Sunday, November 04, 2018

Realizing us the consequenses of what we see

When I summarize the observations of both me and the commentators on my posts lately, I come to the following:

  • Sometimes we don't look at a certain part of the board
  • Sometimes we do look at a certain part of the board, but we fail to realize us the consequences of what we see
The video of GingerGM seems to rehabilitate the method of trial and error. Are the two points above the reason why our trial and error fails so often?

Sofar, the work arounds that we have invented to solve the two points above, were at the cost of introducing redundancy. That is to say, we put system II at work and we got more correct solutions at a slower pace. The use of extra solving time nullified the rating effect, overall.

But now the problem is formulated so clear, we might find a solution. Both the tree of scenarios and the PLF (PoPLoAFun) system focus on realizing us the consequences of what we see.

I'm pretty sure that we are on the right track. We tend to switch to trial and error automatically when we do not fight against it. Maybe we shouldn't fight it but embrace it, since it seems to be the most natural and energy and mental resource efficient method. Meaning that we must just take away the impediments why trial and error doesn't work for us. The impediments being the two points above.

19 comments:

  1. Momir Raduvic (RoaringPawn) has an interesting Chess.com blog concerning Lines of Operation (very similar to Lines of Attack, but based on military strategy):

    Lines of Operation: A Framework for Strategy Design in Chess, Warfare, Sport Contests

    ReplyDelete
  2. There are many masters at youtube solving tactical puzzles, ChessNetwork, kingscrusher, Daniel Rensch ,
    chessbrah, Huschenbeth... might help to watch them too.
    GingerGM is using Chuzakins HE's without knowing it ( at 1:26 ) then searching for a method to make use of it (at 1:34) combining it with the method of candidate
    moves (Bxf4,...) he already have Bd4 in his mind (at 1:46) but he has a reason to pay attention to Bxf4 first ( looks like the natural move ..) and is then calculating it ( i have the impression he already knows here why its wrong ). He finds a refutation for Bxf4 2:04 and then choses Bd4 because he sees that the knight is "hanging" ( which is an other HE )

    So he is combining the methods in a way, looking for the ( next dominanant ) weaknesss , (next) method to make use of it, (next) candidate moves, calculation, if refutation: next
    It dont get way to complicated because he sees the main weakness , the main method, the most forcing candidate move right away
    he can do it because he has a wonderfull "stack"; he can memorize where he was in the tree of thinking ( all the branches ) can get back with ease and continue

    ReplyDelete
  3. Pseudocode
    for ( next dominant ) HE
      for (next ) Method to make use of it
        for (next candidate move
          calculate ( = some : if i make that then they make that, judgement, reasoning )
          If refutation continue
          If move good exit
        next candidate
      next method
    next weakness
    If good move found look for better move for a while

    ReplyDelete
  4. @ Aox: Nice "Pseudocode"!

    That illustrates a "depth first search" algorithm, which may complement Temposchlucker's "trial and error" process, or a process of "elimination" based on prioritization of critical points. The primary issue becomes determining the order of the ply 1 candidates. I understand the idea of carefully working through all of the HEs in the position under analysis, but I don't recall that Chuzakin gave a prioritization scheme for the HEs, i.e., which HEs are more critical in a given position (or if it is even possible to determine priorities from the HEs alone).

    Here's the first position given by GM Alexander Kotov in Think Like A Grandmaster, pg. 16. I remarked about this position in a comment here:

    Tuesday, December 22, 2015 - An example

    [FEN: 1brr2k1/1b3pp1/pp2pqnp/4N2Q/3P4/1B4R1/PP1B1PPP/4R1K1 w - - 0 1]

    Questions:

    (1) How many candidate moves should be considered in this position?

    (2) On what basis (HEs? Motifs? ???) can the candidate moves be determined?

    (3) Which candidate move is the most "natural" (or "most dominant" based on the HEs actually present in this position) to begin the investigation?

    Total irrelevant - and "funny": I "remembered" that I had commented about this position (based on GM Stockfish's analysis), and searched my "comments" document until I found my comment. But there was an "itch" in the back of my mind that I had also seen some other GM's analysis of this position, refuting (actually extending) Kotov's proposed list of candidate moves. I was pretty sure it was GM Andrew Soltis in one of his books, but I couldn't remember which one. I did "remember" that the "refutation" analysis was at the top of the right hand page in the book. (That may seem "weird" but it is a phenomenon that I have had happen many times in many other situations.) So, I started searching through my Soltis books for the position. I finally went "brute force" and checked diagram by diagram. When I finished searching the two books on my library shelf, I was disappointed NOT to find the position. Surely my memory was not false, was it?!? That was yesterday. This morning, as I got out of bed, I spied a third book The Inner Game of Chess: How to Calculate and Win buried in a stack of books by my bedside. (I'm an avid reader, as should be obvious from my comments.) I found the passage I was looking for on pp. 77-79. The "refutation" WAS on the top of the right-hand page! WHEW! At age 70, I am so glad my memory was NOT playing tricks on me!

    ReplyDelete
    Replies
    1. I guess GingerGM dont know anything about HE's, he just sees an idea which pops up first in his mind and makes enough "beep, beep, beep". When this idea dont work.. next idea.
      Chuzhakin is taking about more or less serious HE's but you still have to judge the HEs by experience.
      In your example b6 and Bb7 are examples for HE's which are easy to see but completly unimportant, there is a knightfork in 2 moves at e7.. unimportant e7 is "save".

      f7 is part of 3! HE's !
      Blacks HE's : Bb3-Kg8 , Rg3-Kg8 , Qh5-f7 , Ne5-f7, h6 is 2 times attacked 0 times defended ( g7 is """pinned"""), Ng6 is weak too, 3 times attacked, 2 times protected. a lot of alarmbells are ringing around the blacks king. The candidate move i see first are all possible takes at g6 . Bxh6 and Bxe6 are candidates too, funny to see how f7 is in the center of most of the HE's but to sac at f7 would be late in my calculations. But i think its possible to "see" that its a good move without calculations, because it brings the next piece into play , the Re1, removes a defender of Ng6...
      I wonder how a gm would solve this puzzle

      Delete
    2. @ Unknown: "I wonder how a gm would solve this puzzle."

      I have 3 GMs "opinion" regarding this position.

      I. GM Alexander Kotov, Think Like A Grandmaster, Diagram 1, pg. 16-17:

      Candidate moves:

      26. Bxh6
      26. Nxg6
      26. Ng4, intendiing 27. Nxh6+

      The MASTER [according to Kotov] playing White actually played 26. Bc3?, which cost the game after 26. ... Nf4 27. Qg4 h5 28. Qd1 h4 - White resigned.

      Kotov asserted that White should have played 26. Ng4 Qh4 27. Nxh6+ Kf8 28. Qxh4 Nxh4 29. Nxf7 Kxf7 30. Bxe6+ Kf8 31. Rg4 Nxg2 32. Bb4+ Bd6 33. Bxd6+ Rxd6 34. Bxc1 Nxe1 35. Bxb7 and "White would win."

      II. GM Andrew Soltis, The Inner Game of Chess: How to Calculate and Win, pg. 77-79:

      "After 26. Nxf7, it's easy [for a GM!] to see that if the knight is taken [26. ... Kxf7 or 26. ... Qxf7], 27. Bxe6 wins. So the real calculation involves 26. ... Bxg3 and then 27. hxg3 or 27. Nxh6+ Kf8 (not 27. ... gxh6 28. Bxe6+ Kf8 29. Bxh6+) 28. hxg3. This appears to be at least as promising as the line endorsed by Kotov (26. Ng4 Qh4 27. Nxh6+ Kf8 28. Qxh4).

      "And why didn't Kotov mention it at all? Probably because candidate moves are a lot more personal than anyone likes to admit—and Kotov may have missed 26. Nxf7 himself. What Garry Kasparov considers "natural" may be something that would never occur to Viswanathan Anand or Gata Kamsky. But for each of the remaining diagrams in this book YOU SHOULD TRY TO PICK OUT THE MOVE THAT SEEMS MOST NATURAL TO YOU AND THEN FIND AT LEAST ONE OTHER CANDIDATE."

      III. GM Stockfish 9 64 gives the following analysis:

      1.Nxf7 Rxd4 [1...Rc5 2.Rg5 (2.Nxh6+ Kf8 3.Qxg6 Qxg6 4.Rxg6 Rh5 5.Rexe6 Bxh2+ 6.Kf1 Rxd4 7.Bc3 Be4 8.Rxg7 Bd3+ 9.Ke1 Re5+ 10.Rxe5 Bxe5 11.Rg8+ Ke7 12.Nf7 Bh7 13.Nxe5 Re4++- 8.66/21 ) 2...Kh7 (2...Qxf7 3.Bxe6 hxg5 4.dxc5 Bd5 5.Bxf7+ Bxf7 6.Bxg5 Rf8 7.cxb6 Ne5 8.Qh4 Nc6 9.Qa4 Ne5 10.Qd4 Nc6 11.Qc5 Nd8 12.Be7 Re8 13.Qc8 Bd6+- 10.62/21 ; 2...Nf4 3.Bxf4 Qxf4 4.Rxc5 Qxh2+ 5.Qxh2 Bxh2+ 6.Kxh2 Rf8 7.Rc7 Bc8 8.Ne5 Kh7 9.Bc2+ Kg8 10.Ng6 Re8 11.Rxc8 Rxc8 12.Ne7+ Kf7 13.Nxc8 b5+- 55.60/21 ) 3.Nxd8 Rxg5 4.Bxg5 Qxg5 5.Qxg5 hxg5 6.Nxb7 e5 7.Bc2 Kh6 8.Nd8 Ne7 9.Nf7+ Kh5 10.Nxe5 Nd5 11.Bd1+ g4 12.Bxg4+ Kh6 13.Nc6 Nf6+- 10.45/21 ; 1...Bxg3 2.Nxh6+ Kf8 3.hxg3 Rd5 4.Bxd5 Bxd5 5.Bg5 Qxd4 6.Be7+ Ke8 7.Qxg6+ Kd7 8.Qf7 Qc4 9.b3 Qc2 10.Ba3+ Kc6 11.Rc1 Qxc1+ 12.Bxc1 gxh6 13.Qa7 Rc7+- 9.34/21 ; 1...Qxf7 2.Bxe6 (2.Rxg6 Qc7 (2...Rd5 3.Bxd5 Bxd5 4.Bxh6 Bd6 5.Bxg7 Qxg7 6.Qh6 Bf8 7.Rexe6 Be4 8.Rxg7+ Bxg7 9.Qg5 Bb1 10.Rxb6 Kh7 11.Qh4+ Kg8 12.Qg5 Re8 13.h4 Be4+- 57.73/21 ; 2...Bd5 3.Bxh6 Bxb3 4.Rxg7+ Qxg7 5.Bxg7 Kxg7 6.Qg4+ Kf8 7.axb3 Rd6 8.d5 Rxd5 9.Qxe6 Rcd8 10.Qf6+ Kg8 11.g3 Be5 12.Rxe5 Rxe5 13.Qxd8+ Kf7+- 10.09/21 ) 3.Rexe6 Qxh2+ 4.Qxh2 Bxh2+ 5.Kxh2 Kh7 6.Bxh6 Rd7 7.Bxg7 Rc6 8.Be5 Rxe6 9.Rxe6 Bd5 10.Bc2+ Kg8 11.Rxb6 Rb7 12.Rg6+ Kf8 13.b3 a5+- 8.79/21 ) 2...Qxe6 (2...Bxg3 3.Bxf7+ Kxf7 4.hxg3 Rd5 5.Qe2 Rd6 6.Bb4 Rf6 7.Be7 Bd5 8.Qh5 Bc4 9.Bxf6 gxf6 10.Qf5 Rd8 11.Rc1 Rxd4 12.Rxc4 Rxc4 13.Qd5+ Ke8+- 10.78/21 ) 3.Rxe6 Bxg3 4.Qxg6 Bxf2+ 5.Kxf2 Rc7 6.Kg3 Rf7 7.Bxh6 Bd5 8.Rxb6 Re8 9.Bf4 a5 10.Rb5 Be4 11.Qd6 Rf6 12.Qc7 Rc6 13.Qd7 Rg6++- 9.61/21 ; 1...Rd5 2.Nxh6+ Kf8 3.Bb4+ Ke8 4.Qxg6+ Qxg6 5.Rxg6 gxh6 6.Rgxe6+ Kd7 7.Bxd5 Bxd5 8.Re7+ Kc6 9.Rc1+ Kb5 10.Rxc8 Bf4 11.a3 a5 12.Bc3 Bd6 13.Rh7 Bf4+- 14.37/21 ] 2.Nxh6+ Kf8 3.Rxg6 Qh4 4.Qxh4 Rxh4 5.g3 Rd4 6.Bc3 Rxc3 7.bxc3 Re4 8.Rxe4 Bxe4 9.Rxe6 gxh6 10.Rxe4 Bc7 11.Re6 h5 12.Rc6 Bd8 13.Rh6 Ke8+- 9.32/21 Line

      Errors may have occurred and, if so, are solely because I transcribed from descriptive to algebraic notation (Kotov) and from figurine to algebraic notation (Stockfish).

      My suggestion is to figure it out for yourself, using whatever "system" works for YOU. However, I know most players (including ME) will NOT do this kind of work - which is why we don't improve very much!

      BTW, If anyone can find the game that this position was taken from, I'd appreciate that information.

      Delete
    3. [Event "URS-ch09"]
      [Site "Leningrad"]
      [Date "1934.??.??"]
      [Round "?"]
      [White "Riumin, Nikolay Nikolaevich"]
      [Black "Belavenets, Sergey Vesevolodovi"]
      [Result "1/2-1/2"]
      [ECO "B11"]
      [PlyCount "69"]
      [EventDate "1934.??.??"]
      [EventType "tourn"]
      [EventRounds "19"]
      [EventCountry "URS"]
      [SourceTitle "URS-ch"]
      [Source "ChessBase"]
      [SourceDate "1999.07.01"]
      [SourceVersion "1"]
      [SourceVersionDate "1999.07.01"]
      [SourceQuality "1"]

      1. e4 c6 2. Nc3 d5 3. Nf3 dxe4 4. Nxe4 Nd7 5. Bc4 Ngf6 6. Ng3 e6 7. Qe2 Be7 8.
      O-O O-O 9. d4 Re8 10. Bb3 Bd6 11. Ne5 c5 12. c3 cxd4 13. cxd4 b6 14. Bg5 h6 15.
      Bf4 Qe7 16. Qf3 Ba6 17. Rfe1 Nf8 18. Nh5 Nxh5 19. Qxh5 Bb7 20. Qg4 Qf6 21. Re3
      Rac8 22. Rae1 Red8 23. Rg3 Bb8 24. Bd2 a6 25. Qh5 Ng6 26. Bc3 Nf4 27. Qg4 h5
      28. Qd1 Bxe5 29. Rxe5 Qxe5 30. dxe5 Rxd1+ 31. Bxd1 g6 32. Kf1 Bd5 33. Bb3 Bxb3
      34. axb3 Nd5 35. Ke1 1/2-1/2

      Delete
  5. The candidate moves in a tacticsl position are 1.checks, 2.captures and 3.threats, so Nxf7 is a candidate move. In quiet positions its harder to tell which moves are the candidates, there they are dependend to the chosen strategies , a Petrosian would try to reduce our activity, a Tal would try to increase his activity...

    ReplyDelete
  6. More from the GingerGM:

    Tactic Training #3 Chess is 99 percent Tactics!

    The second puzzle is interesting because of the "sting" at the end. Black can gain a piece, but in so doing, loses control over the mating squares. He states later in the video that he "panicked" on this one because of the timer. Sounds familiar to me!

    The fourth puzzle is interesting. He begins with a "vulture's eye view" and (correctly) assesses that the Black King is "in the box" (back rank mate possibility) AND he immediately "sees" the vulnerability of the c1 square, skewering the Black Queen and taking the Black Rook. Then he goes astray, because he is "not so sure about that." He then switches to another alternative, threatening the back rank mate, and spends quite a long time on it, going deeper and deeper until he is convinced that it just doesn't work. He then goes back to the skewering idea and finally "sees" WHY it works. It demonstrates that even GMs (while working via trial and error and using the process of elimination can sometimes initially wander down the wrong path. The critical thing is that as soon as he "sees" that he is on the wrong path, he immediately abandons it and returns to an earlier idea that had been deferred. As Aox noted above:

    "It don't get way too complicated because he sees the main weakness, the main method, the most forcing candidate move right away [although he does not investigate it deeply at first]; he can do it because he has a wonderful "stack"; he can memorize where he was in the tree of thinking ( all the branches ) [and] can get back with ease and continue."

    ReplyDelete
    Replies
    1. At puzzle #4 he used the method of "combination of tactical weaknesses". The weaknesses are b6 ( unimportant ) d5 ( unimportant ) weak king, weak backrank and the skewer at c1 ( all HE's of course ). During his ""wrong"" exploration of other move sequences he did improve his understanding of the c1 square skewer and moves like Qf5 protecting c8 ( which is not a HE ;)

      So the pseudocode before was a little to simple

      He is changing the model of "what are the used tactical weaknesses" on the fly

      This thinkingprocess is mainly the standard thinkingprocess in any position, you look for the main features of the position and generate the best line from it.


      generate a hypothesis about the used weakness
      generate candiate move
      explore lines and gain insights
      while not finished
      modify hypothesis about the used weakness
      generate / modify candidate moves
      explore lines and gain insights
      end

      thanks to his great memory he does not repeat anything, thanks to his great experience he explore each line to the point where he can tell if the goal is reachable or not

      Some trainer in the net said he told his student to learn chessgames by heart and that the ones who did it, did improve ( a lot ). The skill to be able to memorize games with easy should be helpful for such mindwork.

      Delete
    2. In Tactic Training #4, GingerGM gives his set of "rules" for solving puzzles:

      (Tactic Training #4 Rules for solving puzzles like a Grandmaster)

      When should you search for a tactic? Not on every move!

      1. When ‘Instinct’ kicks in — Smell it!
      2. At critical stages of the game – slow down and spend more time calculating!
      3. In complicated positions
      4. When you have the initiative
      5. There is a clear weakness in your opponent’s position

      When you think there may be a tactic, what should you do?

      1. Analyze “Candidate’ Moves – Look for ONE ‘forcing’ move (CCT) that can take advantage of the position
      2. Use the ‘Process of Elimination’ – stick with one move at a time until you can eliminate it or be sure that it works
      3. Double check and triple check any move that appears to ‘work’
      4. If need be, look for a better move. (A modification of advice attributed to Dr. Lasker.)

      The basic ideas seem very similar to GM Tisdall's "variation processing".

      A strange experience: A couple of days ago, I was waiting in my car for my wife to finish shopping. I decided (on a lark) to try to remember all of the piece/Pawn locations in the Kotov position. I hadn't tried to memorize it or even to remember it until that moment, but I figured it would be a good exercise. To my surprise, I was able to remember everything! When I got home, I wrote down all of the piece/Pawn positions and then checked against the book. I got everything exactly right! That is certainly a first (one and only) for me. Somehow, that position made a deep impression on my System 1. If only I could replicate that experience over and over again!

      Delete
    3. He dont follow his own rules ;) he is looking for tactical weaknesses first as we can see at the next puzzle for example: https://youtu.be/Cp0Ozdc3_h0?t=368"
      First thing he sees are the weaknesses Bb4-Qd2/Ke1 (pin) and Bd5 ( 2 attacker , 1 defender ) uses the method "removing of the defender" candidate move Bxc3 analysis -> Bc6+ ( a typical titfortat reply )
      now he is looking for the next candidate move 6:43 for the same set of weaknesses
      and problem solved.
      what they say they do is not allways what they realy do.

      Delete
    4. what gingergm calls "smell" is "patternrecognition of a tactical weakness"

      Delete
  7. I mentioned GM Tisdall’s “variation processing” method previously in a comment:

    Wednesday, December 28, 2016 Modeling combinations

    Aox wrote:

    He don’t follow his own rules ;)

    what they say they do is not always what they really do.

    Some would think it is hypocritical to give advice that you yourself do not follow. Perhaps it would be more charitable (giving the benefit of the doubt) to look at it from the System 1/System 2 perspective.

    System 1 proposes; System 2 disposes. Simply put, our subconscious “throws up” ideas that are based on pattern matching. System 1 does not just “throw up” ONE idea; it brings up ALL ideas which match the situation based on prior experience. It is then the task of System 2 to “make sense” of all the ideas. The way that System 2 does this is to construct “stories” (narratives) which match between the actual situation and the retrieved “patterns.” But there are problems with this two-system approach. System 1 focuses attention based on “fight or flight”; it makes no judgements concerning which idea is most appropriate. System 2 (being “lazy”) tries to create a coherent “story” out of whatever it has been told by System 1. It is this “story-telling” which is put down in writing by the chess authors.

    They do not intentionally tell “stories” that are not true. They simply write down what their System 2 has fabricated as the “process” that they follow. That they do not follow this process is evident when comparing their “process” list with their annotations and verbalizations.

    Paradoxically, the higher the playing skill level, the more likely it is that the player does not have conscious access to the actual process used, and thus is forced to “make up a story” that “fits” conventional wisdom. The lower the skill level, the more likely that the logical “step by step” process has been conglomerated from things that the person has seen others espouse, and the less likely it is that the “process” bears any resemblance to reality. Thus, we adults “study” and try to memorize less than optimal processes as lists, and fail to realize that is not the way higher skilled players "think" in reality.

    ReplyDelete
  8. Hmm.. lets try to summarize a little

    the thinkingprozess is: patternrecognition to find the dominant weakness(es) , stepping through candidate moves, calculating, reasoning, judging

    so what do we need
    -board vision
    as tempo said :
    "Sometimes we don't look at a certain part of the board
    Sometimes we do look at a certain part of the board, but we fail to realize us the consequences of what we see"
    they are aware where the pieces are and what they ( potentially) do
    -memory of (already calculated ) lines
    -memory of (already analysed ) positions ( related to board vision )
    -memory of own thought process and won insights
    -hyperfast patternrecognition
    -knowledge ( to be able to judge the final positions )

    And i guess a good training should adress exactly these points

    ReplyDelete
    Replies
    1. -memory of (already calculated ) lines
      -memory of (already analyzed ) positions ( related to board vision )
      -memory of own thought process and won insights
      -hyper fast pattern recognition

      I'm not too worried about these four points above. It is typical system I stuff. System I will works its miracles when system II asks for it.

      Delete
    2. I am of course, well aware that you don't believe in miracles.

      Delete
  9. Example for a "correct" solution of a puzzle

    https://lichess.org/training/61614

    The solution is NOT! Nc3

    but:

    Weaknesses: Nightfork at c7, unprotected Nc4
    Method: opening line to c4
    Candidate moves Nc3,Nc1
    Refutation of Nc1: b5

    + other things which had been in our mind and where wrong, unimportant and so on

    ReplyDelete