Saturday, March 05, 2005

About patterns and plateauing.

We talk a lot about pattern recognition.
But what is a pattern actually?
Scientists estimate that an average grandmaster has about 50,000 patterns in his memory. Below you see a graphic of the KNSB, the Dutch chessfederation.


rating (hor) vs #members (vert) click to enlarge

The grandmaster (rating 2600) is the one with the highest rating.
The average rating is 1680.
I don't think it's unrealistic to assume that the average player has the half amount of patterns as the grandmaster i.c. 25,000 patterns.

The most average players have gathered their patterns unnoticed.
You can not say "this is me and there I have stored my patterns".
The two cannot be separated.
Your behaviour (in chess) has changed because of your stored patterns.
You ARE the patterns.

So why is one plateauing?
It is not plausible that your memory is "full".
You don't even know how you stored it and didn't notice when you stored it.
You have no "full" feeling.
How could it be full?

When plateauing the intake of new patterns is distorted.
If talent would put a limit on the mind then it would not be possible to start to grow again after 3 years of plateauing, as I did. (of course talent does put limits on the mind, but not in this specific example)
I see guys play a tournament every second weekend. Still they plateau.
So playing much chess isn't very helpful.
Study in the classical way isn't very helpful. I tried it myself.

When plateauing you work on a sort of automatic pilot.
You play the same sort of games over and over again, and you get no new patterns from them.
Analysing games of grandmasters should give you new patterns.
But here lies a numerical problem.
As average player you have to gather another 25,000 patterns to become a grandmaster.
This is difficult near impossible by analysing games alone.
Besides that, no one is used to go over a game 7 times to imprint them in the brains.

For every 100 ratingpoints you need to etch about 2,800 patterns in your brains.
400 points in 400 days means 168 points in 168 days, which is a MDLM period for romping around with 1,000 problems.
So these 1,000 problems contain 4,700 patterns.
I'm inclined to think that a pattern is a variation. A single line.

It becomes clear now why it is so difficult to improve and why there are so few grandmasters.
Most trainingmethods work in principle, but they just handle too few patterns.
If you want to improve, you have to be prepared to do an awful lot of work.
Work that is easier done when one is young and the synapses are flexible to be programmed.
But at older age it is not impossible at all.

So there still is hope for us!
Apply grain of salt to your needs.

10 comments:

  1. I don't think it's unrealistic to assume that the average player has the half amount of patterns as the grandmaster i.c. 25,000 patterns.
    Why should the relationship between stored patterns and ratings be linear? that is a huge assumption. It also would be the ONLY area of human achievement where the relationship between mastery and mediocrity is linear. Other areas where the thing known and the mastery of the subject is measurable show very non-linear relationships.

    You ARE the patterns.I have no idea what you mean by this.

    So these 1,000 problems contain 4,700 patterns.
    I'm inclined to think that a pattern is a variation. A single line.
    If you dont' know what constitutes a pattern, how can you viably make so many assumptions about them?

    I personally doubt that a "pattern" is single variation. I expect that a "pattern" is a unified idea, such as how to force a smothered mate with queen and knight.

    I think the biggest problem is thinking hwoever along the lines of tactical problems. MDLM is great for tactical weaknesses. But a GM has mastery of a lot more than that. How many opening lines does a GM know compared to a 1600 player? Each of those is in some way a "pattern." When is it best to play for center expansion versus a minority attack in the QGD? Another pattern. When are knights better than bishops on an open board? Another pattern.

    Tactics are important for lower rated players because lower rated players are weak at tactics.

    Of course, lower rated players are weak at everything, and the number one symptom of those multiple weaknesses is the presence of tactically exploitable weaknesses in their position. Therefore, tactics are a great "quick fix" for lower rated players because we're playing people who are weak across the board. But if you want to reach into the ranks of mastery, tactics alone is utterly insufficient.

    ReplyDelete
  2. A friend of mine did his doctoral thesis on adult learning.

    Adult learning has nothing to do with synapses and how they've changed over time. It has to do with learning patterns. Adults learn differently than children because adults already have so much more information stored than children. For adults, the problem is retrieval, not comprehension. Using proper techniques for memory retrieval allows an adult to continue "growing" his knowledge base. I didn't pick up knowledge on a number of advanced scientific and mathematical topics until I was in my 40's, something most educators claim doesn't happen.

    ReplyDelete
  3. Rakshasas,

    Thanks for your comments. I'll try to give a reaction to your points.

    Is the relation between the stored patterns and ratings linear?Well, of course it's not.
    But the intake of patterns overtime is linear.
    If you look at the ratingprogression of Kasparov during his youth it comes near to a line.
    In average the ratingprogression will be a nice S-shaped curve I assume. I only talk about 3 points of this curve, bottom, middle and top. Any curve through these points will do.
    Do you think the middlepoint is higher or lower? Prove it and I'll aggree.

    I'm not a scientist nor am I interested in entertaining discussions.

    This all is rather unknown territory (at least to me). I'm inclined to a practical approach.
    So I'm trying to put a workhypothesis together and say that it is true. And the truth has always to be proven in practice: does it give results. Until someone give a better explanation of things the old statements stay true.

    you ARE the patternsThis is a manner of speaking.
    You can't be separated from your acquired skills.
    A man who has learned to cycle IS a cyclist. Even when he is sitting in a bathtub.

    you don't know what a pattern is, how can you make viable assumptionsAs I explained, these assumptions belongs to the workhypothesis and remain true until someone comes with better ones giving better results.
    A pattern is surely not an idea.
    The scientific research I have based my story on found that the patterns are processed in an area of the brains that is associated with visual processing. This processing works in the very first moments you get a visual stimulus by seeing a chessboard with pieces. It works unconscious.

    I think the biggest problem is thinking hwoever along the lines of tactical problems. MDLM is great for tactical weaknessesI commented on this in a post "tactics,tactics,tactics?" on sun febr. 13.

    I'm the International Knight of Pancakes.
    This is how I bake my Pancakes.
    This is how They shall be baken.

    ReplyDelete
  4. Jim,
    There is a difference between accumulating knowledge and acquiring skills. The former has to do with memory-retrieval problems, the latter not.
    No swimmer ever drowned because he didn't retrieve his swimming-knowledgebase in time.
    The synapses-story is based on a scientific documentary I saw on television. If you don't like it I drop it. It's not essential in my post.

    ReplyDelete
  5. Yep, a pattern can be something like achieving a Lucena's position, while another could be building a bridge in the very same rook ending. It could be knowing under what conditions a sacrifice of the exchange is necessary to save or even win the game.

    ReplyDelete
  6. But the intake of patterns overtime is linear. If you look at the ratingprogression of Kasparov during his youth it comes near to a line.


    Do you think the middlepoint is higher or lower? Prove it and I'll aggree.
    First, Kasparov came up in a system with very controlled pairings. If you read about how the Botvinnik school worked, students who showed promise studied with those who were better than them, but played those that would be tough competition but who they would be expected to beat.

    In other words, Kasparov's linear progression is explainable merely by the type of rated play he was allowed while in school.

    As far as what I think the middle point is ... I'd say that you can look at the difference between mediocrity and mastery in multiple other areas: music, mathematics, language . .. and you'll see highly non-linear relationships between knowledge points and mastery of the subject. Unless you have a reason to presume that chess is fundamentally different than these other areas (who exhibit remarkable similarity to one another) then you're hypothesis is unwarranted.

    ReplyDelete
  7. >Tactics are important for lower rated players because lower rated players are weak at tactics.

    But Rakshasas, we are lower rated players. So we are studying what every lower player should - tactics. Also as far as I can tell, nobody from the knights said that doing this will turn us to masters. Tactical masters maybe? => but a small does of improvement is what is hoped.

    ReplyDelete
  8. Nezha,

    Yes, and I have no problem with that. If it works, great. After all, most of us are under no illusion that we'll ever be master strength, we just want to be better enough that we get a little more enjoyment out of the game and experience a little less frustration with our own limitations. I'm perfectly ok with that.

    But to talk about GM mastery as if it's simply more of the same KIND of things ... is frankly misguided. A grade school students doesn't become a post-doctoral fellow in mathematics by just doing more addition problems. Understanding of the subject matter becomes both more detailed and nuanced, while at the same time gaining a higher level of general understanding.

    ReplyDelete
  9. pattern theory is over ppl, wake up, its template theory now

    ReplyDelete