Sunday, March 04, 2012

Is one plus one actually two?

According to Munich and Aox one plus one equals two. If you know all the constituent parts of a combination, you will be able to find the whole combination. The constituent parts are identified as "patterns". The fact that there are only about 27 tactical elements already shows a fundamental flaw in this reasoning. With only 27 elements you would know them by heart within a few weeks or so. Since it is evident that we aren't able to solve all combinations after only a few weeks, there is more to it.
If you call a combination of tactical elements a pattern, in stead of the tactical element itself, another problem emerges.
If you write out a combination as 27 x 27 x 27 x 27 x 27 etc. you will find that the numbers will grow rapidly. So rapidly that you need to learn at least millions patterns by heart to cover the most. Munich and Aox work their way around this by careful selecting their problemset based on their assumed frequency of occurence in real games. Of course it is logical that they worry about quantities and amounts of patterns per unit of time this way. Hence speed is their motto.

I don't buy this since I have tried all kinds of problemsets from easy to complex, tagged and untagged, big and small, fast and slow, blind and with the use of an analysisboard, with and without writing down the lines, standing on one leg while humming Mary had a little lamb etc. etc.. My results didn't differ from the avarage improvement of the Knights Errant. Just like the results of Munich, Aox and vitoc73 don't differ from that.

I take a more holistic approach, "The whole is greater than the sum of its parts". This means that there is something different to learn than patterns alone. I'm studying Weteschnik's "Chess tactics from scratch" (hattip to Empirical rabbit) and he has a very interesting chapter about this subject called "status examination". Here you learn to guide your pattern recognition by evaluating the features of each piece. Even more reductionistic than holism, I would say, but it is all very logical and it guides the mind towards the right patterns.





.
.
.
.
.
.
In the end the reductionistic approach and the holistic approach probably will turn out to be the same. But at this moment, with overwhelming complexity, holism is a more logical approach.


They say that my method is not proven. But think about this: when I started with the first 50 problems I scored 0 out of 50. After just 100 problems I score 10 out of 50. From 0% to 20% in just 150 problems! This means that there is a massive transfer from acquired knowledge in known problems towards new problems! And if it already has such effect on high rated problems, how massive will the effect be on lower rated problems!


There has been a lot of talking about FIDE estimated ratings. I don't accept that as a convincing method. The only thing I want to know is what the effects are for my real rating. Then I will know how my efforts translate to OTB-play. No reason to muddle around with estimates. That is just a waste of time. I have said this a few times and this is the last time.


Aox didn't see where the speed should come from if you use a slow method in stead of a speedy one. My take is that you don't speed up the same things that you already do. In stead you simulate speed by doing something different. Look at it this way: I you must repair a car by means of trial and error, it will take a very long time. No matter how fast you are able to do the constituent activities. But if you read the manual first, you may speed up by a factor 100 or more, even if the speed of the separate activities is lower. Reading the manual is a very slow process in itself. It will speed up your activities by pruning the useless ones, though. Just like knowledge does to the tree of analysis.

17 comments:

  1. There are not only 27 tactical pattern. Dan Heisman and Dvoretsky are talking about ~2000 tactical pattern. Dan Heisman is using the analogy of a multiplication table.

    He give the example athttp://www.chesscafe.com/text/heisman109.pdf :
    Suppose someone asked an adult, "What is 7x8?" and he replied, "That's easy!
    Of course it is 8+8 = 16 + 8 =24 + 8 = 32 + 8 = 40 + 8 = 48 + 8 =56. See!
    Anyone can do that!"

    You might think that the responder knows how to figure out the answer, but
    that he did not memorize his multiplication tables. I deal with this type of
    problem all the time.


    Its better to learn the multiplication table and better from 0-20 then just from 0-10

    He suggest to learn "the" ~2000 most important tactical pattern.

    You will need several examples of each pattern (100?). Then you will need to learn them = many repetitions. Even if we have everything well prepared ( maybe by Mr Dvoretsky )its lots of work.

    Now we dont have a learning collection of this type we need to learn "random" = even more work.

    There is some research on Experts, especially about Chess Masters. They have a repertoir of about 50 000 chunks ( = ~ pattern ).

    The factor speed:
    1) You can measure how well you already know a pattern or how well you know a special problem if you measure the time you need to solve it.

    2) As higher the rating as quicker a tactician can solve a problem
    Just do an experiment and look at problems at CT rated about 1000.
    The average time of such a problem is much higher than the time you need to see the answer. It is well known that Masters are much quicker in tasks like: "Is the King in check", " who has more material" and so on. Without getting faster at low rated tactics there seems to be no improvement possible.

    3) speed is a (multiplication!) factor of intelligence. The fluid intelligence is ~ STM * Mental Speed. I hope/think that there might/should be an analogy to chess.

    You did the same training as munich, repetition of easy tactics sorted by tactical pattern. Munich just analyse his progress/sucess by watching his solving time and he is selecting the still "unknown" problems for further training. But you did never check, how much you tactical few did improve by this training, munich did! Thats why he is convinced about your former method.

    I did leave my plateau in tactics by a slightly different method. Both methods are based on more ore less easy tactics and on repetition to learn them.

    (eehh, the word Holistic is usually just an other Word for "Magic" )

    Tempo you might be right, maybe you method works

    But

    there are no >facts< that it does and i cant see, that you method speeds up the solving of easy problems and i think, without speeding up at simple problems there is no improvement (no prove, i know).

    And
    MY experience with CT-Standard was negative

    And i think:
    If you would have monitored you tactical strength then you might have see what works and what not.
    To measure tactical strength with OTB success is to unprecise.

    And
    Learning easy tactics by repetition did work with Empirical rabbit, Munich and me.


    So

    0 : 1


    With a (very?) big questionmark, but....

    ReplyDelete
  2. Yeah, I noticed that people with a big heart have hijacked the word "holistic". Do you know a useful synonym?

    ReplyDelete
  3. Nobody knows what a “pattern” is or how many of them there are. By common usage of the word:

    * If you know a pattern you can often recognise it by studying a position that contains it.

    * Studying the position in this context does not involve detailed analysis of moves.

    * If you find a pattern (whatever that is), it may suggest a combination, which may or may not work.

    * Except in the simplest cases, you need to check to see if the combination does work, which may involve spotting other patterns.

    You do not usually have to memorise new patterns to combine tactical motifs, but you often have to remember exceptional cases that are hard to see.

    ReplyDelete
  4. Long story short:
    a) add intelligence
    b) then automate it

    ...but...

    ...it might be good if you find intelligence where you need it the most. Again I am talking of statistical relevance.


    Here we can see why it might be more and more diffiult to improve as soon as we reached A-Class level.
    The relevance is highest at around 1500 CT Blitz rating.

    To add intelligence you need to solve a lot of puzzles to find puzzles that are easy for others but not easy for you.

    If you are around 1800 in rating and solve 1800-rated puzzles, the number of these 1800's are not that statistical relevant anymore. But their number is still reaonable. But at 2000 and above - their relevance gets really bad.

    It is a trade-off between adding intelligence and where to seek it. If you do too easy puzzles, you hardly add intelligence. If you do too difficult puzzles (guidance or tactics does not matter), then you add a lot intelligence - but it is not relevant!

    So are you sure, you can solve 95% of all Blitz rated 1600's ? Or will it only be 70%. Then I'd say go for it, seek there. Add intelligene with guidance and pattern recognition, anything that helps to make this range a 95% success range. You do them only one time. After this, you filter for puzzles you did slow or wrong.
    These trouble puzzles add massive Intelligence while statistically are very relevant.

    ReplyDelete
  5. I deny that my problems lie in pattern recognition. Hence is the statistical relevance of the patterns irrelevant to me.

    What I learn from 2300 rated problems is what I need the most. I learn how to combine patterns. At the Tata-tournament I spilled whole and half points to exactly this problem, the inability to combine well and within a reasonable time.

    For years I have studied at least 6 queen sacs before breakfast. So I have seen a few patterns. It does not work any longer for me. There will come a moment you will reach that point too. Sooner or later. Believe me.

    ReplyDelete
  6. Tempo,

    I completely agree with your conlusions:
    1. The only relevant measure for progress is your OTB rating. There is no other reliable method to measure the transfer.
    2. While doing tactics that are familiar to you, may help you maintain your tactical awarness, they do little to help you to reach a higher level. This would only be the case if you progress is stagnated because of lack of awareness of these patterns during the game. Rising to another level is only possible if you pick up more challenging exercises.
    Keep up the good work!

    ReplyDelete
  7. I just hit a new ATH at chesscube.com with 2006 rating points. I usually play "5 min + 3" sec games
    I am still improving. My name in chesstempo is "J_Munich".

    I dont know if a chesscube rating is comparable with fide elo. I very much hope yes. I dont have a fide rating, only an estimate at CT with 1990 (estimated on 1st timers in Blitz mode), which is very likely also an ATH.
    So I am now an expert player, leaving A-class behind me?


    Anybody in Chesscube? How are your ratings?

    ReplyDelete
  8. Munich you have to play a tournament. Blitz and OTB is different.

    http://www.schachlinks.com/suche/schachturniere/index.cgi

    ReplyDelete
  9. Aox, no Blitz is not different.
    According to your study about speed, blitz games on high level give you a good proxy.
    Besides, I never came above 2000 (meanwhile 2024, my performance of the last 40 games played today and yesterday against players of an average rating of 2000++ is +24/=3/-13.
    This gives me a performance of well over 2100. I never ever had that before in my live in chesscube. Nowhere near that. The games were 5min + 3 sec. So not sooo fast. I posted the last 2 games I did today in the CT forum. Really good play. I did not only win. I crushed them with tactics. This is real.
    o.k. some people may say only long OTB is proof. But how comes I never got into these heights before? I am not winning my games on time. (it is rather the opposite way, I tend to get into time troubles).
    I play amazing. I wonder about myself.
    Boy is that fun!
    It suddenly made "click" somewhere inside of my head and suddenly I play way better. It is more then I ever hoped for in my wildest dreams! I am serious about this!

    Due to my live, (traveling a lot in the world), I will hardly get the opportunity to play long OTB tournament games. I wont come up with this proof, maybe in some years. But is it really necessary? Your scientific study says, that if I beat masters in blitz, it is very likely I translate this into OTB play, too. And it makes sense. In these games I played today, I remembered a lot of CT puzzles. I had really in-depth variations during my thinking.

    Anybody who wants to check and get a feeling for my new super strength is free to challenge me on www.chesscube.com
    my user name there is: J_Munich

    What I worry most about: what if I just have an exceptional good run?
    We all do have ATHs from time to time.
    I really hope I can keep my performance.

    ReplyDelete
  10. P.S. I just checked some names in Chesstempo if they exist in chesscube, too.
    For instance I found luckypatzer, who has a CT Blitzrating of 2203 and has done 1664 puzzles so far. So it should be about stable. He is an IM.
    In Chesscube he has an ATH of 2306 ratings points.
    I found you aoxomoxoa, too, but you have not played there for 183 days. With an ATH of 1906.

    So if I beat people like 2200, I probably have hurt really good players.
    Or in other words: all people are free on whatever time they like to chose to beat people like I did with chesscube ratings 2100++. The average player I played during the last 12 games had about 2100 and I scored 8 out 12.

    ReplyDelete
  11. Munich said. "According to your study about speed, blitz games on high level give you a good proxy."


    Thats a study about player wich do a lot of OTB.
    Many things wich are "possible" at blitz, dont work OTB.

    Just do some OTB, its an experience.

    ReplyDelete
  12. Chesscube ratings are generally inflated +100, 200 elo. Comparing it with chesstempo is double funny.
    For example, here is a chesstempo forum post about rating comparison.

    http://chesstempo.com/chess-forum/empty-t132.0.html

    "I feel like this topic can give a little more, My rating on chesstempo standard now is 2000+ but on sites like Fics my rating in blitz is a horible 1500, "

    In essence, I don't think you really understand what ratings pool are about. Nothing a quick wikipedia does not solve so fire google. You probably play around 1900 live odb 5 min blitz if you have 2100 on cube, god knows how you'd measure in normal time limits odb. FYI blitz ratings in general higher than normal time limit ratings. Kingscrusher, a noted player on the internet, is around 2300 at CCube, but he was around 2400 there, he is 2130 ODB. Orangle/apple.

    ReplyDelete
  13. in the same post I linked:
    (Sorry I did not read all the stuff there, but for entertainment I paste some bits here from the next page)

    "Dear Cheestempo friends! I am very concern about tactics ratings and FIDE rating because we obviously spent lots of time here, but not mooving forward in real chess."

    "I know several players who will solve tactical problems faster than me, when shown a certain position, but who wouldnt stand a chance against me in a 4 hour game. They may be smarter than me.. but in a real game they never reach a position from which they can launch their firework of tactics...

    Tactical ability is only a small part of the tournament player´s strength as measured by FIDE-rating, and to my experience this ability may vary quite a lot among the ordinary players/strong amateurs I know (1200-2200) "

    I wanna laugh, but there is nothing there that makes a new impression on me.

    I'm gonna quote myself here.

    "I told you so."

    z.

    z.

    ReplyDelete
  14. For those who think that there is no correlation between high chesscube rating and high FIDE elo rating - I very welcome you to play me. I am around 2000 in chesscube. Take me as estimate to your own skill.

    I really wonder about the reactions here. Of course there is some correlation.
    High rated users at chesscube are stronger than low rated ones.
    This isnt only in the FIDE elo system so.

    Same valid for chesstempo fide estimate ratings. The higher the rating in chesstempo (in Blitz mode), the higher the real Fide elo of the user.
    Simple prove: look at the ranking list at Chesstempo in Blitz mode. Look where the IMs and FMs are listed. They are NOT randomly listed, but they pile up in the top places of that ranking list!

    ReplyDelete
  15. Of course there is correlation, I'm just saying you are not a 2000+ player odb, so hold your horses. If we play with x10 elo, you'd measure 21.000 in 5 min blitz, it is similarly not true but possible on the internet. I'm the second one here to point out that 5 min blitz on the internet is not reliable measurement, it is a blitz game for christ's sake. On the internet. I don't play 5 min games at all, frankly, I consider them clownish so I can't take on your offer. I don't think Tempo plays them regularly also. I'd similarly pass on an offer of playing tennis on a ping pong table. It is very good that you feel yourself confident in it,and asserting how it correlates with normal tennis, but as an old fashioned tennis player, I like to play on a standard tennis court. Call me crazy! The point is, 5 min internet blitz is not a reliable odb measurement. You can be defensive about it, but it is true. Do your homework already.

    http://en.wikipedia.org/wiki/Correlation_does_not_imply_causation

    duh!

    z.

    ReplyDelete
  16. Munich your blitz rating and your Fide Estimate say: your strength is about 1900-2000 +- (150? 200?) . But your imrovement is wonderful and dramatic. Now you can beat many clowns at OTB in blitzspeed. But it might be interesting for you to play against some 2000 OTB. Tournaments are an experience. I did play often against 2000+ ( i will stop this year with OTB ). Its a different world.

    ReplyDelete