Sunday, February 05, 2012


Munich found this crushing attack by white at CT. See diagram.
It is a 1830 rated blitz problem which is a beautiful illustration of pattern recognition guided by context.
White to move.

You can find the solution here.

As you see a lot of people miss this tactic due to the wrong context. At CT you expect problems of a certain difficulty. You don't expect a simple capture. The 1830 blitz rating and my description of the crushing attack might have put even you on the wrong foot. But then again, if attacking a queen isn't crushing, I don't know what is:)

The message is that we see what we are looking for and that we don't see what we are not looking for. Even when it stares you in the face. Allthough that is very economical for the brain, it means that we have immense blind spots. Both in chess and in life.

If you think this doesn't apply to you since you saw the queen capture immediately, you are wrong.

At the Tata tournament I asked myself during the games what the potential  reach could be of the training of complex tactics which I'm doing. I realized soon that there are, besides the opening, no tactic-less moves in my games. The amount of moves you can do is always restricted by the moves that cost you wood. You always have to look for tactics yourself. This always checking for tactics for both you and your opponent is time consuming, errorprone and draining your energy. Of course not every move is a tactic, but you have to watch out for them anytime. They always play a role in the background. Being able to do this automatically and flawless would have an immense effect on my games.

The training I'm doing I consider not to be about tactics, by the way. It teaches precision in reasoning. You usually recognize the patterns that are involved soon enough. But then you must make the combination work. And when you think it works, you must attempt to falsify your move. From the 37 topics on the list which constituted my thoughtprocess there are only a few left. I left out those topics which I already check for automatically. So it became a highly personalized list:

1. What is this position about?:
  • Mate
  • Gaining wood
  • Promotion
  • Counterattack.
2. What if I could do 2 moves in a row?

3. Can I falsify my move?

The first question helps for orientation in the position.
The second question helps me to identify threats. CCT prunes a lot of garbage moves, but it introduces new ones. Since I'm always looking at checks and captures automatically, there is no need to guide my mind towards that. But question 2 brings above all the relevant threats in the position.
The third question is designed to test the correctness of my combination.

You have seen me change from exercise to exercise the past 3 years. The reason for that was simply that I was never quite convinced that it was the right method.That's different now. I have at least found the right subject to train. Precision in reasoning is imo the subject to be trained to the max. What can change is the exact method. For now the current method seems to be working fine, so I will stick to it. In fact I already do for 2 months now.


  1. The agent provocateur says:

    If your trainingsmethod is good, then your Fide Estimate based on Standard should raise. If your Fide Estimate dont raise...

    An other BASIC "IF THEN" ;) :

    If you where shure your method works Then you would check your Fide Estimate...

  2. It is more logical to measure my real rating in stead of an artificial one. Since I want to be better in real games, not in complex tactics.

    Besides that, I take about two hours per problem to work out all important details and "to get the hang" of the position, so that I would recognize the reasoning "in the wild". I assume that will lead to a lowered Fide Estimate. And I don't want to solve at speed especially for an estimated rating solely.

    The first 200 problems I scored exactly 0%. Now I score about 10%. Which is proof enough for me to keep going.

  3. The fide estimate calculates your tactical strength. CT-Standard-Rating has a Duplicate Rating reduction and dont consider the used time and other factors.
    It is more logical to pay attention to an immediate sucess indicator.
    Success at OTB might be the result of many different reasons. If you want to see if your tactic training shows any effect the fide estimate would be the quickest and most precise method.
    You still may analyse your problems as long as you wish, after you did try to solve them first for ex.
    Then you have the analysis board of CT available where cou can add an engine wich can speed up and check your analysis too.
    If you are writing a subroutine ( to stay in the World of BASIC ) then you check this subroutine first "alone" before you check it in the whole program.

    But i cant confince you...

  4. Do you mean I can get a rating when solving my own selection of problems?

  5. Using the computer and analysis board of CT is a good idea, btw. I use now Arena to play the positions out. Thx!

  6. with an own selection of problems you will get no fide estimate, no. You may choose standard hard to get heavy problems. if you solve "only" max 5 min and analyse then, you dont loose much time on problems you dont want to solve.

    I use a little different method then you. I do many blitz problems and then repeat and analyse my blunders.

    The blitz problems are "necessary" to get a fide estimate, to get datas to analyse my "history" and to find blunders = things i do wrong. At the same time i hope to prevent to get slower and slower. In combination with other things.. i do improve in tactics, slowly, but at least i do.

  7. Yes, it's true, I'm hard to convince. Yet I'm not "unconvincible". Thank you for being persistent.

    Since this is the only training I do (besides tactical speedtraining when I'm too lazy to think) I am not worried that I don't know the cause when my OTB rating rises.

    I assume that one or two years of training are necessary to show significant progress anyhow.

    Besides that, I already know that it works because of an increasing percentage correct.

    I belief that the right order is: first do it right, then speed it up. Not the other way around.

    The difference is that I know now exactly what I'm after and why. That makes a big difference in what I get out of a training!

  8. another crushing attack! rated 1746 CT blitz rating.
    But since you have been warned from the other puzzle, you will probably see it immediatly.
    Nevertheless it is amazing why it took me 37 seconds to solve this one:

  9. Even though it is just a 1657 CT Blitz rated puzzle, it is still amazing how long the average solution time is. I needed 1:04 minutes to solve it.

    In Standard the puzzle has only 1482 CT standard rating, but the average solving time is 1:20 Minutes here.

    Can there be something more obvious than this one? The last move white did was moving his bishop a5-a7.

    How are you going to check mate?

  10. Last one (unless I really find a high rated one):

    I even failed this one (CT Blitz rated 1697) after 25 seconds thinking.
    The last move is so obvious, and in jumps right into your face.

    I dont know how to avoid this blindness. I did attack training with the fritz gui and can almost do 50 attacks per minute (was it 48, cant remember)?

    I reckon pattern recognition can hurt you here. But it is also your expectation. You dont expect white to do such a stupid move.

    I could imagine my a weaker player with less pattern knowledge is more likely to solve the above one.
    But does that mean pattern recognition ist a bad thing to train? No, I dont think so. It is still important.
    It made me improve in my tactic skills, and the prove which was first it bit shaky is meanwhile quite evident.
    I did improve, and I think I am still improving with my system. (which is mostly derived by aox and tempos ideas, while they themselves might have been influenced by empirical rabbit).

    Give me a forum message in CT and I can provide you with proof.