Deiri Counting, Miai Counting, and Tedomari

[Kirby]

Deiri counting is simple to understand. Miai counting is understandable, but more complex.

From the SL entry on miai counting:

You can compare miai values directly. In general, you make the play with the largest miai value. Also, miai values add and subtract like ordinary numbers. Neither is true of deiri values.

I am most familiar with deiri counting, because it's the only method I've learned well, but if miai counting can be used to directly prioritize endgame plays, it is worth learning.

As I am only really familiar with deiri counting, I have come across situations where tedomari comes into play. That is, even though I have a particular ordering of deiri counts for different areas of the board, it's not always correct to play this order in the case of tedomari, because of trying to get the last big play.

---

My question is, is the same true for miai counting? If I learn the way to count using miai counting, is tedomari, for example, still an issue? The quote from SL seems to indicate that you can simply calculate the miai count values, then play in order of descending value. Is this true? Are there exceptions to this?

[Bill Spight]

I am most familiar with deiri counting, because it's the only method I've learned well, but if miai counting can be used to directly prioritize endgame plays, it is worth learning.

Miai counting, or Absolute Counting (TM), tells you how much a play gains, on average. Deiri counting does not.

As I am only really familiar with deiri counting, I have come across situations where tedomari comes into play. That is, even though I have a particular ordering of deiri counts for different areas of the board, it's not always correct to play this order in the case of tedomari, because of trying to get the last big play.

---

My question is, is the same true for miai counting? If I learn the way to count using miai counting, is tedomari, for example, still an issue?

Oh, yes. Getting the last play, or the last play before the size of plays drops, is an issue, regardless of how you count the plays.

The quote from SL seems to indicate that you can simply calculate the miai count values, then play in order of descending value. Is this true?

If I gave that impression, I apologize. It is true that in general the play that makes the largest gain is the best play, but go is not so easy.

Are there exceptions to this?

You bet! And last play situations form a major class of such exceptions.

[Kirby]

Thank you for your responses, Bill.

If I understand correctly then:
1. Neither deiri counting nor absolute counting account for last play situations where one should not strictly play in descending order of the calculated value.
2. Suppose that we ignore the exceptional cases (last play situations, included), and instead use a heuristic: playing in descending order of the calculated value. If we follow this heuristic, absolute counting will be on average more precise than deiri counting, since it takes into account the local tally. Correct?

Given that you can count precisely using miai counting, is iteration of different play combinations the best strategy for solving problems involving "last play situations"? Is there a more organized approach to playing optimally to get the last play (other than trial-and-error)?

[Bill Spight]

If I understand correctly then:
1. Neither deiri counting nor absolute counting account for last play situations where one should not strictly play in descending order of the calculated value.

Right.

2. Suppose that we ignore the exceptional cases (last play situations, included), and instead use a heuristic: playing in descending order of the calculated value. If we follow this heuristic, absolute counting will be on average more precise than deiri counting, since it takes into account the local tally. Correct?

Well, those who use deiri counting make adjustments so that their values come to twice miai values. The problems lie with getting score estimates right. For instance, if you are 3 points behind and then you get the last play, which is a 4 point gote by deiri counting, you win, right? Wrong. Your play gains only 2 points and you lose.

Given that you can count precisely using miai counting, is iteration of different play combinations the best strategy for solving problems involving "last play situations"? Is there a more organized approach to playing optimally to get the last play (other than trial-and-error)?

Well, not everything has been mapped out, but a study of go infinitesimals ( http://http://senseis.xmp.net/?GoInfinitesimals ) will help. This is one area where Western go is ahead of Eastern go.

[Kirby]

Thanks so much, Bill. Things are getting clearer.

I just have one more question that I can think of now. Namely, when you say, "those who use deiri counting make adjustments so that their values come to twice miai values", what adjustments are you referring to? Could you give an example?

[Bill Spight]

Thanks so much, Bill. Things are getting clearer.

I just have one more question that I can think of now. Namely, when you say, "those who use deiri counting make adjustments so that their values come to twice miai values", what adjustments are you referring to? Could you give an example?

Sure. Reverse sente values are miai values. But those who use deiri values double them. They also adjust regular ko deiri values by multiplying by 2/3.

[Kirby]

Great, thanks! Now just time to study...