How is your positional judgement?

Uberdude · **#121**

TelegraphGo wrote:

Admittedly, I don't know just how lopsided Elf likes to evaluate, but my impression is that passing (basically what playing L2 is) would drop you from 50% to maybe 80%

Elfv2 (which I think is what was used for the historical analysis Bill is using, but not sure) is way more extreme than that, a pass for the first move gives white >98.5%. From a 50-50 position, a 10% change is maybe a 1 point mistake. (80->90% is probably more than 1 point). So when Bill is posting these ~90% games, that's about a 5 point lead to my mind (which is a pretty terrible opening for a pro after <50 moves, but it's not 20!). Indeed for game 12 which Elf gave black 96½%, Tryss showed KataGo thought it was a 7.8 point lead and 68%.

Knotwilg · **#122**

Interesting and confusing: in my Lizzie 007 interface, LZ network and Kata network disagree who's ahead in Game 16. Anyone else has noticed that in their environment?

Uberdude · **#123**

Possibly because KataGo read the 0 komi of your sgf whilst LZ assumed 7.5?

Tryss · **#124**

Probably, because my Katago agree with your LZ

Edit : yup, if you put 0.0 komi, you get your result, but if you change the komi to 7.5, you go back to 77%wr

What's interesting is that with 7.5 komi, Katago think white score is +10, while with 0 komi, it's -1, a difference of 11 points !

Knotwilg · **#125**

Uberdude wrote:

Possibly because KataGo read the 0 komi of your sgf whilst LZ assumed 7.5?

Yes, that seems to be the case. So this is structural? Kata reads the komi stored in the sgf and LZ doesn't, making an assumption of 7,5?

dfan · **#126**

To be pedantic, it's Lizzie that reads the SGF. It certainly tells the komi to KataGo; I'm not sure whether it also tells it to Leela Zero (it's just a GTP command), but it doesn't really matter, since Leela Zero can't do anything with the information, as it has the assumption built in that komi is inherently 7.5. The komi is baked into the Leela Zero (and AlphaZero) model, while KataGo is more flexible (time for some more huzzahs for lightvector!).

Knotwilg · **#127**

LZ-Kata analysis (komi 7,5)

Meta-analysis: I've been participating since game 2 and I'm right-wrong 11-4
"Dwyrin's heuristic" (as coined by me up here) is right-wrong 9-3 and 4 times doesn't apply.

Bill Spight · **#128**

Uberdude wrote:

TelegraphGo wrote:

Admittedly, I don't know just how lopsided Elf likes to evaluate, but my impression is that passing (basically what playing L2 is) would drop you from 50% to maybe 80%

Elfv2 (which I think is what was used for the historical analysis Bill is using, but not sure) is way more extreme than that, a pass for the first move gives white >98.5%.

Well, the cat's out of the bag, so I'm not hiding this.

TelegraphGo's impression is uncannily accurate. According to Elf, accepting the bad shape lost Black 30% to par.

A pass for the first move has an expected cost of approximately 14 pts. OC, the relation between winrates and points is not linear, but it's not too far off until the winrate approaches 0 or 1. A linear relationship would suggest around a 3% difference for each point on the board (48/14). That's actually close to what statistics on Japanese pro games from the 1970s suggested for komi differences.

I am using the published Elf commentaries of GoGoD games. The analyses were done in October, 2018, according to the dates on the SGF files. I thought that Elf v2 came out later than that.

Kirby · **#129**

Is elf less useful with such aggressive swings?

Might as well just have a bit that says “White win” or “Black win”, rather than a percentage or point evaluation..

Bill Spight · **#130**

Kirby wrote:

Is elf less useful with such aggressive swings?

Might as well just have a bit that says “White win” or “Black win”, rather than a percentage or point evaluation..

Well, with the Elf commentaries, which use a humungous number of playouts, I like the added precision of the larger swings when evaluating specific plays.

Precision is not the same thing as accuracy, OC.

And for these positions, I think that the difference between an 85% winrate estimate and a 90% estimate is meaningful.

Kirby · **#131**

Good point.. I suppose similar reasoning gives the only reason I consider Fahrenheit worth considering at all as a measurement of heat. Overall, I like Celsius better, but the difference in temperature is easier to express more precisely with whole numbers in Fahrenheit :-p

I’d still choose Celsius, though.

Bill Spight · **#132**

Kirby wrote:

I’d still choose Celsius, though.

Celsius 233, good book.

Bill Spight · **#133**

My discussion of Elf's commentary for Game 14.

Game 14

Kiyonari Tetsuya, 9 dan (W) vs. Takemiya Masaki, 9 dan, 1998-00-00aj.

After :w34:

Elf estimates White's winrate as 91% (24.4k playouts), inherited from Elf's choice of reply. (I try to have the last play the same as Elf's top choice, but that was impossible in this game.) However, Black, Takemiya, won by 3½ pts.

Click Here To Show Diagram Code: [go]$$Wcm34 White is ahead. $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . . O . X O . . | $$ | . . . , . . . . . X . . . . . X O . . | $$ | . . O . X . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . O . . . . . . . . . . . X O X . | $$ | . . . . . X . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . , . . . . . , . . . . . X O . . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . X . . . . . . X . . . . | $$ | . . . O . . O O . X . . . . . X . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Elf's recommended continuation for :b35:

Click Here To Show Diagram Code: [go]$$Bcm35 Elf's mainline for $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . 0 . . 2 . 8 . . . . . . . . . . . . | $$ | . 9 6 O 1 4 . . . . . . . O . X O . . | $$ | . . 7 3 5 . . . . X . . . . . X O . . | $$ | . . O . X . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . O . . . . . . . . . . . X O X . | $$ | . . . . . X . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . , . . . . . , . . . . . X O . . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . X . . . . . . X . . . . | $$ | . . . O . . O O . X . . . . . X . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Black plays the jump attachment on the third line and then intrudes into the top left corner. White prefers to let Black in instead of letting Black make a territorial moyo on the top side.

Click Here To Show Diagram Code: [go]$$Bcm45 Elf's mainline for , continued $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . O . . O . O . . . . . . . . . . . . | $$ | 2 X O O X O . . . . . . . O . X O . . | $$ | . 3 X X X . . . . X . . . . . X O . . | $$ | . 1 O . X . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . O . . . . . . . . . . . X O X . | $$ | . . . . . X . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . , . . . . . , . . . . . X O . . | $$ | . . . . . . . . . . . . . . . 5 X O . | $$ | . . . . . . . . . . . . . . . . 4 . . | $$ | . . . . . . . . . . . . . . . . . 6 . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . X . . . . . . X . . . . | $$ | . . . O . . O O . X . . . . . X . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Next, White expands his territory on the right side into the bottom right, preventing Black's kikashi against the White group.

Both of these variations have been lurking for some time.

Earlier in the game.

Click Here To Show Diagram Code: [go]$$Bc Sanrensei, slight error $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . a . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . 5 . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Elf's winrate estimate for :b5:

is 7½% less than that for a, making :b5:

a likely minor error.

Click Here To Show Diagram Code: [go]$$Wcm6 Takemiya kosumi, minor errors $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . . 3 . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . 5 . . . . . . . . . . . 4 . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . 2 . . . . | $$ | . . . O . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . 1 . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

In 1976 I played the kosumi response in a 5 stone game vs. a pro in Kyoto. I don't know when Takemiya had the idea. In any event, Elf thinks that :b7:

loses 8% to par, and :b9:

loses 9½%. It believes that each should approach the top left corner. :w10:

encloses the corner. I don't usually report minor errors, but I thought that these were of some interest. By :w10:

Elf reckons that Takemiya has lost a net 20% to par.

Click Here To Show Diagram Code: [go]$$Bcm11 Black tenuki, significant error $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . . O . 3 2 . . | $$ | . . . , . . . . . 1 . . . . . X 4 . . | $$ | . . O . . . . . . . . . . . X . . 8 . | $$ | . . . . . . . . . . . . . . . . 5 6 . | $$ | . . . . . . . . . . . . . . . . a 7 . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . X . . . . | $$ | . . . O . . . . . 9 . . . . . X . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

I was never confident of handling the cut at a, and so avoided the kosumi or blocked at 14. Elf thinks that :b19:

loses 16½% to the solid connection at a. Considering Black's low winrate estimate after :w18:

of 33%, we might consider :b19:

to be a blunder.

Click Here To Show Diagram Code: [go]$$Bcm19 Elf's mainline variation for $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . 3 2 4 . . | $$ | . . . O . . . . . . . . . O a X O . . | $$ | . . . , . . . . . X . . . . . X O . . | $$ | . . O . . . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . . . . . . . . . . . . . . 1 X . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . 9 . . . . X . . . . | $$ | . . . O . . . . . , . . . . . X . . . | $$ | . . . . . 8 . . 7 . 6 . . O . . . . . | $$ | . . . . . . . . . . . . . . 5 . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

After

White lives in the corner with the hane-and-connect. OC, the cut at a is smaller than the cut at 19. The territorial footsweep, :b23:

, is interesting. It does not seem to be in line with the general aim of the sanrensei. After :w24:

extends, not really making a base because of :b23:

,

and

attack the bottom side White group.

Game continuation.

Click Here To Show Diagram Code: [go]$$Wcm20 a blunder? $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . . O . X O . . | $$ | . . . , . . . . . X . . . . . X O . . | $$ | . . O . 0 . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . . . . . . . . . . . . . 2 1 X . | $$ | . . . . . . . . . . . . . . . 4 3 . . | $$ | . . . . . . . . . . . . . . . 6 5 . . | $$ | . . . , . . . . . , . . . . . X 7 . . | $$ | . . . . . . . . . . . . . . . . 8 9 . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . X . . . . | $$ | . . . O . . . . . X . . . . . X . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Elf thinks that :b29:

, a standard reducing play, loses 16% to par. We should probably consider it a blunder, because White now has a winrate estimate of 95½%.

Click Here To Show Diagram Code: [go]$$Bcm29 Elf's mainline variation for $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . . O . X O . . | $$ | . . . , . . . . . X . . . . . X O . . | $$ | . . O . . . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . . . . . . . . . . . . . X O X . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . X O 2 . | $$ | . . . , . . . . . , . . . . . X O 1 4 | $$ | . . . . . . . . . . . . . . . 9 X O . | $$ | . . . . . . . . . . . . . . . . 8 3 . | $$ | . . . . . . . . . . . . . . . . 0 . . | $$ | . . 6 . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . X . . . . | $$ | . . . O . . . . . X . . . . . X . . . | $$ | . . . . . 5 . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . 7 . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Black plays kikashi against the White group on the right side, then makes a territorial moyo on the bottom. :w36:

and

intrude into the bottom right. Note that Black's kikashi means that the :b31:

stone has some aji there.

Click Here To Show Diagram Code: [go]$$Bcm39 Elf's mainline variation for , continued $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O 1 . . 3 . . . . . O . X O . . | $$ | . . . 2 . . . . . X . . . . . X O . . | $$ | . . O . . . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . . . . . . . . . . . . . X O X . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . X O O . | $$ | . . . , . . . . . , . . . . . X O . O | $$ | . . . . . . . . . . . . . . . X X O . | $$ | . . . . . . . . . . . . . . . . O X . | $$ | . . . . . . . . . . . . . . . . O . . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . X . . . . | $$ | . . . O . . . . . X . . . . . X . . . | $$ | . . . . . X . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . X . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Next, Black develops the top side, staring with the third line attachment, :b39:

.

I could have stopped here. However, I do not like to end on a play that Elf considers a mistake, because it typically has very few playouts, and its winrate may not be very reliable. In addition, a series of errors follows, which may be of some interest.

Click Here To Show Diagram Code: [go]$$Wcm30 Dueling blunders $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . . O . X O . . | $$ | . . . , . . . . . X . . . . . X O . . | $$ | . . O . X . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . 1 . . . . . . . . . . . X O X . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . , . . . . . , . . . . . X O . . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . X . . . . | $$ | . . . O . . . . . X . . . . . X . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

looks funny, and indeed, Elf thinks that it loses 16½% to par. It is not unusual to see players trading mistakes of approximately the same size (from Elf's perspective). Perhaps from a mutual blindspot, from shared assumptions or evauations, or from mutual unfamiliarity with such a position. Although in this case Takemiya should be on more familiar territory.

Click Here To Show Diagram Code: [go]$$Wcm30 Elf's mainline variation for $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . 5 8 . . . . . . . . . . . . . | $$ | . . . O 4 6 . . . . . . . O . X O . . | $$ | . . . 7 . . . . . X . . . . . X O . . | $$ | . . O . X . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . . . . . . . . . . . . . X O X . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . , . . . . . , . . . . . X O . . | $$ | . . . . . . . . . . . . . . . 2 X O . | $$ | . . . . . . . . . . . . . . . . 1 . . | $$ | . . . . . . . . . . . . . . . . . 3 . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . X . . . . | $$ | . . . O . . . . . X . . . . . X . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

As we have seen before, White expands his territory on the right side. Then :b33:

plays the jump attachment on the third line against White's enclosure. In this position :b35:

pulls back to secure territory on the top side and then :w36:

secures the corner.

Click Here To Show Diagram Code: [go]$$Wcm30 Dueling blunders, continued $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . . O . X O . . | $$ | . . . , . . . . . X . . . . . X O . . | $$ | . . O . X . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . O . . . . . . . . . . . X O X . | $$ | . . . . . 2 . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . , . . . . . , . . . . . X O . . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . X . . . . | $$ | . . . O . . . . . X . . . . . X . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

looks funny, or impressive, depending on your viewpoint. Elf is not impressed. It thinks that :b31:

loses 15% to par.

Click Here To Show Diagram Code: [go]$$Bcm31 Elf's mainline variation for $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . 6 . . . . . . . . . . . . . . | $$ | . . 0 O 5 8 . . . . . . . O . X O . . | $$ | . . . 7 9 . . . . X . . . . . X O . . | $$ | . . O . X . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . O . . . . . . . . . . . X O X . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . X O 2 . | $$ | . . . , . . . . . , . . . . . X O 1 4 | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . . . . . . . . . . . . . . . 3 . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . X . . . . | $$ | . . . O . . . . . X . . . . . X . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Black takes his kikashi against the right side group and then takes aim at the top left corner, in sequences we have seen before.

Click Here To Show Diagram Code: [go]$$Bcm41 Elf's mainline variation for , continued $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . 4 . . O . 2 . . . . . . . . . . . . | $$ | 6 3 O O X O . . . . . . . O . X O . . | $$ | . . 1 X X . . . . X . . . . . X O . . | $$ | . 5 O . X . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . O . . . . . . . . . . . X O X . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . X O O . | $$ | . . . , . . . . . , . . . . . X O . O | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . . . . . . . . . . . . . . . X . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . X . . . . | $$ | . . . O . . . . . X . . . . . X . . . | $$ | . . . . . 7 . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

In this position :b47:

does not connect, but approaches the bottom left corner from the bottom side.

Back to the game

Click Here To Show Diagram Code: [go]$$Wcm32 Not so bad $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . . O . X O . . | $$ | . . . , . . . . . X . . . . . X O . . | $$ | . . O . X . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . O . . . . . . . . . . . X O X . | $$ | . . . . . X . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . , . . . . . , . . . . . X O . . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . X . . . . | $$ | . . . O . . 1 . . X . . . . . X . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

is not so bad, losing only 9½% to par, according to Elf.

Click Here To Show Diagram Code: [go]$$Wcm32 Elf's mainline variation for $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . 5 . . . . . . . . . . . . . . | $$ | . . 9 O 4 7 . . . . . . . O . X O . . | $$ | . . 0 6 8 . . . . X . . . . . X O . . | $$ | . . O . X . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . O . . . . . . . . . . . X O X . | $$ | . . . . . X . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . , . . . . . , . . . . . X O . . | $$ | . . . . . . . . . . . . . . . 2 X O . | $$ | . . . . . . . . . . . . . . . . 1 . . | $$ | . . . . . . . . . . . . . . . . . 3 . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . X . . . . | $$ | . . . O . . . . . X . . . . . X . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

By now this sequence is no surprise.

Click Here To Show Diagram Code: [go]$$Wcm42 Elf's mainline variation for , continued $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . 3 . . O . 1 . . . . . . . . . . . . | $$ | 5 2 O O X O . . . . . . . O . X O . . | $$ | . . X X X . . . . X . . . . . X O . . | $$ | . 4 O . X . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . O . . . . . . . . . . . X O X . | $$ | . . . . . X . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . , . . . . . , . . . . . X O . . | $$ | . . . . . . . . . . . . . . . X X O . | $$ | . . . . . . . . . . . . . . . . O . . | $$ | . . . . . . . . . . . . . . . . . O . | $$ | . . 6 . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . X . . . . | $$ | . . . O . . . . . X . . . . . X . . . | $$ | . . . . . 7 . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

In this position Black prefers to approach the bottom left corner from the left side.

One more time.

Click Here To Show Diagram Code: [go]$$Wcm30 Black returns the favor, once more $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . . O . X O . . | $$ | . . . , . . . . . X . . . . . X O . . | $$ | . . O . X . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . O . . . . . . . . . . . X O X . | $$ | . . . . . X . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . X O . . | $$ | . . . , . . . . . , . . . . . X O . . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . 4 . . . . . . X . . . . | $$ | . . . O . . O 5 . X . . . . . X . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

returns the favor, losing 10½% to par, according to Elf. Then :w34:

is only a minor mistake. I guess these mistakes are caused by a shared evaluation of this and other areas of the board.

Click Here To Show Diagram Code: [go]$$Bcm33 Elf's mainline variation for $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . 6 . . . . . . . . . . . . . . | $$ | . . 0 O 5 8 . . . . . . . O . X O . . | $$ | . . . 9 7 . . . . X . . . . . X O . . | $$ | . . O . X . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . O . . . . . . . . . . . X O X . | $$ | . . . . . X . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . X O 2 . | $$ | . . . , . . . . . , . . . . . X O 1 4 | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . . . . . . . . . . . . . . . 3 . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . X . . . . | $$ | . . . O . . O . . X . . . . . X . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Familiar territory.

Click Here To Show Diagram Code: [go]$$Bcm43 Elf's mainline variation for , continued $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . 4 . . O . 2 . . . . . . . . . . . . | $$ | 6 3 O O X O . . . . . . . O . X O . . | $$ | . . 1 X X . . . . X . . . . . X O . . | $$ | . 5 O . X . . . . . . . . . X . . O . | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . O . . . . . . . . . . . X O X . | $$ | . . . . . X . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . X O O . | $$ | . . . , . . . . . , . . . . . X O . O | $$ | . . . . . . . . . . . . . . . . X O . | $$ | . . . . . . . . . . . . . . . . . X . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . X . . . . | $$ | . . . O . . O . . X . . . . . X . . . | $$ | . . . . . . . . . . . . . O . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

By this time, other pros had learned how to play against Cosmic Go. Well, sort of. They still made a lot of mistakes. BTW, Takemiya won by 3½ pts.

Gomoto · **#134**

Let's burn all go books. :twisted:

Never mind and don't ask. (I prefer Fahrenheit over Celsius only when considering books.)

Bill Spight · **#135**

My discussion of Elf's commentary on Game 15

Bill Spight · **#136**

My discussion of Elf's commentary on Game 16

xela · **#137**

Uberdude wrote:

Elfv2 (which I think is what was used for the historical analysis Bill is using, but not sure) is way more extreme than that, a pass for the first move gives white >98.5%.

Surely a pass on the first move means that white can also pass and win by komi? So actually the winrate should be 100%, but 98.5% is within the margin of error :-) Do you get the same result if you make an SGF file that has an empty board and "white to play" (so there isn't a previous black pass)?

Knotwilg · **#138**

I've started https://senseis.xmp.net/?BillSpightsElf ... tExercises

If someone wants to help out doing the C&P work, be my fellow!

Bill Spight · **#139**

I posted Game 25 today.

During the holiday season I plan to catch up with the comments. I may post another game, but maybe not.

Enjoy your holidays.

Or, as we say in the US, Happy War on Christmas! :mrgreen:

How is your positional judgement?

Who is online