Wednesday, March 09, 2016

Continuing with the thought process

Bad board vision is one of the two weaknesses I discovered during analysis of my blitz tactics at CT. We started a campaign to work that out, and a set of exercises emerged. The first test revealed two things: some exercises work, while others are too complex. I hope that Lain can find the time to tweak his algorithms, so that the exercises comprise the whole area of mate in one. Thus far, he has done an excellent job.

So the work is set out for us, and it will take some time to improve on all improvable exercises AKA the salt mines. It doesn't make much sense to talk about the work in stead doing the actual job. In stead I will talk about the thought process, which is the next thing that needs development, when I rest for a moment from salt mining.

The jury is still out.

28 comments:

  1. I have noticed that I can calculate 3-5 moves variation... only to miss an important second of third key moves (by my opponent or myself). It is really hard to play correctly when you cannot visualise the pieces and miss important refutation moves. And it starts with the lack of noticing ALL the legal moves. That's why I think we should create a tool - find all legal moves (FALM). Most often we (at least me) miss the moves we do think they are "not legal" (very unreasonable ones). IT has been (and still is!) the case with me when I have to improve at find all checks. Some checks moves seem to my mind as "ridiculous" moves (a very bad ones) EVEN if I know I just have to list all of such "ugly moves".

    Have you got the same feeling NOT to play (click) the moves you see as "ugly ones"? At present it is my biggest drawback and if I could break this disadvantage - I would be able to improve at FAC. Any ideas and comments?

    My score at FAC (Find all checks) - initial 33 and present 42 (copied from the table). Let's compare with FAES (Find all escape squares) - intial 49 and present 69. Do you see the big room for improvement? IF I improved FAC to the level of FAES, I would improve M1-e (Mate in 1 easy) from 35 to 40, and probably M1-h (Mate in 1 hard) from 7 up to 12. At least it is my estimation. What do you think about it?

    ReplyDelete
  2. Weeel Thinking Process..
    1) why
    2) how

    1a) Time saving
    1b) preventing blunders
    1c) every position becomes a type of "board vision" exercise because you do always the same. Such a habbit becomes automated

    2) to write a pseudocode which can be used at every position,
    good example for that are the courses of GM Smirnov.

    ReplyDelete
  3. I must generate IAPP/IABP problems only? they are all FAC-like?

    ReplyDelete
    Replies
    1. I lost track about the IAPP and IABP discussion so maybe what i say now is already clear.
      You must not make a process which is already too complex even more complex.
      Let me explain
      If we have a problem with FAC then it dont make it easier if you create pozzles where the user has to Find all Check which satify even an other condition too. That makes the task simply more complicated.

      What does work is : you could make the puzzles more uniform
      you could ask for "Find all Checks" and the puzzles do satify already an other condition already. Its nessesary that the puzzles satisfy more conditions to make the task more unify and must not ask the user to make even more thoughts, that would not work.
      So sorry if i missed that you already discussed that

      Delete
    2. I improved from 11 to 20 at FAC, and I feel that further improvement is possible with the subtask exercises we already have. It's just a matter of doing the exercises. I don't need extra exercises for that. If you want to get better at M1, IAPP and IABP are critical exercises. Which incidentally happen to be necessary for any tactic as well.

      Delete
  4. So these puzzles are like : find all attacking and all protecting pieces of "d4"?
    That sounds improvable

    ReplyDelete
  5. These positions are right?

    http://puu.sh/nBzps/56f90c2f5a.txt

    I will extract more later if they are.

    ReplyDelete
    Replies
    1. 5R2/5q2/2bp1B1N/8/R1P2k2/3P1N2/4p1PP/2K5 w - - 0 1 #protected d6c6 #blocked c6f7
      This position is too complicated. The problem is the discovered check c5+ followed by Bxa4. In general, all positions where a protected checking square equals the square with the attacker should be avoided. For both #protected and blocked.
      8/2p3Q1/p1RNKP2/n2p1p2/1P1k1p1R/1p3Prb/3PPP1b/2N5 w - - 0 1 #protected d5a5h3a6g3
      Ditto: f7+ Rxg7
      2b4Q/4p1K1/4pr2/2P1k3/2Pn1R2/3P3p/7B/8 w - - 0 1 #blocked f6d4
      This position is too complicated too. No discovered checks would avoid these positions (and the positions above).
      b2R1B2/3p3r/nN6/1R6/rn1kp3/P3pB2/1PQ5/K3N3 w - - 0 1 #protected h7a6e3e4b4a8a4 #blocked a8b4
      OK
      4n3/2rR2B1/3b1q2/1pP5/1P1kNQ2/5BK1/N1pnP1b1/3R4 w - - 0 1 #protected e8f6
      OK
      3Rn3/1np3B1/b2b1qP1/p6R/rP1kNQK1/3p2P1/1P3P2/8 w - - 0 1 #protected c7e8b7f6
      OK
      6Q1/2p1kr2/3R4/8/1R2K3/B3p2B/8/8 w - - 0 1 #protected f7 #blocked f7
      OK
      B4n2/5n2/1p4N1/1p1K1N2/2pP2p1/5k2/1R3b1Q/8 w - - 0 1 #protected c4f2f7g4 #blocked c4f2g4
      OK
      1b1R1N2/6p1/1Np2pP1/2p1k3/1pp1BnPK/6n1/q2QRrPB/8 w - - 0 1 #protected c6f4f2g3c5b8a2b4 #blocked f4g3c4
      Too complex.

      If you leave out all positions where white can give a discovered check, then we get rid of most positions which are too complex. As far as I can see now.

      Delete
    2. 3Rn3/1np3B1/b2b1qP1/p6R/rP1kNQK1/3p2P1/1P3P2/8 w - - 0 1 has a discovered check: Ng5+, but the king can only escape. Should I remove it too?

      Delete
  6. In the same position, 3Rn3/1np3B1/b2b1qP1/p6R/rP1kNQK1/3p2P1/1P3P2/8 w - - 0 1, Nc5# is a discovered check and mate, that's sufficient reason to consider the position as complex?

    I'm currently excluding all discovered checks of any kind.

    In 4n3/2rR2B1/3b1q2/1pP5/1P1kNQ2/5BK1/N1pnP1b1/3R4 w - - 0 1 a similar discovered check that is also mate happens: Nxd6#... And I'm not sure if I must keep or not these positions.

    ReplyDelete
  7. Yes both positions are too complex.For mate there are other exercises. This is about training the protector and the blocker in isolation as much as possible. When in doubt: dump it. Since we will have the same doubts when exercising. M1-h is already full of doubts, we don't need those when training subtasks of M1-h.

    We are already working at the edge. We already are piling up these exercises for identifying the protector and the blocker on top of FAC. Which means that only when you fully master FAC, it make sense to do these exercises you are developing now.

    ReplyDelete
    Replies
    1. Tempos said : "We already are piling up these exercises for identifying the protector and the blocker on top of FAC."

      "on top of" is not good.
      i think such tasks could be trained seperatly: Find all not protected pieces, find all not protected squares
      blocking could be trained by a given check and : Find all blocking squares, find all blocking pieces.

      "AND" should be done by the puzzles and not by the user. Find a check AND the chack can be blocked is a complex task for the user instead : there is already a check in the puzzle AND the user has to find the blocking possibilitys is way more simple.

      But i repeat myself...
























      Delete
  8. I am not sure if this helps, but I find this article really interesting:
    http://www.bakadesuyo.com/2016/03/expert/

    The title is - The 4 Rituals That Will Make You An Expert At Anything


    -------------------------------------------------------------------------------------
    The 4 Rituals That Will Make You An Expert At Anything

    We hear a lot about “10,000 hours” being what it takes to become an expert. But the majority of people totally misunderstand the idea.

    So I decided to go to the source and talk to the guy who actually created the theory.

    Anders Ericsson is a professor of psychology at Florida State University. His wonderful new book is Peak: Secrets from the New Science of Expertise.

    So what does everybody get wrong? 2 things.

    First, the “10,000 hour rule” is not a rule and it’s not an exact number. The amount of time varies from field to field. It’s an average. But it’s always a lot and more is better. Here’s Anders:

    In most domains it’s remarkable how much time even the most “talented” individuals need in order to reach the highest levels of performance. The 10,000 hour number just gives you a sense that we’re talking years of 10 to 20 hours a week which those who some people would argue are the most innately talented individuals still need to get to the highest level.

    What’s the second mistake? Becoming an expert is not merely doing something over and over for 10,000 hours. There’s a right way — and an awful lot of wrong ways — to spend that time.
    -------------------------------------------------------------------------------------

    What do you think about it? Let me know what things you can agree and which ones you cannot! Thanks in advance!

    ReplyDelete
  9. I would like to know if there is an excercise "Is the King in a Check" (IKIC). If you know where I could test my skills at this trivial excercise - just let me know. I wonder how well I could manage this task. Sounds very simple, but as far as I know - masters and granmaster are able to recognize the position in a less than 0,3sec! It must be really lightning recognition! :)

    ReplyDelete
  10. What are your thoughts and raw conclusions related to "Salt mine experiment"? Can we draw any conclusions, suggestions or recommendations?

    Or we have to work on these puzzles (all of them or just the selected ones?) much more time - to draw any reasonable conclusions or suggestion?!

    ReplyDelete
    Replies
    1. So far, it seems possible to improve at certain subtasks. Yet the improvement doesn't seem very decisive. It seems that improving the subtasks, helps to improve the main task. Yet I'm underwhelmed.

      I haven't trained M1-h in a while, and I seem to be back at square one. Even though I trained a lot of FAES-h and improved in that subtask. I haven't seen long lasting results yet.

      FAC and FAES seem critical subtasks.

      I did a few blitz exercises at CT and scored terrible. So training the subtasks haven't lead to an instant improvement in that area.

      The raw conclusion is: it works somewhat. But to make it really work, it must be combined with using a thought process. I was actually diving in that (the TP) right now. But I continue salt mining too.

      Delete
    2. you cant judge your performance at ct on just 16 samples. You need to do a few hundred? puzzles to get used to this type of task again.

      A simplified thinking process can helpt to improve at these board vision exercises. You may learn it doing FALM ( find all moves at Lukas chess ), you are forced to step through the pieces starting with the king down to the pawns

      I have the "impression" that the h versions dont work as good as the e version. I suspect that a "single piece aura vision" is hard. The easy versions teach to see groups/configuration of pieces ( chunking ). The effect of speeding up is done by learning such chunks and their effects on the subtask.

      I share your conclusion though. I did work on this idea a while ago and had to give up because of an overdose of salt. But i see no other chance to improve the fast processes / vision and i think that without an improvement in these past processes a decisive improvement in tactics/chess is not possible.

      Delete
    3. The hard tasks seems to be less improvable than the simple tasks. Maybe that's why we called them hard in the first place. I can no longer avoid your conclusion that we must simplify the tasks as much as we can. Otherwise the Troyis-effect soon evaporates.

      Delete
    4. There is a lot of overlap in the exercises. For instance FAES-h makes that you look for protectors of potential escape squares too.

      Further I don't quite understand why FAES-e and FAES-h feel as if it are totally different exercises.

      Delete
    5. We probably must leave the database which is based on composed problems alone for the moment, and focus on more simple positions from real games.

      Delete
  11. What happened to http://chess.emrald.net/?

    ReplyDelete
  12. My reflexion after I did another bunch of puzzles (1421 puzzles at FAC NO QUEENS - NO QUEENS: 43.40 CPM - my best score is 46 so far).

    What I realized is I NEED is to IMPROVE all the tasks individually. After that I can try to break the limits. Why some of these tasks seem 'non-improvable' to us? I have the strong impression it is because we have not PERFECTED all of these and that's why we cannot be very flueant at the "BIG ONE" (FAC and FAC hard respectively as much as M1e and M1h). They are simply TOO complex.

    Is there a chance (hope) to provide ALL the (remaining) tasks as the separate tools?

    What I mean?

    FAC - ONLY pawns
    FAC - ONLY Bishops
    FAC - ONLY Knights
    FAC - ONLY Rooks
    FAC - ONLY Queens

    And it would be VERY helpful to have these tasks ONLY for the one colour (only white gives checks) and AFTER we master this one - the opposite colour (only black gives checks). And finally - both colours give checks. It is hard work to do it, but if you could construct (provide) such tools... I would be the first to test it. I simply DO NOT believe we are so weak at so "simple" tasks. However I notice we have not decomposed these to the simplest parts (atoms) - and that's the reason we cannot SIGNIFICANTLY improve at the "BIG ONES" (FAC and FAC-h; M1e and M1h).

    To give you some motivation to create these puzzles - after the tools will be created - I am going to try to break my best score (at FAC close to 48-50 and FAC NO Queens close to 52-55). It is a VERY HARD and demanding work (most of you probably know how much), but we have to find the atoms of chess tactics (attack the piece in ONE move) as much as the invisible relationships between the pieces. There is NO OTHER WAY (except improving thinking process).

    What do you think about my idea and conclusions?

    BTW. Could you set the minimal score to achieve at any of the salt minining tools - to break (achieve) and say - "yes, Tomasz you have improved the X-tool to the level we can say it is a significant progress/improvement"? I do not mean the BIG ONES, but the additional ones (after that I will try to break "the BIG ONES").

    ReplyDelete
    Replies
    1. Yes, we have to break down the subtasks even further. Only the lowest level works, and only after mastering these lower levels we can work at the higher levels with a reasonable chance of success.

      I will break down the subtasks to it's atoms and provide a list.

      Your suggestion to make separate exercises like
      FAC - ONLY pawns
      FAC - ONLY Bishops
      FAC - ONLY Knights
      FAC - ONLY Rooks
      FAC - ONLY Queens
      is not so much work. The problem is, that the problem sets will be very small. If Aox can use his tools to create a much bigger fenarray for FAC, I can use that to create bigger problem sets for the separate FAC exercises with less material.

      Delete