Forum for Tewari

wondering about sequence dissection. [#1099]

Back to forum     Back to page

New reply

Minue: wondering about sequence dissection. (2007-08-08 17:14) [#3755]

Peronally, I hate using a japanese Go term, and avoid using it if possible. So, I will use a term "sequence dissection" (very literal translation of the orginal term) instead of 'tewari' from now.

Anyway, I'm wondering what kind of argument sequence dissection is...

Let's see a very very specific and clean example of an argument with sequence dissection which I made right now after reading BMQ375.

See the 2 hypothetical openings below.

An hypothetical opening sequence(1)  
An hypothetical opening sequence(2)  

And now, a typical argument with sequence dissection would follow something like this.

1) In the first diagram, (Hypothetical opening 1), Every move from B1 to B7 is reasonble(good move). In other words, it means that the game pos would be kept even for both, assuming both black and white will play a correct follow up.

2) But white's actual follow up, W8 is a stupid move, and B9 is a proper move to punish white's joseki mistake.

3) So, black is leading now.

4) Now, see the second diagram(hypothetical opening 2). Its end result is identical with the first diagram with just different move orders.

5) Since its end result is excatly the same with the previous diagram, Opening positin in the second diagram also must favor black.

Conclusion : So, it must be that white did something wrong in the second diagram.

My questions are as follows.

1) In this specific argument,

Assuming that all premises (from 1) to 5)) are true, the conclusion (white did something wrong in the second diagram) is logically deduced? (just as conclusion is logically deduced in Aritotle's syllogism)

"In my impression", yes. But, Im not totally sure.

(Talking about side issue, all of 1)~5) premises are of course really true)

2) And, if conclusion can be deduced logically in the above specific hypothetical argument with sequence dissection of opening sequence, then does it mean(or imply) that argument with sequence dissection can be a sound dedecutive inference "in general"?

My understanding of logics is quite elementary, so I can not find answers to these questions on my own.

So, I hope someone with specialized (or professional) knowldge of logics would help me.

Thank you~

LukeNine45: Re: wondering about sequence dissection. (2007-08-08 17:29) [#3756]

Wouldn't another possible conclusion be that

A) the second diagram is "normal" (even), because

B) B1 in the first diagram was a mistake, but

C) White makes an equivalent mistake with W8, bringing it back to the "normal" position?

I think that if it can be proved that B1-B7 in the first diagram are 100% good and W8 is the only mistake, then it is necessary that white has also made an equally large mistake in the second diagram.

In other words, I think your logic is valid-- the difficulty is in proving that your data (that B1-B7 are correct) is valid. I don't doubt that it is correct-- I just don't know how one could prove that rigorously (or disprove it, either-- I'll have a hard time believing that the enclosure is wrong, as I've suggested above).

Which leads me to my question: What did white do wrong in the second diagram? Is it W2, an approach that plays into a pincer?

Minue: Re: wondering about sequence dissection. (2007-08-08 18:04) [#3757]

A) Wouldn't another possible conclusion be that...

==> 'No' in 2 senses.

First, second diagram is "really" bad for white.

Second, in my question, I suppose that we know only about premises, but do not know if the second diagram is good, even, or bad for black.

A judegement about second diagram is what we have to guess(deduce ,infer) from the given premises.

what I asked is if we can deduce any conclusion about second diagram logically from the 5 premises (given that they are all true)

B) Which leads me to my question: What did white do wrong in the second diagram? Is it W2, an approach that plays into a pincer? ==>

Yes, W2 in the second diagram wrong. A right move for white is to play a loose approach to the top left black corner.(W2 or white a)

W4 in the second diagram is another mistake. W should have played avalanche.

19x19 diagram  
LukeNine45: Re: wondering about sequence dissection. (2007-08-08 18:11) [#3758]

Ah, I understand.

I can't see any flaw in your logic. Given your premises, your conclusion certainly follows.

Minue: Re: wondering about sequence dissection. (2007-08-08 18:27) [#3759]

ok, if you think so, what do you think about my second question?

Do you think it implies that an argument with sequence dissection can be a sound deduction in general?

If so, how can we distinguish sound argument with sequence dissection from flawed ones?

I 'vaguely' remmeber that I learnt to use venn diagrams to check validity of a certain type of syllogim in a logic class when I was in school. (though I forgot all of specifics of it...)

Would it be possible to have a sort of formal method (like a venn diagram in syllogism) to check validity of sequence dissection arugment? what do you think about?

LukeNine45: Re: wondering about sequence dissection. (2007-08-08 18:42) [#3761]

Hm, I think it's helpful to distinguish a valid argument from a correct one:

A valid argument is one in which the conclusion follows from the premises.

A correct argument (I forget the technical term but this will serve) is both valid, and the premises are factually true, meaning the conclusion is factually true as well.

("All dogs have three legs; Jenny is a dog; therefore Jenny has three legs," is an argument that is valid but not correct.)

I think this argument is always valid-- the problem is verifying the premises every time one tries to apply this reasoning.

The argument boils down to "white has a bad result in diagram 1, so white also has a bad result in diagram 2" which is true by the law of identity (x=x); the trick is figuring out who made the mistake(s) and where.

Bill or Charles probably can be more helpful than I can be though...

Minue: Re: wondering about sequence dissection. (2007-08-08 19:15) [#3762]

So, you think

1) any arugment with sequence dissection is valid in its nature (since it relies on the law of identity)

2) and it can be correct one if its premiese are factually true, if not, valid but not correct.

LukeNine45: Re: wondering about sequence dissection. (2007-08-08 19:37) [#3765]

I would agree with that, and I think Unkx has a good explanation below... Re: wondering about sequence dissection. (2007-08-09 05:46) [#3778]

In terms of using sequence dissection on this position...

19x19 diagram  

We change the sequence like this. Now it is apparent that W8 here is not appropriate. White would normally play away. This is the normal use of sequence analysis, IMHO. The larger-scale (potentially more important) questions of W2 and W4 are not subjects for sequence dissection. Dave

Bill: Re: wondering about sequence dissection. (2007-08-08 19:17) [#3763]

I am not sure, but I think that permutation (changing the order) is what you are talking about now.

As far as logic goes, if White did something wrong in the first line of play, so that the end result is bad for White, White did something wrong in the second line of play, as well, since it produced the same result.

However, the usual case is one in which we do not know whether a result is good or bad. And in general, we know that the order of play matters, so that we cannot freely permute it. A play that might have been good when it was played may not be good when other stones have been played.

So we are talking about a heuristic, a guide for play and judgement that is not infallible.

Minue: Re: wondering about sequence dissection. (2007-08-08 19:49) [#3767]

yes, you got a point.

Changing move order is an essential part in sequence dissection, and to re-arrange move order in a reasonable way, we have to rely on our Go knowldge.

Then, it's no wonder that ordinary amatuer players often feel suspicious about pro's sequence dissection, since they dont share same amount of Go knowldge with professionals.

Unkx80: Re: wondering about sequence dissection. (2007-08-08 19:41) [#3764]

Of course there are mathematicians who are far better than me, but I think there is one relatively easy explanation to it. For this discussion lets ignore the effects of ko or similar stuff.

I will model the problem as follows. For every position we can say something like, Black leads by 10 points or White leads by 20 points. So we can assign each position a score. Lets agree on a convention that if Black leads, the score is positive, and if White leads, the score is negative. So Black leading by 10 points means a score of +10, White leading by 20 points means a score of -20. If a position is even, then the score would be 0. An optimal move by either player does not change the score of the position. A sub-optimal move by Black decreases the score, and a sub-optimal move by White increases the score. Black alone cannot increase the score of the position, and White alone cannot decrease the score of the position.

In your first sequence, you claim that all the moves are (more or less) optimal for both players, except for W8. Hence, the score of the position after each move before W8 should be about 0. But W8 causes the score to be changed to +k, where k is a positive number much larger than 0. If you plot a graph of score against move number, then you see that the graph starts with score 0, is flat until W8, and then an increase up to +k.

Given any move reordering (sequence dissection) of the first position, such as your second sequence, the first point and the last point on this graph are the same since they are the same position. From here we can immediately conclude that there must be at least one White move that is sub-optimal, although we cannot tell whether there are sub-optimal Black moves. The reason is because somewhere along the way the score must increase.

I think this gives a somewhat intuitive idea. I think my modelling is right, but if it is not, maybe Bill or someone else can correct me.

Bill: Re: wondering about sequence dissection. (2007-08-08 20:11) [#3768]

Thoughts about etymology. I think that te may simply refer to a move or moves instead of a sequence, so that move deletion might be close to the original. Putting moves back in a different order produces permutation of the sequence.

Minue: Re: wondering about sequence dissection. (2007-08-08 20:31) [#3769]

Te(手) means literally 'hand'. It is used to mean "a move, or moves (played by hand)" analogically in Go.

And "moves (by hand)" actually means nothing but " a sequence of moves" in Go.

Because Tewari examines a sequence of moves (not a single move), "sequence dissection" is more closer to the orginal meaning of the term rather than move deletion.

Its original Kanji(chinese characters) is '手割論'

手: hand

割: dissect, divide, cut.

論: (usually)theory,(but sometimes, method).

So, 手割論 literally means " a theory(a method) of dividing(dissecting) hand(move, a sequence of moves)

We can not divide any single move. only possible to dissect or divide a sequence of moves.

So, sequence dissection is one of best translations of it which is very faithfull to its meaning.

Unkx80: Re: wondering about sequence dissection. (2007-08-08 20:41) [#3770]

Hmm... I understand the Kanji, and so I understand how you came out with the translation. However, the effect of what you call "sequence dissection" is an analysis of some reordering of moves (the word "permutation" has the same meaning). To an English audience, I wonder how many people will actually understand "sequence dissection"... =P

Minue: Re: wondering about sequence dissection. (2007-08-08 20:53) [#3771]

To east asian audience (who know well kanji, also play Go but don't know about tewari yet), "手割論" would not make much sense in itself, until they learn what it really is, just as English audiece.

Do you really think that 手割論 would make more sense to east asians than sequence dissection would do to english speakers?

At least, "sequence dissection" would give some hint of of what it is better than tewari.

Anyway, I don't see any better translation of it than sequence dissection yet.

LukeNine45: Re: wondering about sequence dissection. (2007-08-08 20:55) [#3772]

What about "sequence deconstruction", to borrow a philosophy term?

Dissection implies that you cut something up and that's it; deconstruction hints at taking something apart and rebuilding it, putting it together another way...

Minue: Re: wondering about sequence dissection. (2007-08-08 20:54) [#3773]

sequence deconstruciton?

sounds good. I like it. but, let me think about it for a while...

Unkx80: Re: wondering about sequence dissection. (2007-08-09 19:26) [#3782]

"Sequence deconstruction" sounds good. I totally forgot about one of the approaches to tewari is not purely about move reordering, apologies for that.

blubb: simple permutations, intricate dissections (2007-08-09 18:53) [#3780]

As long as we merely permute a fixed sequence of moves, starting from the same position, with no stones being captured either way, the ending positions are also the same. As Unkx80 points out, the net score shift (e. g. a loss for White) over the first sequence then necessarily equals the net score shift over the second. We can say for sure that, if one color plays worse in one permutation than in the other, the opponent does so, too. This relation is generally valid, but without "Go knowledge", useless for practical analysis (for reasons illustrated by LukeNine45's three-legged dog).

I think that even more Go knowledge is required to sensibly perform the actual dissection, which goes beyond permuting. In principle, tewari, or "sequence dissection", can help to analyse a particular difficult to judge about subsequence (may consist of just a single move), if both the complement of that subsequence and the whole sequence are easier to judge about. Likewise, it can help to analyse an obscure whole sequence which can be devided into subsequences that are clearer. Both ways may also involve permutations.

Supposedly, the most popular application of this principle is to analyse a whole sequence by:

  1. removing a well-chosen subset of its moves that is considered sufficiently close to neutral, to get a remainder which is easier to evaluate
  2. evaluating that remainder

There is a catch in here: If the removed subset is biased, you can prove about anything! This might be (part of) what makes "ordinary amatEUr players often feel suspicious about pro's sequence dissection" (minue). The undone moves must be significantly easier to judge about than the whole sequence, so we are able to see that their net effect is really close to neutral, or the entire procedure is moot.

blubb: Re: simple permutations, intricate dissections (2007-08-10 10:14) [#3785]

A variation of the application above is to analyse a whole sequence, leading from a reference position A to the ending position C, by:

  1. removing a well-chosen subset of its moves, such that the remainder (A->B) is considered sufficiently close to neutral
  2. replaying and thereby evaluating the undone moves (B->C)

(Note that the roles of the removed part and the remainder are switched.)

Again, there is a catch: If we can't tell whether the subsequence of moves that leads from A to B really is (roughly) neutral (that is, whether B is worth about the same as A), we cannot build a meaningful tewari analysis on (B->C), either.

There are plenty good tewari examples at SL, including the ones in minue's translation of [ext] Moon yong-jik 5p's explanation of sequence dissection(tewari). To clarify the issue, here is an (admittedly, hilarious) counterexample that results from Moon yong-jik 5p's position by choosing a subsequence (A->B) that is not neutral.

position C, White to move  

At one point of time, the position looks like this.

If we reordered the moves that led to here (from A, the empty board) such that the marked stones would come last, and took everything but those last two moves as (A->B), this would be our B:

position B, White to move  

In step 2 we would now play White o and then, Black x. Apparently, in this situation, that's quite a bad exchange. White would be much better off by playing p instead. Even q would still be better than o.

However, this does not mean that the whole sequence is bad for White. The value of (B->C) does not give us the value of the whole sequence (A->C), because our analysis is flawed. (A->B), regardless of its inner ordering, is a very bad choice of a "neutral subsequence", since it strongly favors White. This unrealistic advantage explains the likewise unrealistic disadvantage of (B->C) (i. e. the o-x exchange). Both together (possibly reordered) constitute the realistic, relatively balanced sequence (A->C).

This suggests that tewari analysis, or sequence dissection, does not work any better than the Go knowledge it is based on.

Back to forum     Back to page

New reply

[Welcome to Sensei's Library!]
Search position
Page history
Latest page diff
Partner sites:
Go Teaching Ladder
Login / Prefs
Sensei's Library