Wednesday, August 08, 2012

Mutual influence between tactical elements

For every 2000+ rated problem at CT I make a list with all tactical elements that I recognize in the position. In the beginning it took me about 40 minutes per problem to find all elements. Often overlooking one or two important ones. After about 50 problems, I'm now able to write them down in about 10 minutes. And I overlook important elements less often.
But guess what, it hardly has any influence on my succesrate!

Of course, if you don't recognize all relevant tactical elements, you will not find the solution. But finding all elements does not guarantee that you will find the solution. Moreover, it has virtually no effect. This means that there are other, more important factors at play.

Black to move.
You can find the solution here.

Below you will find a table with all tactical elements for both  sides. Such table is a good starting point.


5.Double attack
6.Discovered attack

Kf1,Qc2 (after decoy Bc2)


7.Discovered attack




If a tactical element has no defrender, this means that you can execute it right away and gain wood or mate. A defender can defend in a lot of ways. It can protect a target, it can protect an attacking square or it can simply stands in the way on the path from attacker to target.

If a defender is listed in two rows, this might mean that it is overworked. In that case you can execute one of the tactical elements and it will decoy the defender from the the other tactical element it is defending.

In the table above you see that element 6 has no defender, so that's where you start looking. It turns out that you have to play element 5 first, in order to decoy the Queen into the discovered attack. Here I stopped reasoning, and played Bc2.

Which proved to be wrong. White takes the Bishop with Qxc2. After Nce3+ white plays Nxe3!

I thought that the Bishop was pinned and couldn't move. And indeed after Bc2 that is the case. But after Qxc2 the rook is protected again by the queen!
If you look at element 4, you can see that the bishop protects against the knightfork. It is all there, but you must learn to manage it.

What is needed is a prelimanary move Bxd4. Removing a defender from e3 which happens to be the first piece of a battery too.  I had looked at Bxd4 in the trial&error phase but rejected it because Nbxd4 seemed to give c2 an extra defender. But since the knight can't take at c2 due to Qxd1+ that is nonsense.

It is strange that mapping all tactical elements took me 40 minutes. Even now, while I was able to bring that time to 10 minutes, I wonder why that has to take so long. What on earth am I doing? I don't know. I assume that I can bring down that time to < 3 minutes. So that it fits in an OTB game.

Important not recognized elements make me fail for sure. But recognizing all elements is no guarantee that you will solve the problem at all. It has virtually no effect on my succesrate.

The crucial point is that tiny factors play a major role. The tactical elements interfere with each other and I must learn to mould the combination myself. Unbelievable I am so bad at this!


  1. Simple tactical elements of the type you describe do not help very much beyond the beginner level precisely because they can combine in ways that change their action. Learning tactics is partly about learning all the simpler ways in which simple tactical elements interfere. Far to many ways for a check list I fear. There are also other basic motifs like discovered defence to contend with. Finding tactics is also about spotting a basic idea, and then looking for moves to implement it.

  2. The list is not limited by simple elements. The list is limited by what's in the position. And guess what, that happens to be time and again simple elements. Learning to combine them at will and so moulding a combination is what I want to learn. Investigating them is the start.

  3. I can tell 2 things that helped me to improve:

    a) with lots of repetition of easy (tag sorted) puzzles I brought down my solving time for puzzles (= resulted in better overal thinking speed).

    b) I learned a lot of simple patterns, which I did not know before (resulted in higher accuracy). Example for a valuable simple pattern many of you dont know (because it is rather high rated):

    How to solve puzzles like you showed in this post?
    I believe that you need to have a better speed at patterns.
    There are other puzzles, where you need to know rather unknown patterns like I showed in point b).

    I am not sure how far new patterns (point b) is going to take me.
    Probably there will be a time where I hit deminishing returns from accquiring new patterns.
    Once this point is reached (I believe it is not reached yet), then there will only be point a) to become better.

    Having said that I didnt hit a point where more patterns wont yield in more elo, I am wondering if that holds true for you, too, temposchlucker.
    I very much doubt that you hit a limit here, but rather believe, that despite that you did 100K puzzles, you did not learn even 10K of them.

    But point b) is not the only way to improve, there is still point a).
    By experience (or by feeling) I think that my speed improves better with solving lots of (tag-sorted) easy puzzles. But I am not absolutely shure about this.

  4. P.S.
    by the way: hard to believe my above "pattern" is really not so well known?

    Here I have a puzzle with the same pattern. The puzzle has even less distracting elements in it.
    It is almost painfully obvious that it a check mate needs to be found:

    For its basic simplicity (you hardly get it any more "pure") the rating is therefore indeed a bit lower than usual.
    Still people need on average 27 seconds, and the rating is still close to ~1700 (measured in CT Blitz rating). But other puzzles with this pattern are usually higher rated.

  5. @Munich,
    you never fail to surprise me. I thought after this post nobody would ever think that it is possible to solve the kind of difficulties I encounter with more low level pattern exercises. For the very reason that it are different difficulties.

    You didn't make any progress since end of march. How do you define hitting a wall?

  6. I was wondering about my CT Blitz graph, too.
    However, Richards new stats-features revealed me the following:

    Between 25th of April and 27th of Juli I hardly did any puzzles.

    From 28th Juli till the end of 8th of August I did 1043 Puzzles in Blitz mode.

    I started with 1918 (28.07.2012) and ended with 1904 (08.08.2012).

    So in total I lost 14 points ("= -14 points gained").

    I had 603 duplicates (seen before) and
    I had 440 "first timers" (non-duplicates). (= 42.2% first timers).

    I gained -81,8 rating points with my 603 duplicates.
    I gained +67.8 rating points with my first timers (non-duplicates).

    That means, that if I had only seen the first timers, then my rating would have gone up to 1918+67.8 = 1986 CT Blitz rating.

    There is a flaw in this, though: while going up, the reward for solving a 1800 puzzle is becoming less the higher I climb with my rating.
    If I am rated at 19000 and receive for solving a 1800 puzzle +3.0 points, then I would not receive +3.0 points if I was rated 1950. I would maybe earn only 2.3 points or so.
    So it is difficult to say where my rating would be if I had only first timers in these sessions.

    But what you can see is, that I suffer enourmously by duplicates.
    There is a whopping 149.6 rating difference between my solving duplicates or non-duplicates.

    My old ATH at 1936 was also achieved with a mix of duplicates and non-duplicates.
    Between 25.03.2012-07.04.2012 I did 1055 puzzles (I have chosen these dates to have a similar amount in Puzzles within a similar amount of days, and this time frame includes my very best 1936 rating).

    746 non-duplicate (=70.7% first-timers).
    I gained 76.7 rating points on my first-timers (but with 746 puzzles that is on average 0.1 points per puzzle, compared to my 67.8 rating points with just 440 puzzles, wich is a bit more than 0.15 per puzzle).
    I lost 69.8 points on my 309 duplicates, which is way more desasterous than -81.8 rating points over 603 puzzles.

    Conclusion: I am better on duplicates AND on non-duplicates. Nevertheless my CT Blitz graph does not show it.
    The gap between last seen my duplicates had widened in comparission to my march sessions, so it is likely I must have remembered less than back in March.

    One thought at the end: Till end of April I did train point a) (see previous post).
    From then on I started with puzzles in the medium range. I learned per SRS all my faults. That is point b).
    By feeling both had an effect in my tactical abilities, but only training point a) can be seen in my rating graph.

  7. P.S. I think I mixed up the number of duplicates with non-duplicates:
    So I lost 81.8 points over 440 puzzles. That means I did not improve in duplicates. and my increase per puzzle in first timers should also be about the same.

    But the main result holds true:
    my first-timers bail out my weak performance on duplicates.

    And althouhg I had at the end of march a peak with 1936, my average rating over those 1055 was 1879.6, while my recent rating over 1043 puzzles has an average of 1897. I keep more often close to 1900 than back in March.
    Also my accuracy is 1.9% better Today than back in march. (a slightly better accuracy at a slightly higher average level in rating.)

  8. Still doesn't sound like a competitor for plan c.

  9. It is not that I am resistant in learning from you.
    Actually I put more or less together your findings - and have success with it.

    I got aware that there are next to patterns some guidance rules, too.
    Sometimes I forget about calling it guidance, and refer to it as pattern, too.
    Because a guidance is a pattern, too.
    For instance: "knight on the rim looks dim" is a guidance. For me it is somehow a pattern, too. A knight on the edge can get trapped with a rook or a bishop by taking the escaping squares. Do I have now a pattern or a guidance here if I trap a knight by taking its escaping squares? I dont know. At the end guidance and patterns dont have always sharp lines, and unless we have an easy basic example the differences get increasingly unsharp.
    Is it a pattern if I think I restrict a piece (not only the knight, but in general other pieces, such as the king)?
    Or is it logical reasoning?
    Example: is this puzzle to be found by a pattern or by logical reasoning?

    You could say it is all three: guidance, logical reasoning or pattern.
    Solving it with logical reasoning:
    "The white king is in the corner and cant move. He would be stalemated if he had not another piece.
    If I want to check mate him, then I need a piece that can give him check on a white square. The only piece I have is my knight. So the knight must be the key actor. Either I check mate the black king with the knight, or I can at least get hold on the promotional square.
    There are only 2 possible squares from which the knight can give a check: b6 and c7.
    b6 is blocked by the white king, while c7 is covered by his bishop.
    If my white king would stand on a6 then blacks king would still be stalemated (=cant move). Blacks bishop can only cover one intruding square at the time: either b6 or c7, but not both - unless he can get to d8 or a5. But can he? It takes two moves to do that. If white can check mate the black king in 2 moves, then black has not time for a 2nd move. So best I need to look for a check mate in 2.

    Or is the puzzle simply a pattern?
    I solved it in 9 seconds, but I did use some logical reasoning like above. Obviously with high speed.

    Learning a pattern starts with reasoning. After some time you derive a guidance rule for some typical situations (trapping a knight). After you applied this guidance several times then you have it automated --> it becomes a pattern.
    I understand it now (at least I believe so).

    Everything you do is correct - but(!): you are trying to learn juggling with 5 balls, while you cant even handle securely 3. The puzzle you presented here is a perfect example: too many patterns/guidance/logical reasoning required. You can draw on a piece of paper how to juggle 5 balls.
    You can draw it down in 40 minutes, and with some training in 10 minutes. And even if you can do it in a minute - you are lacking practice. (there are too few difficult puzzles, and if there were more you would need too much time).
    If you then juggle with 1 ball only and say "that is too easy, no use to train 1 ball, because I can allready do this" then you are right.
    Why difficult puzzles? "because we dont fail them" so you say.
    I agree, but I need to say, that we do already fail the 1400s often enough to consider to train them.
    The 1400s are the 2 or 3 balls.
    (measured in CT Blitz rating).

  10. Sorry for flodding, but I just encountered a perfect example:
    (It is rated 1944)

    You solve it with logical reasoning.
    However, look at the tags:
    "x-ray", "pin", "defensive move"

    If you practiced (trained) many x-rays then this is of enourmous advantage in solving it.
    I cant say it is a "pattern" I have seen before. But I cant say, that solving a lot x-rays is useless, too.
    The e&t process is simply much faster with having practices hundreds of x-ray & pin puzzles.

  11. Making the combination work requires thought. You can't replace that by pattern recognition. There are too many possibilities.

  12. @Munich,
    you said:
    After you applied this guidance several times then you have it automated --> it becomes a pattern.
    I understand it now (at least I believe so).

    That's the idea. Allthough I don't know if it is appropriate to speak about a pattern here.

    Everything you do is correct - but(!): you are trying to learn juggling with 5 balls, while you cant even handle securely 3.

    I said it before. You are overimpressed by those 5 balls. I score 50% at the moment with an average of 15 minutes per problem.

    You can draw it down in 40 minutes, and with some training in 10 minutes. And even if you can do it in a minute - you are lacking practice. (there are too few difficult puzzles, and if there were more you would need too much time).

    You are still toying with the obsolete idea that you need lots of patterns. That's not true. I wrote down all the tactical elements during the past 150 problems. I could describe all 2000+ rated problems with only 7 tactical elements. That are 7 patterns you have to learn to recognize. You don't need to learn all variations by heart. That's why it is called recognition.

    At average each problem contains 6-10 patterns which you must recognize. It takes me now 5-10 minutes. That is about 1 pattern per minute. I'm sure I can improve on that. I already improved from 40 to 5-10 minutes.

    Besides the pattern recognition, you must combine the patterns. That is the troublesome part. But I'm making considerable progress. Combining has little to do with patterns but everything with removing of the defenders which prevent the tactical elements from execution.


    I recommend to review these puzzles (and not only them).

    It is not the only one of its kind, there are other "reds".
    Why not easy ones? Because we dont fail them.

    But what if we do, even though we should not? And you do have puzzles, which you failed even though you should not. Not worth to be repeated in an SRS?

    And dont cheat. Please do look at these 2 links. I promise you - this time you will find them VERY interesting! If they dont touch you I shut my mouth afterwards, I promise!

  14. The 17719 took me 2 minutes to solve with my new method.
    The 48113 took me 4 minutes to solve.
    No cheating, no kidding.
    Even after a considerable amount of beer :)

    Doing it right, in a reasonable time. That's what I'm after. I use this method now for a week, so I expect to speed it up considerably.

    First you must do it right. Then you must speed it up.

  15. Well done.
    I have picked these two, because

    a) you failed 2 days before in your most recent blitz session.

    b) they were comparibly low rated (by the way, I recommend "easy" mode instead of "hard" mode if you do blitz sessions. We can discuss why, if you like.)

    c) of those 4 or 5 puzzles, that meat a)+b) I have chosen the 2 puzzles I found statistically highly relevant/ instructive, because they happen often in a game. They have a very typical pattern/ guidance/ reasoning/ whatever.

    But all the others (which you failed 2 days ago) are probably of similar value. Especially for you (not for others) since they are puzzles which you failed. They offer what you train with puzzles above 2000 in rating: they are puzzles you fail. But they are all below 2000, so their statistical relevance is higher (more typical), they are probably even more instructive, and they dont overload you (5 versus 3 balls juggling).

    You might not want to go into the effort to solve 100 puzzles to gain 20 fails or so. But since you do have them now, you can use them and put your fails into a SRS.
    Everything you do with your training you can apply on these failed puzzles below 2000, too: think about the underlying patterns, maybe you could tag them (vote for tags, because that makes you intensively think about the nature of the tactic and hence you probably automate it better?).

    I will shut up now, unless you want to go into this direction and put your failed Blitz puzzles into an SRS set.

  16. @Munich,
    that's great. That means that I can do any low rated problem within a reasonable time without any exercise or rote learning!

    Just by applying the right method! If I can do the same with a succesrate of 90% with 2000+ rated problems within 3 minutes, I am where I want to be.

  17. 17719:
    Problem Blitz Rating:1703.2
    Blitz Av Seconds:00:27

    "The 17719 took me 2 minutes to solve with my new method."

    "that's great. That means that I can do any low rated problem within a reasonable time without any exercise or rote learning!"

    After 2 days this problem should be still in your memory. I forget easy problems after ~ 8 Days ( my solvingtime is 8 days after my 1. attempt ~ as slow as at my first attempt).

    120sec/27sec ~ 4 = 2*2 so your performance on this problem was something like 1700-2*200=1300

    I think that is not that great. You need to speed up another 4 times... then it begins to be interesting.

    I think its interesting ( it makes me think ) that your blitz-rating dont improve. Its common that the blitz-rating is improving the first ~4000 attempts. ( i think the reason for this common improvement might be the fact that you get used to the ratingsystem and the type of problems ). It happend to me too, and i had solved more than 100000 problems at other severs, before i did switch to CT.

    Its interesting for me too, that you seem to forget puzzles very quick.

  18. @Aox,
    I'm not interested in blitz or low level problems at high speed.

    I needed a benchmark from the type "before". So I did a few blitz problems. I didn't make an attempt to learn something from it, I wanted just the numbers. After a few months or so I will take another benchmark sample "after".

    My method contains quite some overkill for low level problems, so time says nothing here. Only succesrate counts.

    When I can do 2000+ rated problems with a succesrate of 90% in under 3 minutes I reached my goal. It means that I can find any tactical shot in an OTB game within a reasonable time.

    My memory is quite good actually. Even so for chess problems. But not when I don't make an attempt to learn something. As was the case here.

    I expect that the benchmark "after" will yield the same figures as the benchmark "before". I would be very surprised if that wasn't the case. That's why I took the benchmark, to be able to trace unexpected side effects.

    For the moment I solely focus on finding the right moves. The succesrate.

    Learning to do it good first, speeding it up later.

  19. I have a sucessrate of 90% ( in Blitz ). Did not help. I did compare high success tacticians with low sucess tacticians: the improvment of both classes of players seems to be the same ( ~0 after ~4000 attempts ).

    My standard did raise from 1900 to 2300, but my tactical power did only change ( after the initial 4000 attempts ) by high speed repetitious pattern training.
    If you can solve 2000 rated problems quicker, then you should be able to solve 1600 rated problems quicker too. So if your new method works, it has to improve your blitz rating too ( you are still below the doublicate reduction effect ).

  20. That might be. Hence the benchmark. But it is irrelevant. If I can do 2000-2600 rated problems in 3 minutes, I expect to be able to do 1600 rated problems in 3 minutes too. Which is sufficient for an OTB game.

    Overtime, redundancy in the method might disappear when experience gives feedback. That might lead to faster times for lower rated problems.