The hypothesis.
The hypothesis I'm working on right now can be summarized as:
After you reach a certain level, improvement in chess can be achieved mainly by pattern recognition.
What does that say, a certain level?
That's the plateau where people are stalling after initial growth.
When you haven't reached that level yet, probably anything will cause you to improve.
A list of facts that need clarification.
There are a lot of facts that need clarification.
A lot of these questions seem to point in the direction of this hypothesis.
That's why this hypothesis has evolved in the first place.
I have tried to geather a list of these facts/questions.
On a lot of these items we have posted before. So sorry for doubles
The former French blitz champion.
I had the honour to play a lot of games with this guy. He had drunk about a whole bottle of red wine.
During the games he was talking to another guy. So he paid hardly any attention to the game.
But he beated me game after game. What is it that makes this guy so good? It is not calculation, because he hardly looked at the board. It's no positional plan, because he hasn't taken time to work something out.
Beaten by kids.
I'm often beaten by "runner up" kids.
I often talk to them afterwards to find out how they are "phsychologically build."
Most of the time it are ordinary kids which you can easely beat on reasoning, logic, planning, knowledge, memory etc.. That always gives me the feeling that these boys perform a simple trick. A simple trick that is mystified by others as a chess prodigy.
The speed of grandmasters.
The 2 seconds per move average of Susan Polgar during her simul is simply beyond imagination.
How is that possible?
The only explanation I can find is described in this old post.
LPDO.
Grandmaster John Nunn describes that he once played 100 blitzgames with a master.
5 minutes on the clock for Nunn and 10 minutes for the master.
He won with 88-12.
Afterwards the master was disappointed:
"I thought that I would see lots of advanced strategic concepts in these games but actually all I have learnt is LPDO."
"LPDO?" Nunn asked.
"Loose Pieces Drop Off"
Most of the games where decided by relatively simple tactics involving undefended pieces.
Brainscans.
Brainscans of grandmasters and amateurs showed that grandmasters use mainly their long term memory while amateurs have to invent every move.
Papa Polgar.
Papa Polgar proved that any child can be a prodigy with the right training.
How about adults?
Candidate moves.
Research of Prof. Adriaan de Groot cs. showed that grandmasters investigate 2-3 candidate moves per move while amateurs look at 6-8 candidates per move.
He found that both the capability of calculation (!) and the memory of the grandmasters vs. amateur differed very little.
Indication.
To me all these facts seem to point at one single item: Ferociously fast pattern recognition.
Saturday, December 31, 2005
Thursday, December 29, 2005
More about flawed methods
The past 8 years I tried a lot of methods to help me to improve.
The main flaw of these methods was that my rating wasn't influenced a bit by any of them.
Except for the tactical training which helped me gain 200 points.
I strongly reject the idea that that improvement was the cumulative effect of all the training before. I'm sure that is not the case. In the years before I started with tactical training, I gained simply no point at all.
Beside the main flaw, gaining no rating, a few remarks can be made by every method.
This maybe helps to infer why the methods didn't work
To study openings.
At my level the opening proved to have little influence on the outcome of the game.
It made the start of the game easier, but I kept losing from guys I usually lose from, and I kept winning from guys I usually win from.
To study positional play.
This is exactly how DLM describes it. I deepened my insight of the game, but I had no skills to do what I know I have to do. Positional knowledge is only the very first beginning of positional play.
What is needed here is positional pattern recognition. Once we know how to develop tactical pattern recognition it is time to get back to positional play and to train positional patterns.
To study master games.
You are exposed to new patterns at such a low pace and the patterns are so exclusive that it works much the same like the study of positional play.
To study endgames.
Endgame methods like zugzwang, stalemate, opposition etc. are highly comparable with middlegame methods like pin, phork, skewer etc..
The approach has to be the same.
The only reason I don't study endgames at the moment is that the % of games that are decided by endgame failures from me is so low. And I'm so terribly bad in tactics that I really have to fix that first. But there comes a moment for sure that I will study endgames again.
To write a chess program.
Very interesting, it improved my insight in the game. But it helped me not the slightest bit.
To play a lot.
Playing alone is simply not enough. It is fun though.
To play a lot blitz.
Very good for developing bad habits. It doesn't help with long games. It is fun though.
To work with a thoughtprocess.
This helps a short time. But before you know it it is integrated in your system, it becomes automatic, and it is no conscience process anymore. Reduncy in the process will cost you time.
To visualize the board and pieces.
I have been #1 on the chess visualisation training site of Jan Matthies for a long time.
It improved my ability to see a board with pieces in my head. At a certain moment I could even move the pieces as if it was on a real board.
I started to dream about chessboards in 3D and technicolor.
It didn't bear an influence on my rating though.
I posted about my experiences here.
To play blindfolded.
Good to impress family and friends.
No use for actual play. See visualisation.
The main flaw of these methods was that my rating wasn't influenced a bit by any of them.
Except for the tactical training which helped me gain 200 points.
I strongly reject the idea that that improvement was the cumulative effect of all the training before. I'm sure that is not the case. In the years before I started with tactical training, I gained simply no point at all.
Beside the main flaw, gaining no rating, a few remarks can be made by every method.
This maybe helps to infer why the methods didn't work
To study openings.
At my level the opening proved to have little influence on the outcome of the game.
It made the start of the game easier, but I kept losing from guys I usually lose from, and I kept winning from guys I usually win from.
To study positional play.
This is exactly how DLM describes it. I deepened my insight of the game, but I had no skills to do what I know I have to do. Positional knowledge is only the very first beginning of positional play.
What is needed here is positional pattern recognition. Once we know how to develop tactical pattern recognition it is time to get back to positional play and to train positional patterns.
To study master games.
You are exposed to new patterns at such a low pace and the patterns are so exclusive that it works much the same like the study of positional play.
To study endgames.
Endgame methods like zugzwang, stalemate, opposition etc. are highly comparable with middlegame methods like pin, phork, skewer etc..
The approach has to be the same.
The only reason I don't study endgames at the moment is that the % of games that are decided by endgame failures from me is so low. And I'm so terribly bad in tactics that I really have to fix that first. But there comes a moment for sure that I will study endgames again.
To write a chess program.
Very interesting, it improved my insight in the game. But it helped me not the slightest bit.
To play a lot.
Playing alone is simply not enough. It is fun though.
To play a lot blitz.
Very good for developing bad habits. It doesn't help with long games. It is fun though.
To work with a thoughtprocess.
This helps a short time. But before you know it it is integrated in your system, it becomes automatic, and it is no conscience process anymore. Reduncy in the process will cost you time.
To visualize the board and pieces.
I have been #1 on the chess visualisation training site of Jan Matthies for a long time.
It improved my ability to see a board with pieces in my head. At a certain moment I could even move the pieces as if it was on a real board.
I started to dream about chessboards in 3D and technicolor.
It didn't bear an influence on my rating though.
I posted about my experiences here.
To play blindfolded.
Good to impress family and friends.
No use for actual play. See visualisation.
Monday, December 26, 2005
When everything else fails. . .
Ever since somebody lyed to me about the very existence of Santa Claus, I decided to take nothing for granted anymore.
So if someone is telling me something, it is stored internally in a file "unverified".
Only after I checked the facts myself it is relocated to the file "true according to known facts" or "false according to known facts".
At a later stage relocating between true and false can take place when new facts become known.
At that stage doubt is unnecessary. I just flick between files whenever new facts arrive.
This attitude towards the things people are telling you has upsides and downsides.
On the upside:
Especially in a learning situation my attitude has a large impact.
Nothing is believed based solely on the blue eyes of the teacher.
To prevent teachers from being driven mad by me I have a special internal file called "hypothesis".
A special thing about it is that it is a temporary file.
From everything that's stored here I ASSUME it is true. And I act AS IF it is true.
But I have the severe obligation to undertake everything that's within my capability to prove the hypothesis or to flaw it. I regard it as my duty to NOT steer things in a direction I WANT to be true. That is, I'm really trying to be objective.
When a hypothesis is flawed, I feel the obligation to develope another hypothesis, untill I have found a hypothesis that covers all facts and that proofs to be true.
So how about improving at chess?
The past 8 years I allready tried the following:
Until now every hypothesis is flawed.
Except for solving tactical puzzles, which helped me to improve 200 rating points before stalling.
The hypothesis I'm working on right now can be summarized as:
After you reach a certain level, improvement in chess can be achieved mainly by pattern recognition.
Working with CTS in the way I described in an earlier post is intended to train pattern recognition. I consider CTS as a bad way to train pattern recognition, but never the less A way.
Since this is not the time to develop a better method, I stick with CTS until the hypothesis is proven or flawed.
The positive thing I can report is that the hypothesis isn't flawed until today, while all previous hypothesises have been flawed at an earlier stage.
I hope this unenigmatizes things a little:)
So if someone is telling me something, it is stored internally in a file "unverified".
Only after I checked the facts myself it is relocated to the file "true according to known facts" or "false according to known facts".
At a later stage relocating between true and false can take place when new facts become known.
At that stage doubt is unnecessary. I just flick between files whenever new facts arrive.
This attitude towards the things people are telling you has upsides and downsides.
On the upside:
- Nobody can easily cheat me.
- I don't spend energy at good advice which isn't verified by the advicer. It is simply unbelievable how many people are stuffed to the eyeballs with good advice which they haven't checked or live by themselves. You can spent your whole life following such advice without coming a millimeter further.
- I trot on places nobody has gone before.
- People think I'm selfopinionated, arrogant and pedantic. From this place I like to apologize for that. I hope you can now understand a little where it comes from.
- I have to find out everything myself. So experimenting became part of my life. I'm used to it and I very like it.
- I can't be lazy.
Especially in a learning situation my attitude has a large impact.
Nothing is believed based solely on the blue eyes of the teacher.
To prevent teachers from being driven mad by me I have a special internal file called "hypothesis".
A special thing about it is that it is a temporary file.
From everything that's stored here I ASSUME it is true. And I act AS IF it is true.
But I have the severe obligation to undertake everything that's within my capability to prove the hypothesis or to flaw it. I regard it as my duty to NOT steer things in a direction I WANT to be true. That is, I'm really trying to be objective.
When a hypothesis is flawed, I feel the obligation to develope another hypothesis, untill I have found a hypothesis that covers all facts and that proofs to be true.
So how about improving at chess?
The past 8 years I allready tried the following:
- To study openings
- To study positional play
- To study master games
- To study endgames
- To write a chess program
- To play a lot
- To play a lot blitz
- To work with a thoughtprocess
- To visualize the board and pieces
- To play blindfolded
- To solve tactical problems without repetition
- To solve heavy tactical problems with lots of calculation
- To solve tactical problems 7 times
Until now every hypothesis is flawed.
Except for solving tactical puzzles, which helped me to improve 200 rating points before stalling.
The hypothesis I'm working on right now can be summarized as:
After you reach a certain level, improvement in chess can be achieved mainly by pattern recognition.
Working with CTS in the way I described in an earlier post is intended to train pattern recognition. I consider CTS as a bad way to train pattern recognition, but never the less A way.
Since this is not the time to develop a better method, I stick with CTS until the hypothesis is proven or flawed.
The positive thing I can report is that the hypothesis isn't flawed until today, while all previous hypothesises have been flawed at an earlier stage.
I hope this unenigmatizes things a little:)
Sunday, December 25, 2005
Merry Christmas
Things start to become interesting.
Today I played a few games at FICS to test if my skills are growing.
I didn't do too well.
Allthough 4 games is too little to draw conclusions, I didn't feel "muscled".
Doing a tremendous effort, it would be nice if "something" happened.
But there is no proof that anything is working at all.
So the faith in my own reasoning is put to the test.
Usually at this stage of an experiment flaws you couldn't even imagine before show itself.
But no flaws become apparent.
Except one: nothing happens!
And that needs to be clarified.
At this moment I can see only two explanations for this phenomenon.
First explanation:
De la Maza is wrong.
Prof. de Groot is wrong.
Tactical training doesn't work.
Doing zillions of tactical problems is as usefull to chess improvement as solving crossword puzzles.
Chess skills are innate.
I will be a patzer the rest of my life.
Second explanation:
Pattern recognition is the ONLY way to improve at chess after you reach a certain level.
I have the feeling that my calculation ability is close to its maximum after 3 years of continuous training. I just can't imagine that I will improve in this area much further.
On the other hand it is equally clear that there is plenty of ROOM for improvement on the tactical side.
The problemset of CTS is so big that improved pattern recognition at this moment contributes little to my results. The pattern recognition will only be considerable after lots of repetitions of the problemset. I just have to continue. The training in itself does nothing, only after I have stored a pattern I can expect improvement.
Both explanations are equally bizar.
Tuesday, December 20, 2005
Halfway math
Today I'm halfway the 7 circles with 35,000 done of 70,000
I'm starting to recognize the problems.
That doesn't mean that I remember always the solution, but that I realize "hey, I've seen this one before."
I have done some counting and 32 % of the problems look familiar.
The rating stays level but the succesrate has crawled up with 0.9 % to 79.9 % !
I just can't force myself to gamble so I take my time with every problem.
26 % of the problems are solved by pattern recognition alone.
The same in other words:
From the 10,000 problems in the problemwindow:
3,200 problems look familiar
2,600 patterns are stored in my system
5,400 problems I have to calculate
2,000 problems make me fail.
An increase of 0.9 % in succesrate means that I solve 90 problems more than before.
CTS
circle 0: 1470
circle 1: 1500
circle 2: 1520
circle 3: 1520
circle 4: 35,100 /70,000
Highest rating 1567
I'm starting to recognize the problems.
That doesn't mean that I remember always the solution, but that I realize "hey, I've seen this one before."
I have done some counting and 32 % of the problems look familiar.
The rating stays level but the succesrate has crawled up with 0.9 % to 79.9 % !
I just can't force myself to gamble so I take my time with every problem.
26 % of the problems are solved by pattern recognition alone.
The same in other words:
From the 10,000 problems in the problemwindow:
3,200 problems look familiar
2,600 patterns are stored in my system
5,400 problems I have to calculate
2,000 problems make me fail.
An increase of 0.9 % in succesrate means that I solve 90 problems more than before.
CTS
circle 0: 1470
circle 1: 1500
circle 2: 1520
circle 3: 1520
circle 4: 35,100 /70,000
Highest rating 1567
Sunday, December 18, 2005
And now something completely different
How is it to live 6 meters below sealevel?
The above picture give you an idea what we are talking about.
The main idea is that you pump out the water faster than it oozes in.
To that extend we have 4 pumpstations.
One of them you see below.
As you can see it has four pumps.
The total capacity of one pumpstation is 2000 cubic meters per minute.
That's enough to keep our feets dry.
We have a mobile reserve of 5 times what is needed.
Each pumpstation has 3 sources of energy: gas, diesel and electricity.
The pumps are so well engineered that they haven't failed even one hour sinds they started to work in 1956.
The idea is that they keep working when you need them.
Our polder is surrounded by dikes.
Each dike has a height that it can withstand a combination of high water and a superstorm which can occur once per 4096 year.
If the dike leaks, you can put your thumb in it.
If the hole is less than 60 meters, you can put a ship in it.
I'm glad we have no hurricanes here. Otherwise the levees would be much higher.
I agree it hasn't much to do with chess, but maybe you find it interesting.
Wednesday, December 14, 2005
#1 until tomorrow
I'm the #1 at CTS until tomorrow overtakes me at the end of the week!!
#1 in activity that is. With 33,712 problems solved.
The weird thing is that I have the FEELING that it goes better and better every day at CTS, while my rating doesn't show that. That has everything to do with the fact that it is impossible to me to speed up. I just MUST take my time. So I have more answers correct, but not faster. Hence my succesrate steadily increases. But that doesn't look as spectacular as rating improvement.
I'm about halfway the 7 circles now.
In earlier efforts that was the point where calculation gradually is replaced by pattern recognition.
I'm curious if the same will happen with such huge amount of problems as we have here.
When pattern recognition takes over, speeding up must be the result.
I'm VERY curious.
#1 in activity that is. With 33,712 problems solved.
The weird thing is that I have the FEELING that it goes better and better every day at CTS, while my rating doesn't show that. That has everything to do with the fact that it is impossible to me to speed up. I just MUST take my time. So I have more answers correct, but not faster. Hence my succesrate steadily increases. But that doesn't look as spectacular as rating improvement.
I'm about halfway the 7 circles now.
In earlier efforts that was the point where calculation gradually is replaced by pattern recognition.
I'm curious if the same will happen with such huge amount of problems as we have here.
When pattern recognition takes over, speeding up must be the result.
I'm VERY curious.
Monday, December 12, 2005
Review of Deep Fit
Nine month's ago my new chesscomputer did arrive.
I made a long natural selection before I ordered it.
The engine is programmed with genetic algorithms for optimal piece evalution.
"Deep Fit" is build by Darwin Evolvued LTD.
The manual gave a clear description of what to do.
The chess engine came with lots of biological stuff in Erlenmeyer flasks.
First I had to mount the craddle.
Then the different genome-rich potions had to be fertilized at the back rank of the mate-chambers.
The concoctions had to be poured in special IVF-proof glass tubes.
These had to be stored in the craddle at a temperature of 37 degrees Celsius (99 F) for nine months.
Today the nine months are over.
Conform the instruction manual I disassembled the most unfit baby-engines until two of opposite construction were left.
What is left are the most unugly ones.
The only thing what's left to do is to optimize them by adapting them to the internet.
I'm sitting for 3 hours now staring at the two.
My bottle of wine is almost empty.
For some reason I find myself in a post natal depression.
I try to reach for the internet connector.
But somehow my arm seems to be petrified. . .
I made a long natural selection before I ordered it.
The engine is programmed with genetic algorithms for optimal piece evalution.
"Deep Fit" is build by Darwin Evolvued LTD.
The manual gave a clear description of what to do.
The chess engine came with lots of biological stuff in Erlenmeyer flasks.
First I had to mount the craddle.
Then the different genome-rich potions had to be fertilized at the back rank of the mate-chambers.
The concoctions had to be poured in special IVF-proof glass tubes.
These had to be stored in the craddle at a temperature of 37 degrees Celsius (99 F) for nine months.
Today the nine months are over.
Conform the instruction manual I disassembled the most unfit baby-engines until two of opposite construction were left.
What is left are the most unugly ones.
The only thing what's left to do is to optimize them by adapting them to the internet.
I'm sitting for 3 hours now staring at the two.
My bottle of wine is almost empty.
For some reason I find myself in a post natal depression.
I try to reach for the internet connector.
But somehow my arm seems to be petrified. . .
Sunday, December 11, 2005
Riddles
Before a break at CTS my rating hoovered around 1530.
After the break of 3 weeks, my rating nose dived to 1470
Why so extreme?
It took me a month to get back to 1530 again.
Why so long?
In the mean time my succesrate improved from 79.0 % to 79.5 %
Since it is an average over all 33,000 problems that is much.
Why so much?
Most people's OTB-rating is about 300 points higher than their CTS-rating.
Except Celtic Death, who's OTB-rating is more than 50 points LOWER than his CTS-rating.
Why?
My new approach to redo my failures by visualising them in the mind seems to work.
Why does my rating at CTS improve so slow?
There seems to be only one answer to these questions: to continue until I find out.
After the break of 3 weeks, my rating nose dived to 1470
Why so extreme?
It took me a month to get back to 1530 again.
Why so long?
In the mean time my succesrate improved from 79.0 % to 79.5 %
Since it is an average over all 33,000 problems that is much.
Why so much?
Most people's OTB-rating is about 300 points higher than their CTS-rating.
Except Celtic Death, who's OTB-rating is more than 50 points LOWER than his CTS-rating.
Why?
My new approach to redo my failures by visualising them in the mind seems to work.
Why does my rating at CTS improve so slow?
There seems to be only one answer to these questions: to continue until I find out.
Sunday, December 04, 2005
The role of speed in chess improvement.
The role of speed.
(did you see the name on the boat?)
When one starts working with CTS, the most remarkable fact is that the time constraints are very tight.
This easily leads to the wrong conclusion that CTS is primarily beneficial for learning how to play blitz.
I made this mistake myself.
It is easy to underestimate the role of speed in normal chess.
I have posted about speed it in the past. It is still interesting and relevant to read.
A player with a higher rating plays faster.
I found an average difference of 3 x faster per 100 elopoints.
Whether this is 2x or 5x in reality is not relevant for understanding.
This means that 300 elopoints = 3 x 3 x 3= 27 x faster.
I know this sounds unbelievable.
Where does this speed come from?
Of course not from a faster working brain.
The difference in speed comes from switching over from one sort of brainprocess to another. One is "finding out" the other is "recognizing".
The time you need to find out how to solve a chesspuzzle differs vastly from just recognizing it.
Take a look at the next diagram:
Black to move.
If you have to find out the move of black, it will probably cost you about 1 - 2 minutes.
If you recognize the underlying pattern, you are ready within 10 seconds.
Let's take a some somewhat closer look at this finding out stuff.
This finding out is usually referred to as "calculating".
What you do here is a sort trial and error.
What happens if I do this or that?
This trial and error continues until you DO recognize a pattern.
Then you know you are on the right track.
If you have no patterns in your brain-database at all (= you are a genuine novice) it will at least take you a few hours (or days) to "calculate" what this position is all about.
Statistics of this problem:
59 men tried to solve it.
31 succeeded.
28 failed.
From the 31 people who succeeded, 27 finished it within 10 seconds.
The 28 who didn't solve the problem gambled and failed.
It will be clear that the persons who failed have to do some work.
Otherwise CTS is useless indeed.
This work consists of switching over from one brainprocess to another.
First you have "to calculate" the solution. Then you have to store it as a pattern in your brain so you can "recognize" it the next time.
I posted about my experiments with this a few days ago.
So the flaw is not in CTS or it time constraints, but in the way we work with it.
It is a tool that in a very precise manner indicates what patterns aren't stored in your brainsystem yet.
CTS does have a main flaw though. The problem window of 10,000 problems is much too wide to have an effective repetition of problems.
Subscribe to:
Posts (Atom)
