now my new knowledge (gained in April - July) stays with me (in my Long term memory).

    I conclude for myself: beeing able to do puzzles even faster than fast does not make you (much) stronger. (I am talking about somthing like 9 seconds versus 12 seconds).
    Beeing able to do them veeery fast is rather a side effect: Of course one will do them faster, if he repeats his puzzle-sets he already knew before. Nevertheless, repetition is important to fight "forgetting".

    I did around 26K puzzles between 21st April - 8th of July. And I did another 18K in November, but mainly repeating the stuff of April-July.

    Two days ago I also tested my board vision training. Result: I can still find a lot attacks, defenses and checks. I did not need many attempts to come close to my ATHs (Attack training ATH = 47, defense training ATH = 42, check training = 35 ATH per minute).

    ReplyDelete
  16. @CT,
    ACHOO. . .!!
    I'm stil allergic for the words "hard work"

    ReplyDelete
  17. @BDK,
    That's a nice site indeed! I will look trough it to see if I have forgotten important things.

    Given the different nature of building chess intelligence and automating it it's not so strange that DLM missed the point.

    If you don't know exactly what you are after, you will only train in the right way by accident during a short time.

    Usually that is when you start in a new area. Then you have to think first.

    ReplyDelete
  18. @Aox,
    I did skim De La Maza's book, I dont see any automatisation of intelligence ( thoughtprocesses ), at least not in tactics.

    Yes, hence the title of this post.

    ReplyDelete
  19. @Munich,
    I had an dramatic increase in strength with my training between 21st of April - 8th of July 2011

    I think I know exactly what you mean. I always felt that I made the real progress in a period of only 6 weeks. It was when the material was new to me.

    I looked it up, it was in 2001, already. I talk about adding 170 ratingpoints (official Dutch rating). There was no real repetition, although I worked with themed problems (chess tutor and Polgar's brick).

    I felt I improved those 170 points in only 6 weeks. Due to the lagging of playing games and the slow ratingsystem it took 2 years to show in my rating.

    I posted about it, but nobody seemed to recognize it.

    I have been looking after the same trainingeffect for 10 years (as I realize now with upset). To no avail.

    Until lately. With building narratives from complex looking easy problems I have that feeling again. I'm learning something new. It comes with a lot of "aha's" and head slappings. That's why I wrote about renewed joy in tactics again.

    In 2001 it was a coincidence I couldn't repeat. But now I know exactly what I am doing. Let's hope it translates to my play.

    ReplyDelete
  20. Thanks to this discussion here I will adjust my future training like this: instead of repeating all puzzles I could not do within 9 seconds, I will raise this minimum limit and will now allow more time.
    The reason is, that beeing able to do puzzles even "faster than fast" did not help me.

    It is a trade off: Do I invest time to learn s.th. knew, or do I re-learn the stuff I already knew acceptable well enough ("acceptable well" is NOT "very well"). I think it is "acceptable" to solve a very easy puzzle twice as fast as the average solver (who has an "easy" rating). Since there is no option for a relative time limit, I set the new time limit to 12 seconds. That is: if I feel very diligent.
    If I feel lazy, then I put the time limit to 16 seconds. Today I feel lazy and hence I continue to do my "trapped piece" training with a 16 seconds limit.

    ReplyDelete
  21. "I have been looking after the same trainingeffect for 10 years (as I realize now with upset). To no avail."

    Tactics is not about speed but paternrecognision. With other words learn the elements of what makes the tactic work instead of doing repetition after repetition with greater speed.

    ReplyDelete
  22. @CT
    Tactic is about speed too. Mental Speed is a FAKTOR in WM. Double the speed and your WM-Quality doubles too.
    You can only think what fits in your WM.

    Gaining Speed in repeated Puzzles is a sign of having it in the LTM (you have it in your Memory ). Masters use their LTM to be better than Amateurs. You need to put things in your LTM to be able to use your LTM to get better. ( you can only get something out if you have put something in before ) With High Speed repeated Tactic Training you create the "Chunks" wich enables you to extend your WM-Capacity.
    Pattern RECOGNITION is about recognition (= a form of remembering = there is a need for the pattern to be implemented in the Memory AND to be remembered easy =(!)quick) and not about reasoning ( wich is of some importance too of couse).
    Amateure have to calculate, Masters remember (and calculate)

    ReplyDelete
  23. There is nothing wrong with aox arguments.
    However, solving a puzzle within 9 seconds is somehow like a confirmation for the fact, that you know this puzzle well: A solving time of 9 seconds tells you "actually you already knew the puzzle well, so you just did it in vain".

    But can I know BEFORE I solve a puzzle, that I am going to solve it within 9 seconds?! Probably not. And actually I dont even know if 9 seconds is actually still too slow. I mean, if you really "know" a puzzle inside out, you should be able to solve it even faster than 9 seconds (if it is a 2 or 3-mover, like most of them).

    Still I want to save time and concentrate on what I dont know well.
    I think I will handle it like this:
    The CT Blitz range of 1150-1300 I will be content with if I can do them within 16 seconds. Or drop this ultra easy range all together (because a more difficult range will include the ultra easy patterns anyway). Or I will do them just once and then never again (exceptions: failures I will repeat.). Doing them just once in order to get familiar to typical patterns.

    For the CT Blitz range of 1300-1475 I will try to do them within 12 seconds, because I believe that I solved a puzzle within 12 seconds, next time I would reach the 9 seconds goal.

    In that way I save training time for new stuff to learn while at the same time have the puzzles alive enough in my LTM to aid my learning of new stuff.

    And while I learn new puzzles, probably similar patterns of the old puzzles come along, so that I keep them at the end into my LTM as well as if I had trained them till I reach 9 (or less) seconds.

    And for tempos statement "it aint rocket science" - well, I disagree with that. It is somehow rocket science.
    If it was easy to improve, than we all should easily improve with a more or less good training.

    ReplyDelete
  24. @Munich,
    We have to work this out a bit further. New seems to be more important than speed. Or repetition.

    The fact that we can actually feel that we are improving should be our handhold.

    We can learn to drive a car within just 50 hours or so. This means: we can learn new things fast. And we automate things, well, automatically. With little speed and little repetition. Only after we have got our drivers license, we consolidate what we have learned by repetition. By practice. There I see the role for speed and repetition: for consolidation.

    I'm pretty sure that I shift gears like 30 years ago. But if today a racing driver sits near me and tells me how to shift gears, I would improve immediately, I'm sure.

    ReplyDelete
  25. Tempo saidWe have to work this out a bit further. New seems to be more important than speed. Or repetition


    Speed is a sign of "unknown"= not memorised. As more speed as more memorised. ( and as better the access to this memory ). In (mathematical) optimization there is the method of steepest decent ( http://books.google.de/books?id=R2tRiAbct6QC&pg=PA75&dq=nonlinear+optimization++steepest+descent&hl=de&sa=X&ei=rVP0TqmBOI_Nswbg0JH8Dw&ved=0CD0Q6AEwAQ#v=onepage&q=nonlinear%20optimization%20%20steepest%20descent&f=false )


    "To focus the training at new unknown problems" is somehow the same. Usually you do this method in Math with "perfect line search" wich would mean here: you learn this new pattern till perfection=max speed.


    I think its a waste of (learning) time if you dont burn the pattern in your LTM for "ever". If not: you will forget it after a while.
    My method: i do a speed training on my tactical blunders ( and...) .

    ReplyDelete
  26. Excellent advice! I might humbly suggest that de la Maza left out one more thing: why he quit after achieving his goal. My best guess is that he burned out and decided to get a life. I remember reading the essence of Rapid Chess Improvement in article form some years ago. I have toyed with his ideas off and on since, but in general found that his wholesale focus on tactics was both unsound and insufficiently interesting. When he wrote that article, my rating was in the mid-1400 and I was in my 40s. Now, my rating is 1933 and I'm a tad over 50. I like my methods.

    ReplyDelete
  27. I think DLM did forget sevaryl things to tell us, see the last comments here:
    http://empiricalrabbit.blogspot.de/2012/04/michael-de-la-maza-statistics.html#comment-form

    ReplyDelete