It is currently Thu Mar 28, 2024 2:16 pm

All times are UTC - 8 hours [ DST ]




Post new topic Reply to topic  [ 167 posts ]  Go to page Previous  1 ... 3, 4, 5, 6, 7, 8, 9  Next
Author Message
Offline
 Post subject: Re: Values of moves
Post #101 Posted: Tue Sep 18, 2018 1:27 pm 
Lives in gote

Posts: 388
Liked others: 416
Was liked: 198
Bill Spight wrote:
Actually, O's method of calculating territory is the traditional one. He relies on a 50-50 split for gote and privilege for sente. :)


It just occurred to me: this method of calculating value of moves by averaging all possible outcomes, is in fact very similar to what MCTS is doing in computer-go.
Except that humans average over the outcome of moves in number of points (trying to come up to a value of the move measured in points), while MCTS averages out over win/loss ratio, trying to come up with win-rate.

Aren't they similar?

_________________
Sorin - 361points.com

Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #102 Posted: Tue Sep 18, 2018 3:05 pm 
Gosei
User avatar

Posts: 1639
Location: Ponte Vedra
Liked others: 642
Was liked: 490
Universal go server handle: Bantari
Not sure this helps, but here is my take on it:

Short version: There is no free lunch.

Long version: Read on.

1. As somebody said you cannot have numbers without calculating. You don't want to have to count. So this is out.
2. Its seems that what you are willing to settle on is some kind of general designation of Huge, Large, Big, Small, Nothing - or whatever.

So far so good? Ok.

Now, the values of Huge, Large, Big, etc will vary throughout the game. For example - a move which is Small in fuseki, might be Huge in yose. Right? So, it seems pretty useless to generically define how do we designate a move as Huge, Small, whatever.

It seems to me that what you are REALLY after is some kind of way to organize the candidate moves in a given position by approximate value - like this is the biggest (lets call it Huge), this is smaller (call it Big), etc.

So far so good? If not, please feel free to interrupt at any point.

The above can be distilled - for simplicity - to a situation in which you have two candidate moves and want to determine which is 'bigger'. If you have more, then the problem is either longer (you keep comparing pairs until you figure out a winner), or you compare triplets, quadruplets, or all of them. Either way, lets look at the simple case of comparing two moves.

The way I see it, there are several ways to do that:

1. Counting. You might take into account sente and other features, but it basically comes down to counting. You don't want to do that.

2. Brute force. You reconstruct the remaining game tree and see what leads to win and what does not. This is sometimes possible, but usually not, and even when it is, it is often harder and more complex that counting by using any of the simple(r) heuristics we usually use.

3. Experience. You eyeball the options, draw on your experience, and make an educated guess. The stronger you are, the better your estimate. But without counting, this will only get you so far and no further. I find myself doing that a lot, but then I also find myself having to count and calculate more often than I'd like to admit. Guess I am not very experienced. But its fun. Just have to be OK with losing a lot.

4. WAG. You flip a mental coin and hope for the best. This is easy, right? But this will get you nowhere. Might as well try some other game. WoW is pretty cool, heh.

So where does it leave us? Well... we probably should just accept the fact that Go is a game in which counting and calculating is crucial. One can actually call it a pure counting game. You can substitute some of that by experience, but this way you put a narrow cap on your progress if you try doing it exclusively.

I guess what I am trying to say here is that there is no free lunch, no magic wand. If you want to be able to say "this move is worth 5 or BIG and that is worth 3 or SMALL" then you will have to count. The more you are OK with approximation, the more sloppy your counting can be. But you will really quickly find yourself in a situation where your statement will be "this move is 5+/-3 and that move is 3+/-5" which really tells you nothing, might just do the WAG thingie.

So play a lot, get more experience, but also learn to count as well and as exact as you possibly can. This is really what this game is all about.

_________________
- Bantari
______________________________________________
WARNING: This post might contain Opinions!!


This post by Bantari was liked by: daal
Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #103 Posted: Tue Sep 18, 2018 3:28 pm 
Lives in gote

Posts: 553
Liked others: 61
Was liked: 250
Rank: AGA 5 dan
Here is the simplest sente or reverse-sente situation I can think of. Let's see if we can agree on how this is evaluated. No mathlish involved (I hope).

    W to play (reverse-sente) can create a final position worth 0 points. It is then B turn to play, somewhere else on the board.
    B to play (sente) can create a final position worth 2 points after W replies. It is then B turn to play again, somewhere else on the board.
    (The B move, which we are calling sente, leaves a follow-up move worth 100 additional points. This is so much larger than anything else on the board that W will certainly respond locally to prevent it. This justifies treating the B move as sente.)

Questions:
1) What is the value (count) of the initial position (I believe the answer is 2)
2) What is the value of the W reverse-sente move? (I suspect the answer is 2)
3) What is the value of the B sente move? (I know Bill's answer is 0)

Bill or Robert, are these indeed the values you would calculate?
John, would O Meien also calculate these same values?

Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #104 Posted: Tue Sep 18, 2018 4:28 pm 
Honinbo

Posts: 10905
Liked others: 3651
Was liked: 3374
mitsun wrote:
Here is the simplest sente or reverse-sente situation I can think of. Let's see if we can agree on how this is evaluated. No mathlish involved (I hope).

    W to play (reverse-sente) can create a final position worth 0 points. It is then B turn to play, somewhere else on the board.
    B to play (sente) can create a final position worth 2 points after W replies. It is then B turn to play again, somewhere else on the board.
    (The B move, which we are calling sente, leaves a follow-up move worth 100 additional points. This is so much larger than anything else on the board that W will certainly respond locally to prevent it. This justifies treating the B move as sente.)

Questions:
1) What is the value (count) of the initial position (I believe the answer is 2)
2) What is the value of the W reverse-sente move? (I suspect the answer is 2)
3) What is the value of the B sente move? (I know Bill's answer is 0)

Bill or Robert, are these indeed the values you would calculate?
John, would O Meien also calculate these same values?


1) What is the value (count) of the initial position? 2 pts. (for Black)
2) What is the value of the W reverse-sente move? 2 pts.
3) What is the value of the B sente move? Assuming that White's reply gains 100 pts., 100 pts. :o :cool:

----
N. B. It is the disparity between the White and Black moves that allows you to tell that the position is not gote. If it were, the Black move would gain 51 pts., but the White reply would gain 100 pts. That does not make sense. (Unless Black has a large enough reply to White's reply. Which she does not in this case.)

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #105 Posted: Tue Sep 18, 2018 5:12 pm 
Honinbo

Posts: 9545
Liked others: 1600
Was liked: 1711
KGS: Kirby
Tygem: 커비라고해
Quote:
It seems to me that what you are REALLY after is some kind of way to organize the candidate moves in a given position by approximate value - like this is the biggest (lets call it Huge), this is smaller (call it Big), etc.


It wasn't your main point, but this part of your post stuck out as very fascinating to me.

We have established various methods for getting somewhat precise values of moves, even to fractions of a point. There are lots of articles on SL on this, and some experts here on L19. Because of these precise methods to determine the values of various moves, it's possible to establish an ordering of endgame moves. Yada yada yada.

Sometimes these calculations can be pretty complex.

My hypothesis, then, is that there exist methods to simplify the math behind calculating precise move values, at the cost of losing the precision we have with current counting methods.

But perhaps it's possible to lose accuracy, while maintaining the ability to compare the RELATIVE (not precise) value of moves, which is what we really care about anyway.

Maybe it's not possible, but my hunch is that it should be.

_________________
be immersed


This post by Kirby was liked by: daal
Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #106 Posted: Tue Sep 18, 2018 5:14 pm 
Lives in gote

Posts: 553
Liked others: 61
Was liked: 250
Rank: AGA 5 dan
Bill, I was sure you were going to say that the sente move is worth zero. Since the value of the initial position is 2, and the value of the final position (after B sente and W forced response) is still 2, then the value gained by the transaction was 0. Of course B did not have to expend an extra move in this transaction, so perhaps we should calculate the value per move as 0/0 ?

If we make the (incorrect) assumption that all moves are gote, with (incorrect) probability 1/2 that either side will play first, then by averaging all possible results, the initial position is worth 25+ points, and the first move is also worth 25+ points (since it converts a 25+ point position into a 50+ point position). I agree that the absurdity of these values is what disproves the gote assumption.

Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #107 Posted: Tue Sep 18, 2018 6:14 pm 
Honinbo

Posts: 10905
Liked others: 3651
Was liked: 3374
mitsun wrote:
Bill, I was sure you were going to say that the sente move is worth zero. Since the value of the initial position is 2, and the value of the final position (after B sente and W forced response) is still 2, then the value gained by the transaction was 0. Of course B did not have to expend an extra move in this transaction, so perhaps we should calculate the value per move as 0/0 ?


If X - Y = 0, does that mean that (X+Y)/2 is indeterminate? ;)

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #108 Posted: Tue Sep 18, 2018 10:23 pm 
Lives in gote

Posts: 426
Liked others: 186
Was liked: 191
I have seen Myungwan Kim and Howard Landman try to teach this to general go players with mixed results. Partly I think it is because it takes time to get used to. Many players seem to think it is impractical, but it is just practice, I think.

Sometimes I take a close pro game and calculate as many of the late endgame moves as I can. Patterns emerge. You see things that look like a common miai 2.00 situation, but with just a wee follow up, so you see if it gets played before smaller moves. There is a lot of poking into rooms of various sizes, but in real games the rooms are not as well defined as in the SL examples, and some of the rooms have weak walls. These are hard to count. I do not attempt it, but rather notice when the pro gets to them. I work from the end. Once I get back to miai 3.0 or larger it becomes hard. The tactical reading is too difficult for me and I often do not understand why pros choose one large move over another.

It helps to memorize some common values. Not as much as van Zeijst, but some is better than none. Just 10 or so will do to get started.

I don’t do it enough and I cannot recognize the patterns quickly enough yet. I have limited time. But I think it is something that can be practiced.

Also, take the winning side of a won game against an AI and see if you can win. Also do this as close to the end of the game as makes sense for your level. Stronger players can back up further.

Thermographs look cool, but I am not sure if they would excite non-math types. It is worth a try, though.


This post by Calvin Clark was liked by: Bill Spight
Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #109 Posted: Tue Sep 18, 2018 10:38 pm 
Honinbo

Posts: 10905
Liked others: 3651
Was liked: 3374
Calvin Clark wrote:
Once I get back to miai 3.0 or larger it becomes hard.


Way back when, I noticed that in professional 20th century games, at the time limits in use, the miai 3 pt. level was where top pros started to make mistakes. There were inaccuracies with smaller plays, but they rarely cost the game.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #110 Posted: Tue Sep 18, 2018 11:48 pm 
Judan

Posts: 6087
Liked others: 0
Was liked: 786
mitsun wrote:
W to play (reverse-sente) can create a final position worth 0 points. It is then B turn to play, somewhere else on the board.

B to play (sente) can create a final position worth 2 points after W replies. It is then B turn to play again, somewhere else on the board.

(The B move, which we are calling sente, leaves a follow-up move worth 100 additional points.[...]

1) What is the value (count) of the initial position (I believe the answer is 2)
2) What is the value of the W reverse-sente move? (I suspect the answer is 2)
3) What is the value of the B sente move? (I know Bill's answer is 0)

Bill or Robert, are these indeed the values you would calculate?


I take it that you mean that, after the B move creating the intermediate position being a local gote, the follow-up move is a 100 points gain for the next moving player.

***

1) The count of the initial (local endgame) position is 2. This is the inherited count 2 of the black sente follower created after his sente sequence.

2) The value of the white reverse sente move is 2. This value is the sente move value (and equals the gain of the white move).

Calculated as the sente move value, it is the difference value of the count 2 of the black sente follower (created after Black's sente sequence) and the count 0 of the white reverse sente follower (created after White's reverse sente move). Therefore, we calculate 2 - 0 = 2.

Calculated as the gain of the white move, it is the difference of the count 2 of the initial position and the count 0 of the white reverse sente follower (created after White's reverse sente move). Therefore, we calculate 2 - 0 = 2.

The two calculations are equal because the count 2 of the initial position equals the count 2 of the black sente follower.

3) The value of the black sente move is its gain and does NOT equal the sente move value. The gain of the black sente move is the difference of the count of the intermediate position created by it and the count of the initial position. We do not know the count of the intermediate position yet but we already know the count 2 of the initial position.

The intermediate position is a local gote, whose count is calculated as the average of its followers. Its black follower is created after Black moves twice successively from the initial position; its white follower is the initial position's black sente follower (created by the sente sequence started by Black and concluded by White's immediate reply). Since you have not specified an absolute count of its black follower, we need a variable, say A, for the count of the intermediate position. You say that a move in it gains 100 points for the next moving player. Therefore, the count of its black follower is (A + 100) and the count of its white follower (which is the initial position's sente follower) is (A - 100). We subtract when White moves and gains because Black incurs a loss. White's gain is Black's loss.

Since the intermediate position is a local gote, its gote count A is calculated as the average of the counts of its followers. We have the count (A + 100) of its black follower and the count (A - 100) of its white follower. Therefore, we can now calculate the count A of the intermediate position as the average:

A = ( (A + 100) + (A - 100) ) / 2 = (A + 100 + A - 100) / 2 = ( 2 * A ) / 2 = A.

Surprise. We have learnt nothing new for the value of A.

We can, however, derive the numerical value of A from the count 2 of the sente follower. To do so, we use the gain of the white move played in the intermediate position and creating the sente follower. We know that the gain of the white move is 100 (a positive number because White gains points when playing his move). White's gain is Black's loss. We view the count A of the intermediate position from Black's perspective. When White gains 100, Black loses 100. Therefore, White's move from the intermediate position to the sente follower decreases the count.

However, we want the converse: the impact of undoing White's move. The count A of the intermediate position is larger than the count 2 of the sente follower. It is larger by the amount 100 of the gain. Therefore, to the count 2 of the sente follower, we add the gain 100 to calculate the count A of the intermediate position:

A = 2 + 100 = 102.

Next, we can calculate the gain of the black move in the initial position. The gain is the difference of the count A = 102 of the intermediate position and the count 2 of the initial position. The gain of the black move in the initial position is 102 - 2 = 100. This is the value of the black move considered as a single move.

4) Let us study the sente sequence. Black's first move gains +100, then White's reply lets White gain 100 or Black lose 100 so the impact of White's reply is -100. Now, we can calculate the 'net profit' of the sente sequence by summing up the impacts of its moves. The net profit of the sente sequence is +100 - 100 = 0.

Where informal descriptions say "sente gains nothing", what they should mean is: the sente sequence of a local sente endgame has the net profit 0.

It is NOT the single first black move played in the initial position that you might have guessed to have the "value" 0 (what value?). Instead, it is the net profit of the sente sequence that has the value 0. The sente sequence is started by the single first black move but this does NOT mean that the move itself would have the value of the whole sente sequence.

If Black's first move is ignored, the local impact is a change of +100. If Black's first move is replied, the local impact of the sente sequence is 0.

Code:
             2
            / \
      +100 /   \ -2
          /     \
         /       \
      A = 102     0
       /\          ^
      /  \         |
+100 /    \ -100   |
    /      \       |
   /        \      |
202          2 <--- move value calculated from these two counts as 2 - 0 = 2


The leftwards / rightwards lines are Black's / White's moves. The numbers next to these lines are the gains / losses from Black's value perspective. The numbers at the nodes linked by these lines are the counts of the positions. The arrows in the lower right of the diagram are commentary only. The sente sequence is from 2 via A to 2.


This post by RobertJasiek was liked by: mitsun
Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #111 Posted: Wed Sep 19, 2018 12:17 am 
Dies in gote

Posts: 46
Liked others: 109
Was liked: 34
Rank: Euro 1 dan
GD Posts: 7
RobertJasiek wrote:
The value of the black sente move is its gain and does NOT equal the sente move value


So "value of the black sente move" =/= "sente move value"? If this is what is actually meant, then the terminology is too confusing.


This post by zermelo was liked by: Waylon
Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #112 Posted: Wed Sep 19, 2018 12:28 am 
Honinbo

Posts: 10905
Liked others: 3651
Was liked: 3374
zermelo wrote:
RobertJasiek wrote:
The value of the black sente move is its gain and does NOT equal the sente move value


So "value of the black sente move" =/= "sente move value"? If this is what is actually meant, then the terminology is too confusing.


It is confusing. Go players traditionally call a sente with a reverse sente value of 2 pts. a 2 pt. sente. Ordinary language is flexible enough that you can make sense of that, but if you try to get precise, it is misleading. A so-called 2 pt. Black sente is one where White's reverse sente play gains 2 pts., but Black's sente play gains more than 2 pts. Usually that's all we need to know about how much Black's sente gains. We don't need to know the exact number of points. In this case mitsun asked.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #113 Posted: Wed Sep 19, 2018 12:52 am 
Judan

Posts: 6087
Liked others: 0
Was liked: 786
zermelo wrote:
So "value of the black sente move" =/= "sente move value"? If this is what is actually meant, then the terminology is too confusing.


Yes. This is a consequence of tradition. Tradition would speak of "2 points in sente for Black" and "2 points in reverse sente for White".

Now that the theory has become clear, we find the confusion you mention.

However, Bill and I have already offered the solution: we speak of the GAIN of Black's first move. Bill would say: Black's first move GAINS 100 points.

Instead of considering a move value of a move, we might speak of a move value in a position. The initial local endgame position is a local sente. Its sente move value describes how much White gains by playing reverse sente, or how much Black prevents White from gaining.

Actually, "value of the black sente move" is an informal phrase - not a term. "sente move value" is a term. We need not use the informal phrase, except maybe for our initial learning when the confusion hits us the most.

Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #114 Posted: Wed Sep 19, 2018 3:24 am 
Oza
User avatar

Posts: 2508
Liked others: 1304
Was liked: 1128
Bantari wrote:
3. Experience. You eyeball the options, draw on your experience, and make an educated guess. The stronger you are, the better your estimate. But without counting, this will only get you so far and no further.
This is essentially my method, and indeed I have the feeling that I can only get so far with it. I started this thread not to delve into the mathematics of endgame, but because I often found myself in a situation in which I saw four or five possible moves and had nothing at my disposal to help me decide which to choose.

A lot of this discussion does not pick me up where I am. Where I am is not knowing when to play a sente, and not having a rough knowledge of its value or urgency compared to playing a big gote. I realize that what Bill et al. are explaining does address these questions, but I think my lack of understanding starts way earlier.

I also have to say that my feelings towards the density of technical concepts and terms within an English sentence are similar to John's. To me, much of this conversation sounds to me like the ramblings of homeless men wandering the streets, and it evokes in me a similar reaction. I personally was something of a math whiz in high school, but since then my capabilities have declined significantly and logical reasoning is not my strong suit. If A is bigger than B and B is the uncle of C then I don't give a damn.

It makes sense that to express mathematical concepts, using the appropriate terminology is the best way to do it, and I agree that it is not really too much of a burden to learn the meaning of a few terms. What I question is the extent to which such concepts are necessary to talk about go. Somewhere in this thread was a comment by O stating that all that is needed is 6th grade math. I think I can handle sixth grade math. The problem isn't the math, it's that I can't figure what in Sam's hill you people are talking about.

Compare this:

Quote:
If the move is truly sente for W, the probability of W playing first becomes 100%, as does the probability of B answering...

with this:

If the move is truly sente for white, then white will definitely play there first. It is also the case that Black will definitely answer white's move.

I find the second version easier to understand. Some of you probably not. In this case, for this part of the sentence at least, I could figure out what Mitsun was talking about. In other cases, for example:

Quote:
If you have some number, N, of a position such that the total score, S, of all of them together is the same, regardless of who plays first (they are miai), then the average value of each of them is S/N. That works for gote positions. :)


are beyond my ability to rephrase into something I can understand. I assume that I wasn't the intended audience for the above sentence, so it isn't really a problem. I also understand that some of the thinking around the topic of which move is better is in fact quite complicated, and for those that can interpret and understand the situation mathematically, this is indeed the optimal mode of expression. Provided that their audience can follow what they are saying.

The video by Kyle Blocher that I mentioned earlier is a good example of someone making an effort to explain mathematical go concepts to non-mathematicians. He took things slow. He made sure that people understood one concept before he went onto the next one. He encouraged the participants to ask questions, and he asked them to explain their reasoning when they spoke up. After watching the video one time, I was able to explain how miai counting works in a simple gote position. I also understood his explanation of how to evaluate a simple ko. I think this is because the speaker geared his talk to his audience.

Here of course the situation is not so clear. Sometimes the mathies begin by explaining something to the non-mathies and we non-mathies bounce about like dogs when their owners come home, but then often another mathie comments or asks a clarifying question and then off they go to jargon world leaving us to sulk on the couch.

It's not that I am not interested. For example Bill writes:

Quote:
In 2000 I presented a paper that defined sente and gote in terms of (non-ko) thermographs. It's easy. :D The top of a gote thermograph looks like this:

Code:
          ^
          |
         / \


The vertical line at the top is called the mast. The ^ indicates that it goes up forever.


My first thought is "great!" Bill is going to help me understand sente and gote by offering some kind of visual explanation. My second thought is: "how does this image relate to stones on a board?" and I no longer feel optimistic that this will help me play go any better.

I know that you guys are making an effort to help us and I imagine that you must find some of our comments just as baffling as we do yours. I for example am not even sure what question I want to ask. It's really no surprise that I have difficulty understanding your answers.

Quote:
So play a lot, get more experience, but also learn to count as well and as exact as you possibly can. This is really what this game is all about.
I have taken several stabs at it, but typically a point comes in which I no longer understand the material. Even in the video that I praised above, at about 28:30, Kyle says: "I have about ten minutes or so, and I really want to get across this concept of sente, because then you can see what sente is all about and understand how to classify moves in certain ways." I had high hopes for these ten minutes, but sadly even he lost me. Maybe I have to accept that without toughing up for a mathematical approach, some concepts will just remain beyond me.

_________________
Patience, grasshopper.


This post by daal was liked by: Waylon
Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #115 Posted: Wed Sep 19, 2018 3:29 am 
Honinbo

Posts: 10905
Liked others: 3651
Was liked: 3374
RobertJasiek wrote:
Now that the theory has become clear, we find the confusion you mention.

However, Bill and I have already offered the solution: we speak of the GAIN of Black's first move. Bill would say: Black's first move GAINS 100 points.


OC, O Meien and others who use miai counting also talk about how much a move gains (on average). That's the whole point of O's "value of one move". :)

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Top
 Profile  
 
Offline
 Post subject:
Post #116 Posted: Wed Sep 19, 2018 3:40 am 
Honinbo
User avatar

Posts: 8859
Location: Santa Barbara, CA
Liked others: 349
Was liked: 2076
GD Posts: 312
Hi daal,
Quote:
Maybe I have to accept that without toughing up for a mathematical approach, some concepts will just remain beyond me.
Good teaching materials, good writing, and good explanations are very difficult to come by. ( I wish there's a hard copy of Mr. O's book. )
I thought there are various "ask questions" websites where the replies are voted up or down... maybe this is an example.

3Blue1Brown puts in a tremendous amount of work to make his ridiculously excellent lectures:
Another example when 'infinite intelligence' meets a top presentor:

Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #117 Posted: Wed Sep 19, 2018 3:42 am 
Oza
User avatar

Posts: 2408
Location: Ghent, Belgium
Liked others: 359
Was liked: 1019
Rank: KGS 2d OGS 1d Fox 4d
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
What we may have here, in terms of Geoffrey Moore's classic "Crossing the chasm", is a chasm between visionaries like Bill and Robert and the rest of us, who represent an early majority, with John as a peculiar member because he may be a forerunner of the late majority or his interest is somewhat different from ours, as a professional translator of go material.

Bill & Robert have taken mathematics, applied it to the endgame of Go and have created what we may call "end game theory based on mathematical approach". It not only includes mathematical devices such as value trees and statements in formal logic, it also defines new terms as "count", "value" and "gain". Their aim and interest is precise calculation and unambiguous description.

The early majority sees value in their "product" because it answers our needs and interests: being able to make the best possible decision in a game (the practical level) and also to have reference material with which to enhance our understanding (more on the cultural, conceptual level).

The "chasm" to cross is adopting the new terminology, grasping the concepts behind it and finding out how it will lead to better decision making. We have to take Robert's word that he makes better decisions now that he understands the theory. However, we don't want to "be Robert" i.e. we don't want to go through the same hardships, we want to benefit from the same knowledge but with considerably less investment.

I see three hurdles when trying to cross the chasm: the math, the new concepts, some of which verbally overlap with old ones, and the high degree of precision which is held by the current way of articulating things. Of these, I believe the concepts should remain: the definitions of count, value, gain ... counterintuitive as they may be. As Robert points out, precisely the confusion they create when holding onto tradition is the progress we're looking for. The math may have to go, or should not be intertwined with the text so as to represent the mathlish hurdle. The biggest problem may lie in letting go some degree of the precision reached by Bill & Robert in favor of practicality.

I don't know how to do this. I would have to read through the visionaries' writings and see where there is scope for simplification to cross the chasm. I would have to interview potential readers about how much conceptuality and math they are willing to handle. I would have to check all the assumptions above.

However, Robert may already have done most of that work. He sees the value and is a Go author. What he may not yet have done, is understanding his potential audience and what how he has to shape his writings so as to cross the chasm from talking to fellows like Bill in research lingo, to writing to an audience with a high interest but different needs and different capacity.

Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #118 Posted: Wed Sep 19, 2018 3:54 am 
Honinbo

Posts: 10905
Liked others: 3651
Was liked: 3374
daal wrote:
A lot of this discussion does not pick me up where I am. Where I am is not knowing when to play a sente, and not having a rough knowledge of its value or urgency compared to playing a big gote. I realize that what Bill et al. are explaining does address these questions, but I think my lack of understanding starts way earlier.

I also have to say that my feelings towards the density of technical concepts and terms within an English sentence are similar to John's. To me, much of this conversation sounds to me like the ramblings of homeless men wandering the streets, and it evokes in me a similar reaction.


Those people aren't ranting to themselves, they are talking on their cell phones. :lol:

Quote:
It makes sense that to express mathematical concepts, using the appropriate terminology is the best way to do it, and I agree that it is not really too much of a burden to learn the meaning of a few terms. What I question is the extent to which such concepts are necessary to talk about go. Somewhere in this thread was a comment by O stating that all that is needed is 6th grade math. I think I can handle sixth grade math. The problem isn't the math, it's that I can't figure what in Sam's hill you people are talking about. . . .

Quote:
If you have some number, N, of a position such that the total score, S, of all of them together is the same, regardless of who plays first (they are miai), then the average value of each of them is S/N. That works for gote positions. :)


are beyond my ability to rephrase into something I can understand.


Did you understand when I showed that two copies of a corner from O's book had a combined score of 4 pts. for Black, no matter who played first? Each corner had an average value of 2 pts. Was that clear?

Quote:
It's not that I am not interested. For example Bill writes:

Quote:
In 2000 I presented a paper that defined sente and gote in terms of (non-ko) thermographs. It's easy. :D The top of a gote thermograph looks like this:

Code:
          ^
          |
         / \


The vertical line at the top is called the mast. The ^ indicates that it goes up forever.


My first thought is "great!" Bill is going to help me understand sente and gote by offering some kind of visual explanation. My second thought is: "how does this image relate to stones on a board?" and I no longer feel optimistic that this will help me play go any better.


I have written something about thermographs in This 'n' That. Maybe I should do some more. :)

Quote:
Even in the video that I praised above, at about 28:30, Kyle says: "I have about ten minutes or so, and I really want to get across this concept of sente, because then you can see what sente is all about and understand how to classify moves in certain ways." I had high hopes for these ten minutes, but sadly even he lost me. Maybe I have to accept that without toughing up for a mathematical approach, some concepts will just remain beyond me.


The concept of local sente gets tricky when it entails a sequence of plays. But if you take the simple prototype where the reverse sente is one move and the sente threat is a simple gote, then a position is sente for Black when White's reply to Black's play gains more than the White's reverse sente. That means that normally Black will be able to make her play with sente before White can afford to play the reverse sente. Does that make sense?

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #119 Posted: Wed Sep 19, 2018 4:06 am 
Judan

Posts: 6087
Liked others: 0
Was liked: 786
As long as everything is simple gote without follow-ups, understanding it is easy and sorting move values is easy. Gote with one follow-up move is maybe half as easy.

Already a simple sente is much less easy, unless you ignore most of its behaviour and pick only one aspect.

Local endgames with several follow-ups or combinations of several local endgames (of which at least some are not simple gotes) are much more difficult than simple gotes. This is so because there is no simple, general move order. The selection of local endgames and the values of follow-up moves or positions decide which kind of decision-making applies correctly. Some selections are solved, others work very well with good approximations (not guesswork, but using calculated values), some are complex and not understood yet.

The jump from understanding simple gotes to understanding other selections is steep. Effort is required. Attempts to avoid effort and resorting to guesswork or half-understanding are bound to fail quickly. Maybe they are good for creating an illusion of manageability but what they really do is talking away the necessary effort.

Even simple combinations of follow-ups can quickly become complex. Nevertheless, theory is very good at managing most situations well. Theory, for which the learning effort is roughly 10 times that for simple gotes. We can avoid application of advanced theory, which is 100+ times as demanding.

Top
 Profile  
 
Offline
 Post subject: Re: Values of moves
Post #120 Posted: Wed Sep 19, 2018 4:30 am 
Judan

Posts: 6087
Liked others: 0
Was liked: 786
Knotwilg, when writing, one of my major questions to myself was: should I emphasise or deemphasise rounding? When I tried early rounding, I ran into problems quickly. Therefore, usually, I have used exact calculations, shortly described rounding and left it to the learner to round when and as much he wants. He can then compare to my precise calculations to verify whether his calculation with rounding produces the same (but rounded) results.

In endgame evaluation, rounding errors cannot just add up but they can also cause wrong comparisons, assignments of (gote / sente etc.) types and multiply errors in follow-up calculations. When values are very different, we need not calculate at all (visual guessing is good enough). We calculate to distinguish and order similar values, and for them rounding prevents good comparisons quickly and the effort of managing rounding margins (to be sure not to create conflicts) easily exceeds the effort of calculating accurate fractions.

As said before, rounding can sometimes make sense for deep iterative follow-ups when the impact of rounding errors is significantly smaller than 1 point for a move value.

Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 167 posts ]  Go to page Previous  1 ... 3, 4, 5, 6, 7, 8, 9  Next

All times are UTC - 8 hours [ DST ]


Who is online

Users browsing this forum: Google [Bot] and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group