Now that the people on Sensei's Library have done such a good job figuring out miai values of endgame plays, I'd love to an endgame study tool based on those relative values.

For example, if there were cards with positions/moves on the front and miai values on the back, the following could be a fun study activity:

You'd deal out a few random cards and you try to arrange them in the right order. You don't need to guess the exact miai value specifically; you just need to sort the cards you have. And then you can flip them over and see if the numbers are ascending, in which case your idea of the relative urgencies was correct.

This could be a good way to get a good intuitive sense of endgame urgencies, right? Sounds pretty fun to me.

Has someone already made a miai-value–based study tool? ← tl;dr: main point of the post right here

Looking for cards to print or order, or a digital app, doesn't have to work exactly like I outline here, I just feel like this could become a shortcut to getting good (or at least better) at what's the least creative/subjective/feels-based part of go in a few days.

Because there are two problems with endgames that make them an exception to the "the best way to study go is to play a lot" school-of-thought.

1. There is a right & wrong answer based on math that only very few specific board positions are the exception to, and
2. Just playing the game won't show you whether or not you were near that right answer the way something like this will.

Things like filling a teire before filling a ko, for example. Or when it's time to block a monkey jump.

Gomagic has a free light set here: https://gomagic.org/endgame-flashcards/
I didn't use it so I can't say if it's useful. I think they use deiri instead of miai counting.

Although there is nothing wrong with your suggestion of cards or app, the main focus on becoming strong at the endgame should be the practice of calculating the values. Study books with miai values for that because there you find correct values and the necessary explanations.

Robert, I wanna get a good-enough intuition for the most common things like teire and ko and crawl and the descend "double sente" that shows up twenty times per week.

I don't plan to calculate it in mid game, making a full tree and summing up and divide and stuff, for the same situations that show up again and again.

Your own joseki books were a good read for me at creating a sort of good-enough intution for fun games so I can get to the part of the game that I personally enjoy. The midgame.

Y'all (starting with Ban'nai Jun'ei's work in the sixties) have already calculated it out—https://senseis.xmp.net/?MiaiValuesList has some remaining question marks but is mostly good—and I don't look forward to sitting down at the goban and try to do these branch calculations in my head.

Doing the calculations is practice for doing the calculations, not for using the relative position derived from values.

When I first discovered Sensei's Library and all the wonderful and detailed endgame work on temperatures and komastery and miai values, I liked what I was seeing but I also got overwhelmed. It caused my endgame to get a lot looser, thinking "I'll have to learn to play the endgame properly one day since there's obviously a lot more precision to this than I thought" and since then my games are usually won or lost in the midgame, through fighting. Then the other day I thought of the idea in this post: people have already made the work, which is wonderful, and I can just practice with that work.

Similarly to how in chess I use tablebase apps to practice the endgame and that process of practicing is very, very different from the process of making such tablebases.

For the most common practical shapes, a valueless study can also help as long as values are not too close to each other.

For what is informally called double sente, never use the traditional value as local double sente does not exist. Rather these local gotes with almost as large follow-ups can be played as global sentes at suitable timings but this is global strategy and local card training gets you nowhere.

As to Sensei's Library, this gives you a very wrong impression of miai value calculation because SL concentrates on the arcane difficulties often from research perspective rather than the practically useful shapes and basic calculations.

Valueless endgame study (such as sente before gote) can be good for kyus but it brings you only so far and then value calculations cannot be avoided whenever decisions at a glance are impossible, sente or gote might be correct and the initiative on the rest of the board must be taken into account.

I'm also thinking about the small situations at the very end that show up in many of my games every week. One-stone teire, one stone first line crawl, ko when both sides are super alive, capturing a small number of stones in an alive position etc.

If someone were to print out all those positions on the miai values list, hide the values, and ask me to sort them, I don't think I'd do very well. That's why I'd want to do this practice. Like https://wikitrivia.tomjwatson.com/ game

Value calculations can not be done in a rapid time-controls game. A good-enough intuition is needed there. Card study can help me get part of the way towards an intuition faster.

Bill Spight wrote on Sensei's Library: "How do I play yose myself? Pretty intuitively. But my intuition is informed by having figured out many positions. Also, it was good even before I learned how to calculate the value of plays. "

Reading that, I thought I could give my own intuition a massive boost (not to get anywhere near the level of Bill's but a huge delta from where my own yose is currently at) by piggybacking on all the fantastic miai calculations that have already been done.

I'm not a pro, go is not my job, I'm an amateur, I just play for fun, and it is very fun and I love it very much, and for me at this state I am pursuing the kinds of curiosity-driven practice and study that is intrinsically fun as opposed to being means to an end. And for me as an emotional creature, doing the calculations is not fun while practicing the "feel" of the relative values of positions seems like it'd feel very fun and rewarding since I know that kind of study is, for me, intrinsically fun in chess.



Which is why a more usable format (that still used miai values rather than just deire) would be a great tool for many kyus, so great that I was wondering if it already existed. I mean, some of the shapes in that list are practical and common, but the format isn't.

While server-style byoyomi-only games are the enemy of value calculations, I regularly do positional judgement and at least some endgame value calculations. The latter is hard in 20s per move and rote memorisation of standard shapes helps here, especially when a ko is involved. However, such byoyomi is too fast by far to calculate everything.



Under time pressure, I prefer recalling values else guess-approximising by identifying the obviously larger else quick and dirty calculations. I do not believe in the existence of intuition but speak of subconscious thinking where explicit thinking and educated guessing end.



A great researcher but a weak model for endgame practice during playing...



Winning is fun and more winning eventually also requires calculations!



See https://www.lifein19x19.com/viewtopic.php?f=17&t=19136

It's not fun. Especially if it involved doing boring things.



I'm not talking about those "eventually" situations.

I know how to calculate it. It's just too boring to do it midgame. Or ever.

Doing these calculations for the first time, I believe that was fun. Exploring & mapping new territory. But redoing that work now is a pointless chore.

Conclusion: these cards or app don't exist and it might be worthwhile for me to make. That would be a terrible nightmare of tedious activity to redo if it already existed, but interesting & fascinating to do if it doesn't.

Many of the books & lists I've seen are backwards, starting with the least-urgent plays first.

Instead, the last chapters should be the tiniest and meaningleast things like filling dame.

That's why cards would be good, you could shuffle them up

Sensei lists or research articles start with tiniest followers. Books, other than Mathematical Go Endgames and research by CGT writers, avoid infinitesimals and dames, and do not start with the tiniest fractions!

Let' take an example:

Click Here To Show Diagram Code

[go]$$
$$ | . . . . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . X . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . . X . . . . . . . .
$$ | . . X . O . O O O . . .
$$ | . . X . . . . . . O . .
$$ | . . X X O O . . . . . .
$$ | . . a O . . . . . . . .
$$ -------------------------[/go]

What is the value of a yose move at "a".
My view is that an accurate calculation is a very boring job and in addition this accurate value has no real meaning because the hypothesis made for this calculation (ideal environment without ko or ko threat) cannot be really fullfilled.
In practice I estimate the value at about 3 points (may be a little more) and, as amateur, I am happy with this.
What would be your practice here?

This problem is difficult because lengths of sequences worth playing successively so endgame types incl. ko are not obvious and there are even alternative options. In practice, available time is relevant. With a few minutes time and without an obvious komaster, I would ignore the big ko option and check whether the moves gain about 3 to justify a sente hypothesis.

It always seems more reliable to me when one can figure out the value top-down instead of bottom-up; that's an argument in favor of cards. Intuitively it's more than 2 points since black can capture a white stone after playing "a". Is it closer to 2.5 than 3? I think since the move following after black "a" is 1 2/3 sente or 4 point gote (assuming black follows up with the sente) the value in the diagram is 2 5/6. I don't know if it's better to think of this as 2.5 or 3 points. Maybe it's not accurate but I'm sure it's more important to be consistent than it's to be accurate That is if you can compare two endgames accurately, then it doesn't matter if you can't come up with a precise number.

Besides, coming up with the right move is half the victory. See this problem.

Click Here To Show Diagram Code

[go]$$W white to play
$$ | . X . . . X . . .
$$ | . . X X . . . . .
$$ | . X O X . . . . .
$$ | . X O X O O . . .
$$ | X O O O X O . O .
$$ | . . O X X X O . .
$$ | . . . . . . . . .
$$ +------------------[/go]

If you want to invent a fundamentally new method, you are on your own and it is your burden to prove that / whether it works correctly. It is a matter of seconds to claim a different method but can take years to establish it. Others suggested different methods (e.g., dispensing with fractions) but never proved that they would be working.

When I developed my endgame theory without relying on combinatorial game theory, Bill, who had experience with such endavours, said that I would be on my own and have to prove everything from scratch. I knew and did not fear the prospect. I needed three years to develop and prove my theory.

Have fun with doing alike for yours! Spoiler: you cannot ignore earlier discoveries but must embed them in your theory :)

I guess what I meant is that it's 3 2/3

Click Here To Show Diagram Code

[go]$$
$$ | . . . . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . X . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . . X . . . . . . . .
$$ | . . X . O . O O O . . .
$$ | . . X . . . . . . O . .
$$ | . X X X O O . . . . . .
$$ | . X O O O . . . . . . .
$$ -------------------------[/go]

My point was meant to be that if you define this first position as zero and call the second one +3 or +4 for black then that is an estimate. Someone that is good at these estimates can quickly find a good enough solution to how to approximate the value of the next (local) move. Someone that is not good at it will frequently not be able to make a useful estimate. How can someone be good at estimating values of local positions? There are different ways to make estimates: assumption about follow ups (i.e. lesser of the sente and gote follow ups), counting very conservatively, memorization, experience, and even pure guessing. I have even come up with my own way to make estimates that might require years of investigation (as you correctly hinted at)

Click Here To Show Diagram Code

[go]$$
$$ | . . . . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . X . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . . X . . . . . . . .
$$ | . . X . O . O O O . . .
$$ | . . X . . . . . . O . .
$$ | . . X X O O . . . . . .
$$ | . . X O . . . . . . . .
$$ -------------------------[/go]

I do not understand how you can reach these 2 5/6 or 3 2/3 figures. Can you explain?

Because I agree with you that the move following after black "a" is 1 2/3 sente, I guess the problem is with white to play:

Click Here To Show Diagram Code

[go]$$W
$$ | . . . . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . X . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . . X . . . . . . . .
$$ | . . X . O . O O O . . .
$$ | . . X . . . . . . O . .
$$ | . . X X O O . . . . . .
$$ | . 2 1 O 3 . . . . . . .
$$ -------------------------[/go]

the above sequence is gote for white.

After this sequence if black plays the following move then we reach the following position:

Click Here To Show Diagram Code

[go]$$B
$$ | . . . . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . X . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . . X . . . . . . . .
$$ | . . X . O . O O O . . .
$$ | . . X . . . . . . O . .
$$ | . 1 X X O O . . . . . .
$$ | . X O O O . . . . . . .
$$ -------------------------[/go]

Otherwise if white plays the followimg move we reach the following position:

Click Here To Show Diagram Code

[go]$$W
$$ | . . . . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . X . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . . X . . . . . . . .
$$ | . . X . O . O O O . . .
$$ | . 2 X . . . . . . O . .
$$ | . 1 X X O O . . . . . .
$$ | 3 X O O O . . . . . . .
$$ -------------------------[/go]

I assume it will later follow

Click Here To Show Diagram Code

[go]$$W
$$ | . . . . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . X . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . . X . . . . . . . .
$$ | 2 . X . O . O O O . . .
$$ | 1 X X . . . . . . O . .
$$ | . O X X O O . . . . . .
$$ | O . O O O . . . . . . .
$$ -------------------------[/go]

and I assume it will also follow the following exchange

Click Here To Show Diagram Code

[go]$$B
$$ | . . . . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . X . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . . X . . . . . . . .
$$ | X . X . O . O O O . . .
$$ | O X X . . . . . . O . .
$$ | 1 O X X O O . . . . . .
$$ | O 2 O O O . . . . . . .
$$ -------------------------[/go]

The resulting position is

Click Here To Show Diagram Code

[go]$$B
$$ | . . . . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . X . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . . X . . . . . . . .
$$ | X . X b O . O O O . . .
$$ | a X X . . . . . . O . .
$$ | X O X X O O . . . . . .
$$ | O O O O O . . . . . . .
$$ -------------------------[/go]

In this position the value of a move at "a" is not equal to 1/3 because when white win the ko white can play at b and black loses at least ont other point. For me the value of a move at "a" is equal to 1/2.

Taking all that stuff, for the initial position my calculation gives the result 3 5/24 but I am not quite sure

What I am sure is that such calculation is for me too difficult and far too boring in actual game

Gerard, a difficult part is verifying whether White 1 to 3 is worth playing successively.

The value of the follow up after white tenuki seemed too close to the original move. I assumed it would not affect the value much if we are not playing the ko as was your assumption. Maybe I should have beem clearer, I just wrote what I thought which was "it is" not "it is an estimate". I understood you to mean something similar. If white is to tenuki then that is a specific move on the board, one that I assumed is clearly less than the local move anyway, that is we pick the local move before the tenuki move. Granted that in many positions white would actually tenuki, in that case shouldn't one try to work out the whole sequence involving play in different areas instead of this move value? I guess I relay on my (faulty) spider sense in case there are close moves that should be played first, that is why I agreed that cards with endgame values could be useful.

"work out the whole sequence involving play in different areas instead of this move value?"

One can do so if the ensemble of sufficiently large moves is small. Otherwise, we need the local move value to relate it to the environment and must know for how long local successive play is.

Click Here To Show Diagram Code

[go]$$
$$ | . . . . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . X . . . . . . . . .
$$ | . . . . . . . . . . . .
$$ | . . . X . . . . . . . .
$$ | . . X . O . O O O . . .
$$ | . . X . . . . . . O . .
$$ | . . X X O O . . . . . .
$$ | . . a O . . . . . . . .
$$ -------------------------[/go]

BTW Robert, how do you calculate the miai value of such position? Is it too difficult or too time consuming?