Sunday, December 14, 2014

tactical nonimprovers


Munich was mentioning improvement in his last comment. So i did copy his , upandcommer's and my ratings at chesstempo here.

Munich's rating.
The small improvement after march  2014 is result of a change in the Ratingsystem of chesstempo

My rating
I did start training at ct when there was complete different ratingsystem, that explains the peak at the beginning. 
The small improvement after march  2014 is result of a change in the Ratingsystem of chesstempo


The rating of upandcommer
The small improvement after march  2014 is result of a change in the Ratingsystem of chesstempo


In my eyes there is no "relevant" improvement to see.


Here a nice paper to read : The Role of Domain-Specific Practice, Handedness and Starting Age in Chess by Fernand Gobet and Guillermo Campitelli Centre for the Study of Expertise Centre for Cognition and Neuroimaging Brunel University 

In summary, starting to play seriously not later than the age of twelve, carrying out individual practice such as reading books, playing with others, and receiving feedback from a coach seem to be all important factors to attain a high level of expertise in chess. There was some evidence that individual differences in abilities not related to the chess environment differentiate between players and nonchessplayers (direction and degree of handedness)

Monday, December 8, 2014

Is chess improvement a matter of talent or a matter of work?

Intro:

Its often said that it takes roughly ten thousand hours of practice to achieve mastery in a field. There had been scientific studies telling that "talent" would be unimportant. You find plenty of further information about this if you search with a internet search machine for "rule of 10000 hours".

There are many chess players which have spend much more than 10000 hours on chess and still: they are not Experts. I am watching often the progress of Users of Chesstempo.com ; usually there is none. These users get better on puzzles they did see before, but the performance on puzzles they never have been seeing before remains more or less constant. So i was thinking the tactical skill develops in the youth of a chess player. Now i found a hint that this might be wrong.

Time limits

When i started to solve tons of tactical puzzles in the hope to get better in chess i did improve my rating at several server continuously. It did take me a while to recognize that this improvement was an illusion. I simply was taking more and more time to solve the puzzles or i did see these puzzles before. My performance on new puzzles, with limited time given, did not improve. You can read at my early posts in this blog about the progress of my insights in the relationship of time, score and rating in tactics. You may read in Empirical Rabbits blog more detailed analysis about this.

A strange tactician

Several weeks ago an other blogger did start solving puzzles at chesstempo. His speed in solving puzzles was about twice as quick than i so i did expect him to get a rating about 140 - 200 Points above me.  Surprisingly he did just reached about my level. Seemingly he was unable to solve complex puzzles, probably because of a weakness in calculation, visualization  or thinking process. His rating in not timed puzzles ( Standard ) was even 500 Points below my Standard rating. Of course , i reached a high standard rating because i was sitting sometimes hours in front of these puzzles before i made my first move. The calculation that a tactician gets 100-200 points better with every doubling of thinking time is only correct as long as the tactician is able to visualize all positions. But as higher rated the puzzles as deeper to calculate.
As soon as the visualization stops.. longer thinking is not beneficial anymore. 

Suspicion 

No doubt, many kids do improve in chess. Some kids are seemingly "talented", you get immediately the impression that they will have big success in chess if they just continue to work on it.


So whats about their tactical skill? Do these kids improve their Calculation, Visualization, Thinking process -skills or do they improve their tactical skill = speed?



A boy on the road to get GM

Several years ago i met in the Internet (chat and blog ) a rapid improving boy. At this time he had about my  rating. I was watching his improvement. Wikipedia told me about his parents which are titled players. Today he is a strong IM and seemingly not very far from getting GM

Analysis

Chesstempo had statistics about his Blitz performance when the boy was about 13 years old. I took a sample of 37 puzzles from his statistic. His average rating was 2307 but the average rating of the puzzles had been 144 points higher than today ( rating drift ) so his "real" rating was 2163. His average speed was about 5 times as quick than the average tactician. With Empirical Rabbits K of 200 points per doubling calculation time we get a tactical skill of 2623

Today his average rating is ( in a sample of 30 puzzles ) 2323 but his speed is only about 3 times as fast as the average tactician. With Empirical Rabbits K of 200 points per doubling calculation time we get a tactical skill of 2647, that's about the same rating as he had 6 years ago.

His Fide Elo did improve 500 points during this time and is now almost 2500.

Analogy to Intelligence

One of the more surprising findings to emerge from the intelligence literature is that an individual’s ability to think in highly complex and abstract forms is related to speed in tasks as simple as “press the lighted button.” Simple reaction time tasks like this have amazing predictive power for performance on much more elaborate tasks, leading some theorists to propose that such reaction time (RT) measures grossly index the integrity or speed of processing in a way that benefits all tasks. ( taken from here ).

So i wonder if the tactical skill ( measured as Empirical Rabbit describes)  represent a "natural talent" of a type of "chess intelligence" ?

Sunday, November 23, 2014

Gaining Speed

I did not solve too many Mate in one puzzles the last days, i was watching the World Chess Championship. Now back to work!

Next experiment will be: Spaced repetition of Mate in one problems. I already experienced that Spaced repetition does not help ( at least me ) with complex puzzles. But i had some seemingly positive effects with simple puzzles and Munich estimatingly  too.

I have now 200 000++ M1 puzzles, i will create buckets with  300 puzzles and speed them up by a factor 2. Without knowing my future experiences i guess : 10 repetitions with 10% increase in speed will do. I will start with a space of 4 Days and 4 Buckets with 300 puzzles each.

So : Back to the Saltmines

Thursday, November 6, 2014

Conclusion

A small summary of the last posts

We know : 

  • improvement in chess is hard for adult players, learning of many facts not 
  • improvement in tactics (measured in a timed rating on non duplicates at a chess server after 4000 attempts) seems to be virtual impossible
  • rating, score and time_used are related, twice as much thinking time produces moves which are ~ 200 (+-100) Elo points better
  • the improvement in "Mate in 1" measured in Mates/Min  is not easy ( maybe impossible for an adult too????? )
  • improvement in board vision exercises, skill trainer and some other special chess related exercises is often easy
It is not easy to implement a rating system for chess puzzles but if they are easy enough so that the score is always close to 100% then we can simply use "the time used to solve these puzzles" as "rating". We know: as quicker as better and a gain in speed is equivalent to a gain of  Elo points ( not in chess in general but on this set ).

Conclusions

So i was wondering why is it so hard to improve in  "Mate in 1"?
Mate in 1 is a microscopic small subset of tactical problems, there are Forks, Skewers, Pins, ... Mate in 7...Hanging Piece.. and there are "uncountable" combinations of such Motives. If it IS hard to improve in "Mate in 1" then its easy to believe that it is virtually impossible to improve in tactics because there are hundred times more motives in tactics as a whole than in "Mate in 1", the tiny small subset.
So if you cant improve ( easily ) in "Mate in 1" by solving 10 000 of Mate in 1 puzzles you will definitely not improve in Tactics by solving 10 000 tactics puzzles

Well is that true? Why do we think so?

Many people think, the performance in chess is based on the knowledge of pattern. This is not exactly correct, if i explain a pattern to a beginner then, even if this beginner understand the pattern, he/she will still apply this knowledge less good than me, much more experienced player. It is important to make such a knowledge to a skill by tons of practice.

What pattern are we talking about?

When we think about tactics we think about tactical pattern like a mate at h7, the bishop sacrifice at h7 in the advanced french and other typical positions.

Now: if the improvement in "Mate in 1" is hard because of the number of pattern, then why don't we improve solving thousands of such problems. We should learn the pattern... weeell..

Seemingly the number of unknown pattern is close to the number of puzzles!! Learning new pattern don't help, the next puzzle will be a new pattern or its a pattern you already know. You don't learn anything new ( or you would improve ;)

But to solve a "Mate in 1" puzzle does not only require the knowledge of mating pattern. We need to check: how can we give a check, does the king has any squares to escape, is there any piece of my opponent which can intersect, is it possible for my opponent to take my piece...

If i start to analyse which of the (improvable) skills are related to "Mate in 1" then the result is astonishing.. almost all. Even the Fork-training has some relationship to "Mate in 1" because if you want to give check with a piece ( say a knight ) then you need to find the (knight-) forking square of the present position of that piece and the king. ( As example: say the opponent King is at e8 and your knight at g8, then you need to put your knight at f6, a knight at f6 forks the squares g8 and e8 ).  This is of course a very weak relationship ;)

An other example of related exercises is "chess minefields" ( see the link at the right side of this page ). Here we are looking for a save square for our pieces which we have to put "in bughouse style" on the board. The skill: "which square is save" is needed at "Mate in 1" to judge if the opposite king can escape ( has a save square ) or not. If the opponent king has a save square to escape than its not mate.
Especially this skill is very important for bullet player. It happens quite often that in Bullet and Blitz games someone drops a piece. Of course this skill is important for OTB players too, they need to know if at the end of their calculation any of their pieces are hanging.
So it might be easy to understand why blitz-players can solve "Mate in 1" puzzles better.

So i think an improvement in "Mate in 1" should result in a higher performance in many , if not in all tactical situations, because it might be necessary to improve in many sub skills of all tactics.




Monday, November 3, 2014

Confusion

While it seems to be easy to speed up at simple exercises like Board Vision Exercises it seems to be hard to improve at "Checkmate in one" or "Checkmate in two". Hard to believe that the "ability/disability" to improve in tactics might be ( about ) the same as the "ability/disability" to improve in "Checkmate in one" but i start to wonder...
I work now at CT-ART Mating Combinations   but i have doubts that this will have any effects. If Tactics and Checkmate in one are really that close related ( well are they ? ) then the mating pattern cant be the reason.  It might be the flexibility of the memory, how quick we can push the position into our inner board and how quick we can pull valuable information from this inner board. Or is it the ability to make decisions?..
How to improve here? Spaced repetition? Thematic subsets? Specialised exercises like "Find all checks"?

Checkmate in one, that is a step
On which I must fall down, or else o'erleap,
For in my way it lies. Stars, hide your fires;
Let not light see my black and deep desires:
The eye wink at the hand; yet let that be,
Which the eye fears, when it is done, to see.