Tuesday, September 05, 2006

Time out

We have found a lot of interesting stuff lately. Especially the analogy with pattern recognition and clouds is very strong (see the comment on my previous post). On the other hand my fellow bloggers Mouse, Takchess and Sciurus remind me that there are still some loose ends to cover.

What have we found thus far?
The mind is very economical. From objects only a few remarkable outlines are stored in memory. Most of the picture isn't stored at all.
When you need the picture because you want to think about it in your minds eye, the outlines are recovered from memory and the details are reconstructed on the fly.

You can easely see this for yourself.
If you try to visualise your attic (or garage, or shed or whatever) for your minds eye, what you see is very incomplete. It is not as if you are in your attic and can easely look around. You see only a few outlines, and when you try to focus on details, you have to fantasize them, to reconstruct them. But most details remain simply in the mist. They are not there since you haven't stored them in your memory. (Or if you have, you can't reach them.)

The same is true for a chessboard. You can easely see for your minds eye a rook going from d1 to a6. That is because we have always worked with tables with columns and rows. But if you try to see a bishop going from c1 to a7 in your minds eye, it is very difficult to focus on the crosspoint. That is because we are only familiar with the long diagonals and not the remaining diagonals.
Isn't that funny?
Allthough we have spend days and weeks on end behind a chessboard, we have only so little of its geometry stored into our memory.

A pattern you haven't stored in your memory you can't recognize and you can't visualize it.
Once you have a pattern stored in your memory, you can recognize it even in a very distorted position, and you can visualize it.

This is the reason why grandmasters are always good blind players (and good blitz players), without special training.

Until now I always used the hypothesis that 1 problem is 1 pattern.
Science did an educated guess -solely based on the amount of hours a grandmaster spends to become a grandmaster- that a grandmaster has stored 50,000 - 100,000 patterns in his memory. This gave me no reason to cast any doubts on the correctness of the hypothesis of 1 problem = 1 pattern.
This hypothesis lead to my maniacal behaviour at CTS, since I wanted to store these 23,508 patterns of CTS as quick as possible.

My fellow bloggers Mouse, Takchess and Sciurus casted doubts on this theory in their respective comments and posts.
Sciurus drew our attention to the fact that grandmasters in the past had become grandmaster without computers and without massive repetition.
I dismissed that fact with the argument that they were young and we are adults so what they did has no relevance for us.
Which is a complicated way to say I have no idea how they did it.
But thinking things over, I realized that MDLM didn't make use of massive repetitions either and he was an adult. I hope you don't mind that I call 1209 x 7 NOT massive repetition of patterns.

MDLM used, according to my hypothesis, only 1209 patterns. Why did he become so good so fast? I always had the feeling he forgot something important to tell us. He did something what was natural for him -but not for us- and he didn't realize that it was important to mention it. I often wondered what that could be.

The scientific estimation of 50,000-100,000 patterns stored by a grandmaster has put me on the wrong foot. Let's have a closer look.

White: 2 pieces, a rook and a bishop
Black: 2 targets, a king and a queen
How many ways are there for a discovered attack for white to attack both the black king and queen: 65,280
How many patterns do you have to store to recognize this patterns a tempo in any position?
Somewhere between 1 and 65,280.
Can it be 65,280? If that was the case then probably no one would ever recognize this pattern.
Can it be only 1? I suggested this in a previous post. Most of us feel that 1 is to little. If I look at this specific discovered attack, I think that 4 or 5 positions are basic positions while the rest are lookalikes. So you have to store only 4-5 patterns and you can recognize all 65,280 positions with it. That's the power of pattern recognition.

We have to take a fresh look at it.
People have an amazing ability to recognize a once stored pattern. Even if a cloud is very distorted, you can see a rabbit in it. What you can't do however, is to recognize a pattern that you haven't stored.
How simple is it to store a pattern? We allready found that the brains are very economically with the storage of patterns. Even if you have looked for weeks on end at a chessboard during your games, most geometric details aren't stored in our memory. Why not? We have an idea of the global shape of our country on a map. But how many details can we draw of the various states, provinces, cantons or regions?
When we find out how to store patterns, and we identify the relevant basic patterns, we might be able to make a quantum leap forward.

Yesterday I said I don't have the patterns of Oklahoma or Switzerland stored in my system. I have looked at the map of both. Today I have a clear picture in mind of Oklahoma, but the picture of Switzerland remains pretty vague. The reason for this is that Jim said that Oklahoma looked like a frying pan on an open fire, while Switzerland hasn't such characteristic features. The map of Oklahoma has a lot of straight lines, what is a familiar pattern, while Switzerland consist of ravels, and I don't know how to remember these.

So it might very well be that there are only a few hundreds patterns to master, but that we have no idea how to do it.
Research continued.


  1. Two remarks on your interesting post.

    First: Switzerland looks like a little fat pig: Hind leg south-west (Geneva), tail north-west (Basle), belly south (Valais), foreleg south (Ticino), head east (Grisons).

    Second: Just storing patterns in memory can be very misleading. In fact I made many CTS errors by a (too) fast reaction to a stored pattern. Besides pattern recognition you must also have a powerful brain tool to filter out the right patterns and to reject the delusions.

  2. Quote: "MDLM used, according to my hypothesis, only 1209 patterns. Why did he become so good so fast? I always had the feeling he forgot something important to tell us. He did something what was natural for him -but not for us- and he didn't realize that it was important to mention it. I often wondered what that could be".

    To find a universal formula of chess improvement which works for everyone with the same success is, I think, a pipedream. It's a well known fact that different people maximise their intake of data with different methods. Some people remember more easily what they read while some remember better what the see in forms of animated images etc.

    To make things even more complicated we also have personality traites that effect our over OTB play. Some people might not have enough faith in their skill with tactics when they play, some might not be able to handle the pressure of tight competition...and the list goes on.
    In this case we have to supplement the MDLM plan with training that might not include chess at all, but more with our personality.

    We should also remember that MDLM played a lot of tournaments and competitions to gain experience and a skyrocketering rating.

  3. one other thing. it actually looks like the brain stores thing in exhaustive detail instead of simple guidelines. it has even been hypothetisized it may store everything it experiences, a 'normal' person just can't access it.

    there are a lot of examples of people who actually can access that stored data, practically flawlessly after a single exposure, and even a after a long time. usually such persons have some kind of more or less harmful condition from birth (like autism, asperger), sometimes caused by some kind of a traumatioc or disease induced brain damage. sometimes not, like with the synesthetes. my ex-girlfriend for example, can remember dozens of pages of text after a single reading. I can't usually even remember a single quote correctly after just one reading, and I do have quite good memory.

    so, the data most likely IS there, we just have to train our brain to access it. that's what I think repetitive training is about (in one aspect). we have a filter built in, which stops us from accessing all the data the brain accumulates, thus making normal life possible. -imagine if that filter didn't work, and we'd experience all incoming information as equally important. that would be like watching literally thousands of tv-channels simultaneously, and only 6-7 of them actually had relevant information to our situation. like standing under a deafening niagara falls of information. and there actually are people like this. ones who can recall every single detail of every visual image they've ever seen or song they've ever heard. no wonder they have problems communicating with the 'outside' world.

    the brain is a self-organizing neural net. because of its plasticity, we are able to actually change the way it works. even after after a massive brain damage, the surviving tissue/net can re-organize itself to perform actions it was never 'meant' to perform. it just takes a lot of time and repetition (I have a friend with a 40% brain damage, who now functions perfectly normally after 15 years). learning chess is probably not exlusively about rewireing your brain, as there are a lot of other ways it can adapt to/learn new behaviour, but I do believe learning tactics, or more precisely 'chess vision', is in large part just that.

    of course all this is mostly my own crackpot theory, but the facts I've mentioned are all real. :)

  4. Worm,
    the question is, WHERE do you have to hit yourself on the head and HOW HARD to inflict the right brain damage to become a good chessplayer:)

  5. Mouse,
    thanks for the pig. That might be a clue for my problem. With these simple comparisons, it is easy to remember those complicated forms like the map of Switzerland. But how did I learn to recognize a pig in the first place? Or a frying pan, for that matter?

  6. Around 12 years ago, I was very interested in the area of Mnemonics, I read a number of books on it and practiced some of the systems especially those found in the Harry Lorayne books. In most cases what you are attempting to remember is "Linked" to something you know very well. An example is the Loci system which was used by the ancient Greeks in remembering speechs . Each point of the speech was linked to an area in a location one was very familar with like the rooms of a house. So the first thing one wanted to remember was associated through visual imagery to the first thing you came to as you walked in the front room such as a door. The second item was linked to the next thing in the room a chair. Often times this Links were strengthened in Loraynes books by making the image ridiculous, adding action to the visualization, substituting the item to be remembered for the chair etc. This type of system would be good to remember a list of general rules or a checklist of points regarding chess . However I am at a loss how it would apply to remember a large amount of visual patterns as in mdlm training

  7. tempo, I'm experimenting with the hitting and how hard. no brain damage as of yet, but if you see me suddenly jump on CTS, you'll know I finally hit the spot. :)

    tak, I don't think the different memorization techniques are fast enough or give enough 'storage' to be of any use with chess. they're good for remembering fairly long lists, but they're awfully slow and take a lot of processing when you make that list.

    I've been toying with the idea of re-training your brain to work more synesthetically, as it's the way we processed information as infants. most lose that ability growing up, but not all. - what I was thinking, was to start with 'feeling out' letters, trying to 'see' if I had a 'visual' for each alphabet. then, repetitively train that connection, until it comes firmly established. then I'd move to small words, then bigger words, all the time trying to visualize the words as compound visuals of the alphabet-visuals.

    it would be a gigantic project, much like training tactics, and there would be no guarantees of it having any effect. also, I'd need to devise a test to measure if my memory had get any better (more accurate, faster memorisation, bigger storage), and test myself with regular intervals. I think it might work, but it would take a Lot of work and time, so I'm not that keen to start the experiment. :)

    it probably wouldn't help with chess either, but it would be interesting.

    I think I'll just go back to hitting my head with a hammer. *KLONK!*

    (btw, I do experience improvement in visual recollection of letters and numbers, which I credit to chess/tactics. anybody else noticed similar results?)