Wednesday, August 30, 2006

Finding matching socks in the dark




















What color do these socks have?


I have done a lot of research the past days about the working of the human (and rat-) memory. That convinced me of the following:
We amateurs must indeed fill our procedural memory with patterns. This has to be done by repetition. No matter how a grandmaster has done this in his youth.
Scientists admit that they have no idea why the results of the study of a grandmaster are comitted to LTM while that of amateurs are not. So we are on our own here.

I have reviewed all my old sessions at CTS during the past week. I use the "method J'adoube" for revision. In the mean time I did research on a hierarchy in patterns. I have not yet results that I can formulate, but it is a very interesting area.

I (never start with "I" for the third time, that's impolite) noticed a strange effect during the revision of problems. A problem at CTS starts with 6 seconds silence, then the computer does a move, then 3 seconds silence again and then my clock starts ticking. The first time I see a problem, I have trouble to see anything at all during the 9 seconds total silence before my clock starts ticking. Those seconds are gone with the blink of an eye. I certainly doesn't have the fast hawk eyes that Mousetrapper describes. I tend to start my investigation of a new position at the beginning, that is to say, by Adam and Eve. . .

When I repeat the problem for the 5th time however, I have time aplenty the first 6 seconds. I can finish my paper, eat a muffin, drink my coffee (yawn) before the computer makes its move. This different experience of the flow of time is just amazing.

Monday, August 28, 2006

Topsy turvy

Please all give a warm welcome to our newest Knight SamuraiPawn who chops his way thru CT-art! May he be able to find matching socks in the dark!












I've been two days down due to food poisoning. Still a little feverish, but I can hang behind my computer again at least.

It's nice to see how a discussion among the CTS-users is evolving. Sciurus asks if repetition is a neccessity for learning pattern recognition, since grandmasters in the past became grandmaster without computers and without such repetitions.
I like this kind of basic questions which can put everything topsy turvy. I have of course a certain amount of repetitions under the belt, but that's no reason for not asking these kind of questions. On the contrary.

Where does the idea of repetition come from? Cognitive rechearchers asked themselves how many patterns a grandmaster had stored in his LTM. Based on the amount of hours a typical expert needs to spend to become an expert in any area (=grandmaster in chess), which is about 10,000 hours, they made an educated guess that possibly 50,000 to 100,000 patterns are stored.
We concluded from this that repetition was necessary. How else would you store so much patterns in your LTM?
What we didn't realize, is that the technique of spaced repetition is developed for EXPLICIT memory and not for IMPLICIT memory. My experiments on CTS indicated that a much different regimen would be better for implicit memory.

Mousetrapper raised the question how many patterns there are at CTS.
It's funny that I'm doing research along the same lines as he. At first I thought there would be at least a few hundreds or thousands. Now I have evaluated an 200 CTS positions I'm inclined to say just 80:
  1. Trap
  2. Double attack
  3. Discovered attack
  4. Skewer
  5. Pin
  6. Promotion
  7. Invasion
  8. Overworked piece
  9. Eternal check
  10. Exchange/take/take back (lower rated problems)
Sometimes along with a preparational move:
  • Push
  • Deflection
  • Decoy
  • Annihilation of defence
  • Blockade
  • Interception
  • Clearance
  • Intermediate move
10 basic motifs X 8 preparational moves = 80 different patterns.
In every problem at CTS is at least one of these basic tactical motifs present.
So this can shed a different light on what we have to accomplish. If I see a problem at CTS I now have to ask myself the question: which one of the 10 basic motifs do I have in front of me with what kind of preparational move? I have to find the answer on this question within 6+3 seconds. That's the other way around!

So the job is how to recognize just 80 patterns in 23,803 different situations.

Actually Mousetrappers Target Feature Count and my Thoughts About Tactics were in a way unconscious attempts to find a systematic answer to this question. If we can find a systematic answer, it could alter our training efforts dramatically! It can make repetitions possibly superfluous indeed. Now I know what I'm after, I will have a new look at it. I'm curious where this will lead us!

Thursday, August 24, 2006

Oops, wrong memory.

From time to time, when I feel like a good laugh, I go to a chessforum and read about what they say about DLM and his system. For years on end they say the same things. Some think it would work miracles IF they would do it, without ever getting the impuls to actually start, others thinks it is worse than useless. What keeps surprising me is that they are content with an OPINION. A sort of opinion-fetishism. What's the use of an opinion? I never need one! Just try it and see for yourself.

Very funny were the posts of a person who claimed that it was impossible to follow the program because it consumes so much time. When I looked at his profile, he had more than 2300 entries posted the past few years!

What everybody seems to agree about, is that you have to have a "balanced approach" to chess. This is the single most heard critic against a heavy tactical study program. There's nobody who can deny a balanced approach.
But you have to see things in the right perspective.

If you know only a few hundred words of a foreign language, and you try to write a novell in it, you can worry about your style of course. But that's rather silly. You first have to adress the most unbalanced part, which is the lack of a good vocabulary of frequently used words.
The same holds true for chess.
But for some reason it seems to be impossible to see how bad you are in tactics. It is a strange blind spot that almost everybody has. Every grandmaster who gives a simul says that we are so bad in tactics. But we stay in blessed ignorance about it.
Not me.

Everyday at CTS I'm stunned.
The problems are SO incredibly simple, if you look objectively to it.
Everybody can solve the problems within minutes.
But WHY on earth does it take me so long?
AFTER I solved a problem I tend to slap my forehead (figuratively spoken) and say "aha! is it so simple?"
But the next problems will cause me trouble again and again.
As long as this is the case, it is simply silly to adress other parts of the game.
Yes, I have to take a balanced approach to chess. Since my play is so gruesome unbalanced, I have to adress the most unbalanced part of my game first: tactics, tactics, tactics.

When you work at CTS, you have to keep in mind what you are after.
As brainscans revealed, the difference between a grandmaster and an experienced amateur lies in the perception of the chessboard during the first few seconds a new position is presented. The grandmaster uses his UNCONSCIOUS working LTM while the amateur sees the position as new. The amateurs had at least 10 years experience. The researchers raised the question "why do grandmasters store their patterns in unconscious motoric memory during study while amateurs don't?"

Which means we are at untrotten territory.
We can predict that if we don't take measures, all our efforts will be in vain.
At our club there are a lot of members who study chess for over 40 years but whose ratinggraphs over the last decade can be represented by a straight horizontal line, slightly tilting downwards. . .
So it is easy to study chess in the wrong way. What is learned by the experienced amateur will be crystalized in the wrong memory. You can actually know for sure that if you follow common advice, you end up like those guys with 40 year experience.

What are the characteristics of the memories that are used by the grandmaster and the experienced amateur? What is their "taste"?

Grandmaster:
  • LTM
  • Procedural memory (knowledge how to do something)
  • Memory where complex motor skills are stored
  • Implicit knowledge. You can't formulate it in words.
  • Unconscious recall of information
  • works on a stimulus - response base.
  • It works with tremendous SPEED.
  • Storage by fast repetition.

Experienced amateur:
  • LTM
  • Declarative memory.
  • Explicit knowledge. It's possible to formulate it in words.
  • Conscious recall of knowledge.
  • It works very slow.
  • Storage by understanding (the formulating part) and spaced repetition.
Who have benefitted the most of the program?
J'adoube and Celtic Deatch.
The least?
King of the Spill, Mousetrapper, me.

What's the difference between us? It's rather difficult to try to characterize a person thru the small medium that internet is and based on what everybody tells about himself, but if I don't try it I have nothing at all.
So let's see.
What seems to be a common characteristic in the approach of J'adoube and Celtic, is that they both lay their emphasis on speed and #repetitions. If I recall well, Celtic is even the inventor of the "minicircles".
King, Mouse and I are all laying emphasis on understanding. We are all systembuilders. Look at King's famous Fundamental Checkmates, Mouse' Target Feature Count and my 'illustrious' Rake-Time continuum to understand what I mean.
The point with understanding is, is that it's just the beginning. I always thought I was finished when I understand something. But actually then the work (achoo) begins. To transform knowledge into skill. It's difficult to work when you think you are ready. . .

What got me thinking were the following figures:
To get 1 ratingpoint at CTS I have to commit 34 problems to memory, which takes me 570 problems to realize. I mean I have to solve 570 problems at CTS to commit only 34 problems to memory, so my measurments revealed. I asked myself if it wasn't possible to simply concentrate on those 34 problems and try to squeeze them into memory. 34 problems a day must be doable (=400 points in 400 days:)
So now I concentrate on those 34 and I repeat them alot.

At CTS there is a school which focus on succesrate. That is typical for someone who believes in understanding. But since that is only the first part of the job, they tend to forget the transformation into skill. A common mistake under experienced amateurs, as we know. . .:)

The first round at CTS I look upon as a selection of problems. I need all patterns I don't recognize in under 10 seconds. Because that are the problems where I still need understanding to solve them. I work rather fast during this first selection, so my succesrate drops steadily.
Then I copy the session to harddisk and work my "minicircles" untill I can do all problems while conversing with Margriet. Since it is useless to commit wrong patterns to memory, I use the first round after selection for understanding the problems. Since I use this method, my improvement at CTS is steady at about 1 point a day at average. Which is twice the increment I'm used to.

Sunday, August 20, 2006

60,000 down, 10,000 to go!!

60,000 done at CTS, 10,000 to go. If I should stop at 70,000, that is.
But since things are going very well at CTS, I want to continue.
I reached a new all time high of 1618.



Date
Rating avg
Rating max
Gain
#probs
#probs/pt
start
1470




july 21540156470
50,000
570
aug 3
1580161640
56,529
163
aug 20
1590161810
60,000
347


There is continuous growth.

Friday, August 18, 2006

Chug chug

I'm from an old school that doesn't allow me to have a big mouth when I have no results. I know, in these modern times that is rather old fashioned, but you can't learn an old dog new tricks. So that is why I blog so little lately.

But we can always do some crude calculations.
The table below is derived from a bellcurve at CTS giving the rating distribution of the problemset.
The first column is your rating.
The second column tells you how much problems you have to commit to your long term memory to gain another 50 ratingpoints.
The third column indicates the amount of problems you have to commit for 1 ratingpoint increase at CTS.


Rating

#probs to commit
#probs per rtg.point
1600170034
1650
150030
1700
150030
1750
120024
1800
100020
1850
800
16
1900
700
14
1950
500
10
2000
500
10
2050
300
6
2100
300
6
rest
200
4
total
10,200



As you see the #problems per rtg.point diminishes. That means that the effort to grow diminishes. On the other hand the reliability of the rating will diminish at the edge of the bell curve. Measurements in the past have lead to the hypothetical conclusion that there are 30-33 problems needed per rating point OTB.

Boy, I wished I had a reason for a big mouth.

Tuesday, August 15, 2006

In memoriam










Yesterday prof. Adriaan de Groot passed away at an age of 91.
His famous doctoral dissertation about pattern recognition in chess was the main inspiration why I started with the circles. May he rest in peace.

I'm working my ass off at CTS, but it is difficult to blog about that. There are enough other things to blog about, but I don't do that since it takes too much time. Chessclub is still on vacation.

Friday, August 04, 2006

Restarting endgames

Susan Polgar had a poll on her blog. She asked "Which part of your chess game needs the most work?" Much to my surprise I was the only one from fifteen who answered "endgames". There clearly is a mental barrier in people which makes them to neglect endgames. I stumbled for the umphteenth time over this barrier.

But today I restarted the study of endgames again. I use PCT, and I do only the endgame modules. I continue my work on CTS ofcourse, aiming at 200 problems a day.

CTS 56700 done
Avg rating 1580
Max rating 1616
PCT end mod-01