Life In 19x19

I recently watched Dwyrin's video on that game. It's a game played between the first two AlphaGo iterations, if I'm not mistaken, so AI hadn't really gotten into it (neither moves nor commentary).

I also put the game through a mobile Leela (64b network, I believe), for a couple of hours, giving me between 10 and 20 thousand playouts around the moves here. IME, weaker networks might take a while to realize a move was game changing, and their graphs might have slightly different shapes, but their overall evaluation is not that different (and my PC has been dying for a while).

Spoilers from now on...



Now, the video implies that W is a difficult position from the very beginning. I don't doubt that, but except for a tiny dip around MV 25, Leela considers White is ahead the whole game. I find it quirky, considering bots have a reputation for territory.

But my main point of interest is around Tengen and MV 90.

There's a group right there, a Black 4 stone L shape, that gets netted. Black proceeds to free the liberty at L9... so far, no surprises. White thows a stone again at L9, I assume to force B into filling up... And Black ignores it. For a score+ of moves.

Not only does Black ignore it but White does, too. Now, I might be completely off, but as I see it, Black considered the group important enough to save some moves ago OR those two stones pushing the liberty important enough to eat... And then decided that something else... enough something elses to fill 26 moves... was more important than 8 points. And so did White.

Now, I don't disagree with most of those options. The AI does, sometimes, and it's always in the radar, but still... I understand that some of the moves are big, there's a lot of empty space, still. But what I mainly want to know is...

Am I right assuming that

a. There are six points in play there (I'm substracting a space to settle L9 at L10)
b. There's a double gote (a move to kill and the one above to settle L9)
c. Both players consider the rest of moves are bigger than 6 + 2 x gote (for White), 2 + gote (for Black).

...or did I miss something else?

Thanks. Take care.

I think the point of is not to save L10, but to make an eye with sente.

Sorry... MV 83? I don't question that he might be trying that, but... how does that relate to the priority of eating, or not, those four black stones in an L 7-31 moves later?

Take care

removes an eye, so that White can keep pressure on the central black group while threatening to capture 4 stones.

Between and move , Black doesn't save the central stones in gote because Black can make attacking moves on the bottom group with sente.

kills the corner and strengthens E2, this is much bigger than capturing 4 stones.

is a gote move, but protecting the corner is bigger than saving 4 stones.

Between moves 101 and 117, Black has sente moves against the bottom group and the white invasion on the top.

Finally at move 118, Black no longer has urgent moves so can afford to save 4 stones.

Capturing/saving the 4 stones is worth 8 points of territory (white gets 4 captures and 3 territory if white takes, black gets 1 capture of L9; 4+3+1=8) and locally speaking is a double gote. That's pretty small for a move in the middle game. There is some added value from saving them though in that it creates a cut at k12, and you can see when black saved them white helped the 3 stones at k11 with the h10 atari, which is an active defence that also aims at taking away black's eyespace in the h7 area, so black spent a move to defend there. As jlt says black managed to make an eye in sente, and 79, 83 etc are about trying to defend the m7 cutting point which white peeped at with 72, though I can't help but wonder if black should have not tried to save the n7 stones as although black did save them and make an eye, white got a big wall at L12 and the o10 group was safe enough for now, so threat of the attack of the p6 peep then cutting at o6 was nimbly dodged in an efficient manner by white who avoided connecting on dame in gote at o6 and played more valuable moves instead.

P.S. And also the attack 101-119 allows Black to build a wall K16, which makes the cut K12 effective.

Thanks, guys

Ferran wrote:...Now, the video implies that W is a difficult position from the very beginning. I don't doubt that, but except for a tiny dip around MV 25, Leela considers White is ahead the whole game. I find it quirky, considering bots have a reputation for territory.

Didn't watch the video, but I'm curious if anyone else think White is behind early on? My impression at move 14 was that White was doing quite well. Black has put all of her first seven moves on the right side but there's still room for play by White. On the rest of the board White is developing quickly, and I'd be looking to see if I couldn't overconcentrate Black in the lower right.

Click Here To Show Diagram Code

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