[Welcome to Sensei's Library!]

StartingPoints
ReferenceSection
About


Referenced by
BottleneckTheories
DietersIdeasOnImp...

 

Improvement
   

Improvement in general is obtained through the following three aspects

  • The acquisition of new ideas
  • The eradication of wrong ideas
  • Consolidation of acquired knowledge

Since Go is neither a fully solved game nor an exact science but instead relies on a combination of intuition and reasoning?, these three aspects of improvement can be obtained on both the intuitive level as the analytical level.

It is widely believed that the following practices lead to improvement on some level:

  • Reading Go books or online articles and having a teacher enables one to acquire new ideas on an analytical level.
  • Replaying professional games? enables one to acquire new ideas on an intuitive level.
  • Playing fast games is a (the) way to test and consolidate intuitive knowledge.
  • Playing slow games is a (the) way to test and consolidate reasoning.
  • The analysis of games? enables
    • the evaluation of analytically based judgment, as exercised in slow games
    • the evaluation of intuitive judgment, as exercised in fast games
    • the identification and eradication of wrong ideas, bad habits or wrong attitude?

The overlying aspect of improvement is attitude?. The willingness to acquire new ideas, to investigate flaws in one's own play, to reconsider what one thought was correct, are all indispensable in the quest for improvement. In particular, a preocuppation with rank - the desire to win or the fear of losing - can distract from the real objective of becoming better at playing Go.


See also:


Andy Pierce: How about watching more skilled players playing each other? I'm not talking about pros but just people three or four stronger than you are. I 'think' this is useful anyway, but if I can be convinced otherwise I'll stop wasting my time doing this. There seem to be certain discrete classes of overplays and mistakes that are made at any given level. Watching slightly better players shows those mistakes for what they are and how they get punished. If you just study pro games, you never get to see these basic errors (and the correct responses to them).

Anon: I think this is the basic idea behind for example AlexandreDinerchtein's commented KGS games. These are useful for amateurs to study because the amateur's games contain more basic mistakes than in pro games --- mistakes an amateur player is more likely to make. But, there is a difference between reading a pro's comments and just watching the game yourself: it is quite likely that you will not realize many of the mistakes made, and further more that you risk picking up these bad habits.

ilanpi This is somewhat irrelevant as it only refers to the above statement: "Since Go is neither a fully solved game nor an exact science..." so won't help anyone improve. I claim that the above quoted statement is irrelevant! For example, multiplication of positive integers is fully solved and an exact science, but that doesn't help human performance, for example if you were to have mental multiplications competitions. In fact, there might be some similarity between "human calculators" and go professionals (early familiarity with the game, etc). Moreover, human multiplication may even have hindered the scientific theory of multiplication since the first theoretical speed improvement ever was in 1965, that is, corresponding with the implementation of multiplication on a computer. For a game example of this, there is Nim, which is fully solved and can be programmed in a few lines of code. I claim that human Nim playing cannot take full advantage of this and will eventually resort to some of the usual game playing skills that are refered to above.

Dieter:

The statement is "Go is not fully solved, hence it is necessary to become familiar with heuristics to become better at it"

Your statement is "For solved games/activities too, heuristics can lead to mastery"

The latter doesn't disprove the first, neither does it make it irrelevant. It would do so if heuristics were necessary for all solved games. There are numerous games, fully solved, where heuristics are not necessary at all: you just learn a/the method. Of course, if you do it often enough, you'll incorporate the method in your intuition. I.e. the method is not necessary to be good at a fully solved game.

As often, "A is not necessary for B" is used to disprove "(not A) is necessary for (not B)", which is logically wrong.

ilanpi Nope,your interpretation of what I said, that is: "For solved games/activities too, heuristics can lead to mastery" is not at all what I meant to say, sorry if I failed to make my point clearly enough. The statement I wanted to make was: "For many of the solved games/activities I know, the theoretical solution doesn't help human performance, which must therefore rely on heuristics, intuition, memorization, etc." Therefore, my conclusion that theoretical solutions are often irrelevant to unaided human game playing.

Of course, I should add that the original statement is valid if one accepts that writing a strong computer program to play a game means that one has attained mastery of that game.

Dieter: Hi. I'm not defensive or anything, I just like discussing these things #:-7. Having written a strong computer program is not the same as having solved a game/problem (well, lesser purists may disagree and give the map colouring problem as a counterexample). You are absolutely right that some games are solved yet the solving method lies beyond most humans' reach. So they have to rely on heuristics anyway. This doesn't render the fact that Go isn't solved irrelevant to it having to rely on heuristics. All solved games should have to rely on heuristics for "being solved" to be irrelevant, which clearly is not the case. Many patience-like games are solved, and the method can be taught to humans.

So, since Go has not been solved, we have to rely on heuristics. It may be solved one day, and we'll may still have to rely on heuristics.

But, I may also want to rewrite the statement as "Since Go is a very complex game (and by the way not yet solved) we ..."

Cheers

ilanpi Thanks Dieter, and I agree with your final formulation.

One of my main interests is the comparison between human play and theoretical optimal play. This is not yet possible in go (I suppose you can't count 7x7 go, which appears to completely solved by humans alone). I have devoted some thought to this on my Mather of the Game site [ext] http://cf.geocities.com/ilanpi/dots.html. Also interesting are the comments of John Nunn in his book Secrets of Pawnless Endgames, where he writes about the meaning of "understanding", in the sense we have just discussed. In fact, this series of books by Nunn is exactly about trying to understand in human terms the pronouncements of a perfect game playing oracle. Once again, Go seems to be out of reach for such considerations, which is maybe why I am playing it now -- I was frustrated with chess because I couldn't get my brain to think like a computer.



This is a copy of the living page "Improvement" at Sensei's Library.
(OC) 2004 the Authors, published under the OpenContent License V1.0.