Endgame: Lost Points

RobertJasiek · **#21**

Uberdude, endgame is concerned with playing the bigger moves. Move 6 plays in the smaller gap and therefore is a smaller move than if played in the wider gap. Since the gaps are of simple nature (one stone at each end of the gap), such a simple consideration of width is valid and good enough to assess that the move is a mistake. Its exact loss value is a matter of further study.

quantumf · **#22**

Sennahoj wrote:

quantumf: well but it's linear right, any factor you apply to the gross needs to apply to the net....

I suppose so. I just meant I was only considering the matter of how many extra points a single player could gain.

Sennahoj wrote:

I also think that the number is too high, but disregarding double counting etc, I think the main reason that it seems so high is that Robert include lots of moves that people wouldn't usually think of as endgame. It seems likely that a very strong player could win consistently against White with 100+ points, that could come from attacking at moments early in the game (where Robert credited points for mistakes).

Largely agreed. Certainly a strong player wouldn't in general bother about endgame when playing a much weaker player, they'd just crush them in a fight. Having said that, why I think there is something to be said for Robert's idea, when I play slightly stronger players (maybe 3 stones stronger), and particularly when I play teaching games against MUCH stronger players, where they're not trying to crush me as soon as possible, I'm generally struck by how well their early or middle games moves work out when it comes to the end game. Somehow their moves which seemed to only have a meaning in the middle game enable them to consistently get better endgame moves. This is not a coincidence, because I too consciously consider the endgame when I play middle games moves (although I would not make this claim for the opening), it's just that stronger players do it better than me.

HermanHiddema · **#23**

Ok, because the whole methodology is interesting, but very poorly defined, lets have a simple example.

Robert, correct me if I'm wrong, but the idea is basically as follows, as I understand it:

Here is a position. It's black's turn. The biggest move is a but black didn't see it.

Click Here To Show Diagram Code: [go]$$B black to play $$ +---------------+ $$ | . . X O . . X | $$ | a . X O O O X | $$ | . X X O . . X | $$ | . X O O O O X | $$ | X X O . . X X | $$ | O O O O X X . | $$ | . O . O X . X | $$ +---------------+[/go]

Here's perfect play

Click Here To Show Diagram Code: [go]$$B end score B+3 $$ +---------------+ $$ | . . X O . 3 X | $$ | 1 . X O O O X | $$ | . X X O . 2 X | $$ | . X O O O O X | $$ | X X O . 4 X X | $$ | O O O O X X . | $$ | . O . O X . X | $$ +---------------+[/go]

Here's actual play:

Click Here To Show Diagram Code: [go]$$B pass, end score B+4 $$ +---------------+ $$ | . . X O . 1 X | $$ | 5 . X O O O X | $$ | . X X O . 2 X | $$ | . X O O O O X | $$ | X X O . 3 X X | $$ | O O O O X X . | $$ | . O . O X . X | $$ +---------------+[/go]

Now, the analysis:

Move :b1:

is a mistake. Now perfect play is:

Click Here To Show Diagram Code: [go]$$B end score: W+1 $$ +---------------+ $$ | . 4 X O . 1 X | $$ | 2 . X O O O X | $$ | 3 X X O . 5 X | $$ | . X O O O O X | $$ | X X O . 6 X X | $$ | O O O O X X . | $$ | . O . O X . X | $$ +---------------+[/go]

So black's mistake costs 4 points.

White's move :w2:

also misses perfect play, and also loses 4 points, because black could revert to the original perfect play.

:b3:

again loses 4 points, because white can revert back to the W+1 diagram.

:w4:

pass loses 5 points (white can move from B+4 to W+1)

:b5:

is perfect play.

So in the end, black lost 8 points in this end game, and white lost 9 points.

Correct?

HermanHiddema · **#24**

RobertJasiek wrote:

Uberdude, endgame is concerned with playing the bigger moves. Move 6 plays in the smaller gap and therefore is a smaller move than if played in the wider gap. Since the gaps are of simple nature (one stone at each end of the gap), such a simple consideration of width is valid and good enough to assess that the move is a mistake. Its exact loss value is a matter of further study.

Robert Jasiek is German, therefore: If you meet someone who is German, that person is Robert Jasiek.

This :w6:

is not endgame by any definition.

Bill Spight · **#25**

RobertJasiek wrote:

Uberdude, endgame is concerned with playing the bigger moves.

Takagawa wrote:

The endgame is fighting strength.

RobertJasiek · **#26**

Herman, White 6 is not endgame, but endgame-like because it can be avoided and corrected by an endgame-like principle. Part of what made me stronger from 4d to 5d was a more liberal and tolerant view of what constitutes or can be played like endgame. Expert system computer go programmers have known that for at least 20 years. (It does not require a German to get this insight.)

Your example is correct (although I have not checked each value). Now, which conclusion do you want to draw for it being poorly defined?

Bill, if you are reading this, which insights did token go reveal?

EDIT: fighting strength and bigger moves are not mutually exclusive:) Also compare GoWorld 41.

Everybody, I forgot to ask: which GoWorld issues have Ishida's value estimates of beginner mistakes?

HermanHiddema · **#27**

RobertJasiek wrote:

Your example is correct (although I have not checked each value). Now, which conclusion do you want to draw for it being poorly defined?

Well, it really wasn't defined at all. I tried to sort of work backward from the posted game example.

You speak of rounding fractional values and mention miai values, but also equate one rank to 14 points, which is a deiri value. My example avoided the issue of sente/gote altogether because it was not clear to me how you handled it.

Bill Spight · **#28**

Click Here To Show Diagram Code: [go]$$B pass, end score B+4 $$ +---------------+ $$ | . . X O . 1 X | $$ | 5 . X O O O X | $$ | . X X O . 2 X | $$ | . X O O O O X | $$ | X X O . 3 X X | $$ | O O O O X X . | $$ | . O . O X . X | $$ +---------------+[/go]

Play...Gain.....Gain from correct play.....Loss

:b1:

.....0.5..........2.5....................2.0

:w2:

.....0.5..........2.5....................2.0

:b3:

.....0.5..........2.5....................2.0

:w4:

.....0............2.5....................2.5

:b5:

.....2.5..........2.5....................0

Total loss by Black: 4.0

Total loss by White: 4.5

RobertJasiek · **#29**

Herman, 14 points = 1 rank is a miai value.

Bill Spight · **#30**

Uberdude wrote:

Here is some evidence your numbers are too high. I was a 2d and played the endgame of a game against a 4k. We swapped colours and did it again. The difference was about 15 points. Now of course there are problems from this like not playing optimally if you are behind or the 4k learning from my first game, but this suggests the point difference in endgame skill is about 3 points per rank, which is a lot less than 14. I would suggest more experiments of this nature to get a better idea of how the difference in endgame skill of different ranks relates to points.

Back in the 1970s a go magazine ran some articles with endgame tests. You could play out a position with an opponent and then compare your result with the pros. I was 3 dan and played against a shodan. I averaged 11 points better than the pro result. We did not switch sides, so maybe I would have gained 11 points that way, too, for a difference of 22 points.

Anyway, this suggests that of the 2 stone difference between us, around 28 points, 11 points was due to the endgame difference. That's a little high, but my endgame is pretty good.

What counts as endgame is not very well defined, but my guess is that around 1/9 of strength difference is attributable to the endgame, on average. That comes to 1-2 points per stone. That is consistent with your results.

HermanHiddema · **#31**

RobertJasiek wrote:

Herman, 14 points = 1 rank is a miai value.

Huh? As I understand it:

Deiri counting counts the swing between "black plays first" and "white plays first", while ignoring gote/sente/tally/etc. So, on an empty board, the deiri value for the first move is twice the perfect komi, which is thought to be 7. (Additional stones of handicap seem to roughly adhere to that 14 points per stone.)

Miai counting counts the gain (net count / local tally) from a move.

IMO, one of the easiest methods to determine the miai value of a move in a position is to replicate it a number of time (e.g. three so-called half-point kos can be used to show that the miai value of a move in the ko is actually 1/3).

So suppose we set up two empty boards. If black plays first on both, he should win by 14 points (summed board values, ignoring komi). If both players play first on one board, the result should be a tie. So the difference of 14 points, divided by the tally of 2 is 7 points per move. So the miai value of the first move is 7.

Right?

Bill Spight · **#32**

HermanHiddema wrote:

RobertJasiek wrote:

Herman, 14 points = 1 rank is a miai value.

Huh? As I understand it:

Deiri counting counts the swing between "black plays first" and "white plays first", while ignoring gote/sente/tally/etc. So, on an empty board, the deiri value for the first move is twice the perfect komi, which is thought to be 7. (Additional stones of handicap seem to roughly adhere to that 14 points per stone.)

Miai counting counts the gain (net count / local tally) from a move.

IMO, one of the easiest methods to determine the miai value of a move in a position is to replicate it a number of time (e.g. three so-called half-point kos can be used to show that the miai value of a move in the ko is actually 1/3).

So suppose we set up two empty boards. If black plays first on both, he should win by 14 points (summed board values, ignoring komi). If both players play first on one board, the result should be a tie. So the difference of 14 points, divided by the tally of 2 is 7 points per move. So the miai value of the first move is 7.

Right?

Suppose that correct komi is 7. Suppose also that the (correct) first play gains 7 points in gote. Then the proper komi for the rest of the board is 0. But that cannot be right. We know that the proper komi for the rest of the board is 7 or almost 7, and that the value of plays gradually reduces, except for some exciting sequences.

Now suppose that the first play gains 14 points in gote. Then we know that the proper komi lies between 14 and 0. Rough estimate: 7. Working backwards, if correct komi is 7, then we estimate that the first play gains 14 points.

RobertJasiek · **#33**

Lost points of endgame(-like) mistakes for the sake of assessing potential rank improvement. Not counting mistakes that were corrected with delay. Avoiding double counts due to overlapping / changing locales as far as possible. Therefore the totals are smaller than the sum of all losses in terms of per move values.

Total:

Black = 50 points = 3.6 ranks
White = 35 points = 2.5 ranks

The score is greater because Black made more other mistakes especially during the opening.

1 rank = 14 points.

Rank improvement refers to real world ranks, not to server ranks. Rank improvements must be put in relation to the fraction of scored games versus early resigned fighting games.

Endgame: Lost Points

Who is online