Tuesday, December 05, 2017

Wrapping my head around the tree

So far, I haven't been able to internalize the knowledge which is confined within the tree of scenarios.  Several methods were tried, but none of them stood up to the challenge.

I wondered why that is, and while pondering about it, I came up with an analogy. Quite some time ago I was a member of an economy study group who studied the book Progress and Poverty of Henry George. It took me about six years before I was able to wrap my head around the matter that was described in the book.

Once that was the case, it usually took me only seconds to dissect a complex economic problem and to describe it within the context of the book. I realized that despite all my efforts so far, I still am not able to wrap my head around chess tactics though.

I can investigate parts of it well enough if I take my time, but there is no coherence in my tactical chess knowledge. I feel that I'm not at all that far off, though.

So I'm rereading my posts from about April to now, and extract the essence of it. I'm going to reshuffle the knowledge until the core of it sinks in. Until I'm able to wrap my head around it. From there, the magic of the unconscience is supposed to kick in.


11 comments:

  1. The secrets of the top player : https://www.youtube.com/watch?v=qeQ_szmbb7k

    ReplyDelete
  2. For me, it's about "over-learning" a position. I took out my book "Perfect Your Chess" by Volokitin and Grabinsky. "Find the Win", problem #1 is this game http://www.chessgames.com/perl/chessgame?gid=1311226 White has just played 32.Kg2, now find the win for Black.

    Okay, here's the thing though, try blindfolding the position, and then try and solve the answer.

    The tactics in a position are less important than blindfolding the position. A tactic is one move, so you should be able to see those by now. A combination often contains a tactic a move or two out from the initial position. So, the key is to analyze variations. It's fine to say memorize patterns, but those should be mostly known first, anyway.

    If you play bullet-chess, or even blitz chess, say a week or two after a blindfold exercise, you will notice that your results are better (in my experience) even though you might not know why, because you say to yourself that your chess understanding hasn't gotten better.

    It's very important to know "in your head" what is going on on the board, and not just solely determining what is going in by looking at the board. Without an effortlessness of holding variations in one's head, I don't see how one can make significant jumps in strength, unless it involves a field of knowledge, such as learning opening variations or endgame technique from such a book.

    GMs go over variations very quickly. I find I have gotten a bit quicker at following variations, instantly noticing why some side-move doesn't work. Sometimes, you can even follow what is going on on the board intuitively, while playing at bullet speed. Keeping up involves your knowledge of the board, understanding where everything is makes it easier to intuitively understand what is going on.

    ReplyDelete
    Replies
    1. linuxguyonfics said : It's very important to know "in your head" what is going on on the board, and not just solely determining what is going in by looking at the board.

      After 1 year of learning strategy i am back at tactics again and experience again that one of my problems is the lack of awareness:

      Please try this
      Awareness Test

      I might be aware of a main idea of a tactical puzzle but i am so concentrated on the few things i see that i cant see an other suddenly important aspect and.. blunder. Given enough time i can find such new ideas and treat them correctly. Here we see the bottleneck of the STM (Short Term Memory) for solving problems. The traditional method to overcome this difficulty with the small stm ( size of stm ~= intelligence ) is to create chunks. chunks are informations stored in ltm (long term memory).

      In computerlanguage : LTM is the superfast memory at the processor whils LTM = memory on the slow harddrive . Better to have a 64 bit processor than only an 8 bit processor

      I think we get aware of things ( = see things ) via our memory so i guess improvement in chess is more or less identical to improvement in chess memory. To play good you need to be able to store ideas about the game quick into your memory ( see it) and be aware of it ( keep it in sight ) as long as necessary.

      So i suppose that all type of chess memory training should be beneficial for the play: playing blindfolded, learning games (with variations?) by heart, board vision exercises,, solving many puzzles of one typ of tactical pattern and so on..

      Still..
      we need a proof that the exercises we do are realy helping because we like to fool ourself. I often try to help weaker chessfriends with their game and explain for example that they should not move a piece twice in the opening. The reply is always the same : " i already know, BUT... ". They know it "better". Kids can easier accept that someone else knows it better.
      Well: so im not sure if my ideas of chessimprovement are right, its necessary to test them and THE method to test improvement in chess is CHESSTEMPO. If you cant improve your (blitz)tactic rating there with your method then.. your method is junk.

      Delete
  3. A computer did teach itself chess for 4 hours and then did beat stockfish ( the best engine programed by humans ) in a 100 game match with 28 wins and no losses. Soon human intelligence is not needed anymore
    https://www.chess.com/news/view/google-s-alphazero-destroys-stockfish-in-100-game-match

    ReplyDelete
  4. PART I:

    @ Aox:

    I think there's a typo in here: In computerlanguage : LTM [STM?] is the superfast memory at the processor while LTM = memory on the slow harddrive.

    I once was an in-house Navy consultant to "scientists" from Oak Ridge National Laboratory who were contracted to evaluate why applications on a Navy computer system were running so slow. I knew the hardware architecture, and they did not. The system performance diagnostics showed that the applications were I/O bound: the CPU was in idle mode for over 90% of the time. The natural (and incorrect) conclusion was that faster block controllers and/or faster hard drives would speed up the applications. I tried to tell them that their conclusion was wrong. The computer CPU had a built-in governor which only allowed one out of every three CPU cycles (the memory bus) to be "stolen" by any I/O device, including the block controllers connected to the hard drives. Even though the CPU was idle, it still limited the quantity of I/O that could be accomplished. The existing block controller and hard drives were not maxxed out, but the CPU wouldn't let them pass more data through the channel to the memory bus. Both CPU AND I/O channels were effectively idle and not running at max capacity. They convinced the Navy that I was "wrong" and recommended buying faster hard drives. When the new hard drives were installed, the system diagnostics showed NO change in application speed. (Several hundred thousands of dollars of taxpayer money spent needlessly.) I finally convinced TPTB to get an upgraded CPU (3 times faster) and run the tests again. VOILA! The applications ran three times faster. Eventually, I convinced them to scrap that system and go with a newer, much less expensive (less than 5%) system that ran at least 25-50 times faster.

    Sometimes it's not apparent where our real bottlenecks are occurring. If we don't know the entire "system," we can sub-optimize some component(s) and still have little overall system improvement to show for our efforts.

    ReplyDelete
  5. PART II:

    I think we've hit on this attention test before. The original video is here: The original selective attention task

    Web site: the invisible gorilla - Christopher Chabris and Daniel Simons

    If our conscious attention is "focused" (by either specific instructions or by a formal thinking "process") then it is almost certain that extraneous factors (the invisible gorilla) will NOT be given any conscious attention at all, even though we may "know" all about gorillas and their tendency to "pop up" in unexpected places at unexpected times. Our minds are definitely a "zoo" with only a few "obvious" exhibits "open" at any given time!

    When I was living in California back in the 1970s, I had a school teacher friend who loved chess. He set up an after-school chess program for kids. I was considerably stronger than him (1800+ to 1400-1500) and he knew it. He would invite me in to give demonstrations and lectures to the kids. At one point, he asked me to give him lessons. I suggested just playing, and I would point out some things as we played and allow him to take back moves that I knew would cause an (eventual) loss for him. In the very first game, as soon as we reached a point where he got sidetracked and failed to respond to my threats, I stopped and pointed out the error. He immediately started arguing that it was NOT an error, based on his assumption that I couldn't possibly know how the game would go, nor know what he was going to play during the rest of the game. I tried to reason with him, to no avail. I finally just gave up, and requested that we just play and he could then see what the result would inevitably be. From that time forward, he never got close to winning or drawing another game with me, even though we played several hundred games over the years. My wife asked me after every visit if I had let him win at least one game; the answer was always "No. If he wins a game, it will be because he has learned how to play better, and he will have won the game legitimately. I will not just throw a game to make him feel better." That may sound harsh, but we remained good friends throughout because he stated many times that he did NOT want me to "take it easy" on him and just let him win.

    There are none so blind as those who WILL not "see"!

    ReplyDelete
  6. On a different note:

    I just watched this video about Dan Heisman's Thinking Process.

    Learn The Dan Heisman Thought Process with IM Valeri Lilov (Webinar Replay)

    IM Valeri "Tiger" Lilov presents a game and analyzes it in terms of gaining and maintaining the initiative.

    1. e4 d5 2. cxd5 Qxd5 3. Nc3 Qd6 4. d4 Nf6 5. Nf3 a6 6. Bc4 Bg4 7. h3 Bh5
    8. g4 Bg6

    What should White play here to maintain the initiative (take advantage of the tempi already gained)?

    9. h4! Nxg4

    Maintain the initiative! Under these circumstances, material is not as important.

    10. Ng5! h5

    Note that Black is defending/responding to White's threats.

    11. Qf3

    Double attack against f7 and b7!

    11. ... Qf6
    12. Bf4

    "To take is a mistake!" GM Ilya Smirin

    Exchanges tend to dissipate an initiative unless it is converted into a more permanent advantage.

    12. ... c6
    13. O-O-O Nd7
    14. R(h1)e1 O-O-O

    What is the tactical "pattern" that White should "see" here?

    Boden's Mate!

    The "telltale" sign(s) are the White Bf4 cutting into the "box"
    around the Black King, the "possibility" of the White Bc4 moving
    to a6 with checkmate, and the White Qf3 "attacking" c6.

    What prevents Boden's Mate?

    The Black Pb7 has two functions: "protect" c6 and a6, and is thus overloaded.
    The Black Queen protects c6 (function).

    Solution:

    c6 is a critical PoP.
    First, cut the Black Queen's LoA with a forcing move.
    How can White remove the protection of c6?

    15. Re6!! fxe6

    Second, force the overloaded BPb7 to make a choice.

    16. Qxc6+ bxc6

    And the denouement becomes obvious.

    17. Bxa6#

    ReplyDelete
  7. Off topic:

    Deep Mind's AlphaZero crushed Stockfish 8 28-0 (25 wins with White, 3 wins with Black), with 72 draws in a 100-game match. Interestingly, AlphaZero searched just 800,000 nodes per second to Stockfish's 70,000,000 per second.

    And how long did it take for AlphaZero to reach this stupendously high skill level? Approximately 4 hours, given only the formal rules of chess and NO HUMAN KNOWLEDGE (no opening book, no endgame tablebases, no strategies, no tactics, no previously played grandmaster games). None of the state-of-the-art computer chess techniques used in Stockfish are used by AlphaZero.

    Link: Mastering Chess and Shogi by Self-Play with a
    General Reinforcement Learning Algorithm


    Hmmm. . . Maybe mister Lasker was not so far off when he hypothesized that a beginning HUMAN player could be trained to learn chess in 200 hours to the level that a master could not give any odds or he would surely lose. A very smart man, that mister Lasker, and certainly far ahead of his time!

    ReplyDelete
  8. Off topic:

    Correction: according to the linked paper, AlphaZero searched just 80,000 nodes per second, NOT 800,000 per second (which I got from some news article). The amount of training time for the neural nets was 9 hours, NOT 4 hours.

    So much for trusting the accuracy of journalists!


    Deep Mind's AlphaZero crushed Stockfish 8 28-0 (25 wins with White, 3 wins with Black), with 72 draws in a 100-game match. Interestingly, AlphaZero searched just 80,000 nodes per second to Stockfish's 70,000,000 per second.

    And how long did it take for AlphaZero to reach this stupendously high skill level? Approximately 4 hours, given only the formal rules of chess and NO HUMAN KNOWLEDGE (no opening book, no endgame tablebases, no strategies, no tactics, no previously played grandmaster games). None of the state-of-the-art computer chess techniques used in Stockfish are used by AlphaZero.

    Link: Mastering Chess and Shogi by Self-Play with a
    General Reinforcement Learning Algorithm


    Hmmm. . . Maybe mister Lasker was not so far off when he hypothesized that a beginning HUMAN player could be trained to learn chess in 200 hours to the level that a master could not give any odds or he would surely lose. A very smart man, that mister Lasker, and certainly far ahead of his time!

    ReplyDelete
    Replies
    1. Just checked the last days the progress of CT-User in tactical strength again.. and again: i found no one with a relevant improvement after the first 4000 attempts. That are roundabout 140 hours max for tactics ;)

      Delete
  9. Here is Munich:
    I avoid (deep) tactics, if I can. For the rest, my tactical skill is sufficient.
    Chess is not 99% tactics, unless you count the really easy tactics to those 99%.
    But when you only consider difficult tactics, I have had many games where no such tactic was on the board.

    In Blitz, I can not even see fast enough relatively easy tactics, so my Blitz and bullet rating is far behind my rapid and classic rating in lichess.com.
    Out of ~80.000-90.000 players, I am in the top 50, so I am better than 99.9% of all players.
    Well, I suppose the average rating of the player pool is relatively weak, and no GMs and IMs bother to play longer games at lichess. But many FMs are about my rating.

    Looking at their bullet rating, I am sure they are much better than me in tactics. They often have 2300 Blitz ratings or so, whereas my blitz/bullet is rediculous (= below 1900 most of the time).
    But when I play them in a REAL game, I do fine. It doesnt matter that they see things within 10 seconds, which I see only after a minute. The most important thing seems to be: I see most tactics at the end, too, and my positional accuracy is way better.

    Which means: I guess Temposchlucker could improve in rating in longer games - without better tactic skills. Or maybe, in my case, I reached just about a sufficient tactical skill, so that I can stand my ground.
    I noticed that there ARE GMs and IMs around, who are shy in playing anything really fast. Probably, like me, they are not so good in tactics, and they know it.
    I could imagine that some IMs struggle to get to 2000 CT Blitz rating.

    ReplyDelete