Path: Endgame =>

Doubled swing value [#3235]

tamingo: Doubled swing value (2016-03-21 23:49) [#10693]

In the paragraph Is the Swing Value Sufficient?, the difference between the two endgame sequences should be 1 point and not 2 points as stated.

In a 3 point swing-value endgame, counting -3 points if white takes it and +3 points if black wins, doubles the final score.

Normally, you'd count 0 points if white wins and 3 points if black wins (with arbitrary offset).

The explanation is quite clear and understandable, just the final result is irritating. -- tamingo?

X
75.179.4.177: Re: Doubled swing value (2016-03-22 00:00) [#10694]

this is a wiki :) if you're even semi-confident you can make the change yourself, inline

107.210.159.110: Re: Doubled swing value (2017-03-27 17:50) [#10929]

Bill:

Well, it has been about one year since tamingo pointed out the error. Obviously he did not edit the text, and rightly so, IMO. The problem is that the calculations needed to be done with miai values to get the right answer, and he did not seem to know that. (It looks like he did understand the problem, just not that the solution already existed in the form of miai counting.)

Neither did whoever wrote that section, which suggests that maybe it does not belong on a beginner page.

Most go players learn swing counting, and then imagine -- rightly so, IMO --, that they can use it to calculate how much a sequence of plays gains or loses. Even strong players can make that mistake, because they were not taught correctly. That is why, over 20 years ago, I went on a crusade to introduce miai counting to the West. There is still a long way to go, it seems.

HermanHiddema: Re: Doubled swing value (2017-03-27 21:38) [#10930]

I wrote most of the page, originally. At that point in the page, the purpose of that section was to show beginners that simply using the swing count is insufficient, because it does not account for sente and gote. The page then goes on to introduce miai counting off of that a bit later. So the problem is: I want to show that swing counting is insufficient as a lead-up to miai counting. Meaning I can't use miai counting here yet. So yes, it is wrong, but how do we make it right without using miai counting?

IMO, the original is actually quite reasonable. The point really isn't to give beginners accurate values, it is to show them that order matters in the endgame. And those wrong numbers do show the right order. (As an aside, I am of the strong opinion that it is fine, and even desirable, to lie to beginners in order to teach. Gloss over the details to give them the big picture, then give them the details when they're ready for them). Perhaps just adding a note saying "These numbers are not entirely accurate, but we'll get to that in a later section" suffices?

Regardless, I would really appreciate if you could re-edit is so it doesn't have signed stuff in the middle of the page. It suddenly turns it into a conversation/discussion kind of thing, which really breaks the flow of the page, IMO. Either improve it or restore it, I'd say.

107.210.159.110: Re: Doubled swing value (2017-03-28 17:03) [#10931]

Bill: There are (at least) three issues here. The first is that swing values need adjustments. The initial value of a sente is doubled, and the initial value of a simple ko is multiplied by 2/3, for instance. (These adjustments have also led to confusion, BTW. If you double the value of sente, why not quadruple the value of double sente, for example.) Why teach swing values without adjustments?

If you want an example for doubling the swing value of sente, then you can make a slight change in the example by giving the smaller yose a swing value of 3. Then it does not matter whether White plays the 3 point sente first or the 6 point gote. And it does not matter whether Black plays the 3 point reverse sente first or the 6 point gote. The results will be the same for each player playing first.

The second issue is that of the order of play. Swing values, even adjusted, do not tell you the correct order of play. But neither do miai values. Taking the largest play is not always correct. But it is in the vast majority of cases. That is why it is useful. So that issue does not belong before miai values are introduced, but afterwards.

The third issue is that you can't add and subtract swing values, even adjusted ones, to tell you the score difference between two different lines of play. You need miai values for that. In fact, that is the basic reason to use miai values. Using swing values to compare the results of two lines of play is not just lying to beginners, it is leading them down the wrong path.

HermanHiddema: Re: Doubled swing value (2017-03-30 13:49) [#10932]

Ok, sure, I don't disagree with that at all. What I'm asking is: Can you rewrite it, instead of putting comments in the middle? (Keeping in mind that this page should also be readable and useful to a random 15 kyu with no prior knowledge, who wants to get a little bit better at endgame but is not interested in delving into the math behind it all)

tapir: Re: Doubled swing value (2017-03-31 15:35) [#10933]

This page was one of the good things coming from our attempts to make treatment of endgame more readable and more useful. It is otherwise largely tuned to curiosities and not utility. (E.g. we have lists calculating fractional miai values but not even a rough comparison of large endgame moves, keima slide, 2-2 push into corner etc.)

The edit on 26 March turns "life & death status of many groups has been settled" into "the board has divided up into a number of independent or nearly independent regions" - beginner page much? No one complained about that?

"Naive approach" etc. are all markers that the following is wrong for didactical reasons. And the page continues to introduce miai counting right afterwards... (this is wrong, read below how you should do it). But no, not even this protects this page? It is just like the people interrupting an explanation of life & death (2 eyes mumble mumble) with (put seki on board) "And what about this? Don't believe what he says."

107.210.159.110: Re: Doubled swing value (2017-04-01 04:31) [#10935]

Bill:

I made the change from life and death being settled, because that is not the point. There are many endgame plays that involve life and death. The main thing about endgame theory is deciding which independent (or nearly independent) region of the board to play in. A play over here is bigger than a play over there, for instance.

tapir: Re: Doubled swing value (2017-04-02 13:01) [#10937]

For me this sounds like a complicated way to say the same - a bad change in a beginner page. The "settled" does not refer to a fixed life-and-death status, but that the outcome assuming local answers as suitable is not in doubt, if I read correctly. (This is something I can determine as player, while the independence of regions is some high level concept.) Of course a lot of local moves will threaten the status to some degree and where suitable large exchanges might happen when the player thinks it is beneficial. The independence of "regions" is the result of groups not changing status anymore to my mind, when life and death status changes, e.g. an exchange happens, the independence breaks down as well.

107.210.159.110: Re: Doubled swing value (2017-04-02 16:33) [#10938]

Well, you may be able to determine easily whether a group's life and death are settled, but beginners certainly can't, and it is not so easy for a lot of SDKs, either.

In any event, it is certainly characteristic of the endgame that life and death are settled for most groups, and nearly all of them are independent, forming what Ogawa calls separate territories. I have copied her quote from the endgame page, and made other edits addressing life and death.

107.210.159.110: Re: Doubled swing value (2017-04-01 03:58) [#10934]

Bill: Thanks, Herman. :)

I will make a few unsigned edits in a few days.

107.210.159.110: Re: Doubled swing value (2017-04-10 16:13) [#10939]

OK, I have rewritten the section. But it still needs a rewrite. Probably next week, after doing taxes.