Monday, July 17, 2006

Playing around

I'm toying around with the issues from my previous two posts and the valuable comments on it. Maybe new idea's arise when formulating the same issues in other ways.
Let's try.

Say, we have a complex middlegame position. What does our grandmaster do with it?
In the experiment with brainscanning, the people played against a computer. The scan was taken during the first 5 seconds after the computer made a move. That stood not in the original article, but research on the web revealed that.

Our grandmasters Long Term Memory (LTM) is triggered and analogous positions from the past are consulted. This happens at a subconscious level. Most of the elements (both tactical and positional) of the position are recognized. As response to the stimuli (the new position) the LTM releases 2-3 potential moves. The GM starts to calculate, and one of the moves is chosen.

What happens in the mind of our experienced amateur (rating 1700 or higher and at least 10 years experience in tournament play) when confronted with the same position?
The LTM is almost not consulted at all. This sounds unbelievable, but the brainscans proof it. Furthermore, it is consistent with my own experience when I played a tempo (moving within 3 seconds) against the computer which was allowed to think only 1 ply deep. What you can find within 3 seconds is quite based on what is stored in your LTM. The quality of the emerging move within 3 seconds clearly indicates if there was something analogous stored in your LTM.
Our amateur sees the position as new and starts a process of trial and error. Heavily depending on the Short Term Memory (STM)
But this process can't emulate the responses of the grandmasters LTM.
Even a sheer unlimited amount of time is of no help. Nor is a thoughtprocess. A thoughtprocess at this place is a try to make the trial and error process more efficient. A more efficient trial and error process can only give some relative results against opponents of the same level who use their brain in the wrong way too.
From this trial and error process emerge 6-8 candidate moves. The amateur starts to calculate and one move is chosen. Scientific research revealed that the ability to calculate doesn't differ much between grandmasters and experienced amateurs. Which is consistent with my own experience. After 3 years of doing tactical exercises, my calculation ability has reached its top. I can't imagine there is something to gain for me along the road of calculation improvement anymore.

While the grandmaster actually SEES most elemental structures in the position with his minds eye, most of these elemental structures remain hidden in the blind spot of the amateur. You can only see with your minds eye what is stored in LTM, which was what I found out with my experiments on visualisation.

My all time high at CTS improved with 13 points to 1590!
My estimated average rating improved to 1550 (was 1470 when I started with CTS)


  1. Interesting post and a challenge to my own intentions. I won't say that a better thought process is useless, but I agree that its use is limited. An important distinction has to be made: Is the position equal and quiet, is it equal with tactics on both sides, or has it just shifted in favour of one player? My first tries of a thought process has been a candidate move search by the «checks, captures, threats» scheme. This is just what you call trial and error. But now I am convinced that this method is too inefficient and time-consuming. I think the first step must be to find the most rewarding target. And then the candidate moves to attack it. This will limit the number of candidate moves to two or three in most cases. In many cases there will be only one candidate move that does the task.

  2. having thousands of positions in your long-term memory bank is certainly and almost obviously better than short term memory even with unlimited time. congratulations on the new CTS rating! you're doing awesome there. i also agree with your comment on the thought process. if you have those thousands of positions stored in your head, a great thought process is less necessary. the only benefit i can see in a good thought process is to shore up parts of your game that are weak. for example, i often ignore my opponent's threats and moves and just focus on my own plan. a thought process might help with that problem. but i definitely agree with you in general.

    p.s.i also responded to your comment on my blog.

  3. Tempo,

    One of the biggest differences between patzer and GM is, as you know, board vision.

    If you test a GM by giving him a position for 3 seconds to look at and then ask him to recreate the position from memory, he will almost certainly do this.

    Try the same with the patzer and he will only be able to give bits and pieces.

    In both cases, it will be because the board vision of the GM is more highly trained than the patzer, but the patzer can greatly improve once he/she has a residual of patterns to draw from.

    The technique is simple - we use "board vision" in the Martial Arts - some people call it the thousand yard stare or "combat vision". Essentially you don't stare at your opponent but through them to a point past the opponent. This allows you to use the Moro Reflex - some call it the Startle Reflex - to react to your opponent. You then use your trained responses - stored in LTM - to defend or attack.

    In chess it is the same - using the same technique you can take in the whole position very quickly and then process it - but it requires individuals to train their mind and eyes to not look at pieces or squares, but the relationship between them all - a snapshot of the position.

    Once you can do this and you have a repository to draw from, then sifting through LTM becomes automatic and a trained reponse is equally automatic.

    Good job and congrats on CTS BTW.

  4. We all use some thought process, though I realize that in chess the term is usually reserved for an explicit, almost algorithmic, procedure for selecting a good move.

    However, more generally, for some the thought process is automatic and unconscious. I have found having an explicit thought process helpful. For example, I have been saved by the blundercheck step of my thought process more than once. Also, my opponents have helped me with their blunders quite often, which they could avoid by simply adding 'blundercheck' before moving.

    If I had more time in games, I would play much better. Most of my big errors are due to laziness and time concerns, so I don't bother thinking things through enough.

    Also, sometimes a tactic doesn't jump out at me, but when I scan the board specifically for forks (for instance) one will pop out at me. My attentional system, tuned to particular patterns, pulls out patterns I wouldn't see otherwise. It's like looking at a complicated picture that makes no sense, and then someone tells you there is a dog in it, and suddenly it pops out at you. Given more time, I'd go through all the major patterns looking for opportunities.

    If it was all about rapid, CTS-like pattern recognition, then we'd all play blitz. It's because we need to think that we don't play blitz. Since there are better and worse ways to think on the chessboard, why not make explicit and fine-tune what you are doing anyway?

    I can see one reason not to make it explicit. People who are very skilled at things often do worse when they consciously think about their performance (when hitting a backhand in tennis, don't think about your form, just hit the ball). However, getting to this point of expertise takes a ridiculous amount of training, and in the initial stages things are very explicit and conscious. This beginning stage is crucial, as after a lot of practice, the training leads to habits, and bad habits (e.g., not blunderchecking) are very hard to break. I think this is why teachers tell their students not to play much blitz. Long games cultivate better thinking habits.

  5. Mouse,
    I don't want to disencourage you because I think it is very valuable what you are trying to accomplish. But, you know, I have to go where reasoning leads me.

  6. Jim,
    there will come a time when we will speak about the Startle-Moro gambit.

  7. Blue,
    there are two stages in the process of making a move.
    I would like to call them diagnosis and remedy.
    For our grandmaster the diagnosis is based on pattern recognition, the remedy is based on calculation. In an ideal situation, the remedy would be based on pattern recognition too. But in reality this often will not be the case.

    In the diagnosis stage, the grandmaster doesn't need a thoughtprocesss. In the remedy stage, though, a thought process would come in handy. Of course, in our hypothetic ideal situation, we don't need a thought process in the remedy phase either, since calculation is replaced by pattern recognition and the solution jumps on you. But since that is not often the case, not even by grandmasters, a thought process is needed in the calculation-based remedy phase.

    The amateur though, hasn't a database for pattern recognition which helps him in the diagnosis phase. So he replaces this by trial and error. As said in the post a thought process AT THIS PLACE has some relative value.

    But of course it is better if you are able to replace the whole diagnose phase based on trial and error by a diagnose phase that is solely based on pattern recognition.
    In that specific case, on that specific place, you don't need a thoughtprocess anymore.

    But since we are all still in the patzer phase, a thoughtprocess can come in handy anyhow.

  8. Some questions:

    So chess is just a factor of memorization then?

    if this is true, lets say we have two hyphotetical grandmasters with the same exact replica of chessic LT memory, the end should end in a draw? everytime?

    And if more amount of patterns stored in lt memory equals better chess strenght, then one will improved in chess the longer one plays as one would "hypothetically" encounter/memorize more and more positions. But otherwise, what explains the drop in performance of 40 something chessplayers? Does this mean that at that age, LT memory begins to fade also?

    just curious questions...

  9. Very interesting discussion here! It brings me to more ideas about grandmasters, patzers and thought process. We all agree that grandmasters see patterns behind the pawns and pieces. Patzers just see pawns and pieces. A thought process is useful for patzers because it helps to see more patterns behind the pieces. GMs don't need a thought process for this purpose. But they still may need a thought process on a higher level, helping them to see the complex patterns behind the simple patterns.

  10. Hi Nezha,
    good to hear from you again! I hope all is well?

    I will try to answer your questions along the line of reasoning of this post.

    So chess is just a factor of memorization then?

    The diagnose phase (pattern recognition) is based on memorization, the remedy phase (calculation) is not.
    Memorization is not a very good word since that is generally associated with conscious intake and retrieval of knowledge.
    It is more like riding a bike. You don't say I memorized how to ride a bike. The knowledge how to ride a bike has become an integral part of your system.

    if this is true, lets say we have two hyphotetical grandmasters with the same exact replica of chessic LT memory, the end should end in a draw? everytime?

    They would agree on diagnosis, but they would differ in apprecation during the calculation phase. One gives a higher value to a positional move, the other is fond of attacking. What is in your LTM are just your tools. What you are going to make with these tools differs from person to person ofcourse.

    In practice there will be considerable differences between the toolboxes too.

    And if more amount of patterns stored in lt memory equals better chess strenght, then one will improved in chess the longer one plays as one would "hypothetically" encounter/memorize more and more positions. But otherwise, what explains the drop in performance of 40 something chessplayers? Does this mean that at that age, LT memory begins to fade also?

    This is a very important question. Because if a GM doesn't know how to maintain his database consciously, it degrades over time. During the youth the brain is so flexible that the intake of patterns requires no conscious influence. But when you have reached adulthood, it does.

    Part of the patterns are subject to change overtime. That is easiest to see with openings. But there are other area's too.

    I will try to give an example. Margriet has done her circles with TCT. At a certain moment during a game she made a brilliant combination. It was based on a cross-pin. When she showed it to older people (I don't think there is anybody in dutch tournament halls who hasn't seen it:) they were all very excited with oohhs and aaahhs. But younger people, who had done TCT too reacted very tepid "hmmm, just a standard crosspin".

    The building of a pattern database works along the lines of survival of the fittest. It constantly adapts. The patterns with the best results become stronger and survive. The ones with bad results or which are rarely used die away.
    During the youth this goes automatic, later on you have to maintain your database conscious. Otherwise your rating will plunge. I don't think there is a physiological reason for degression at 40+. Kortchnoi is the proof of that. If he notices during a tournament he fails at tactics, he flicks in half an hour tactics training every day the coming months. Thats the way to maintain things.