Jlt's game reviews

[jlt]

Following this thread Long & live go Dojo 🏯 I played a LL game against @FinrodFelagund. The game was without komi, I was Black, and time controls were 2h + 5x20s. We played for 2 hours, then took a 2h break. I prepared myself for another 2h marathon, unfortunately I had to resign 30 minutes after at move 119. Nevertheless, several positions were quite interesting and worth reviewing.

This move is the first significant mistake by White, although I didn’t manage to punish it.

I hesitated between the kick E2 and connecting my stones with F3. I chose the kick, thinking that if White extends then I can play kosumi

and if White jumps then I can clamp.

The last diagram corresponds to the game, however White had a better response:

So Black should have played like this

move 10 being necessary otherwise the stone 6 can be captured. Black loses a few points but gets a strong group, damages the lower right white group, and gets sente.

If White 6 is an atari at E3, then Black can let White capture!

Then White has either to live with 2 points while being completely surrounded, or crawl on the second line towards the right.

Coming back to the previous position, Black has sente to take the lower right corner. White can’t really block Black because after this sequence,

Black can play A in sente, threatening B.

The following move is a mistake:

White doesn’t need to respond and should play A in the diagram below. So instead of the game move, Black should have played B.

The followup of B is this sequence:

In this position my first thought was to play A but after overthinking I played B. This was the beginning of my problems.

In this position the lower right corner is a dead L-shape, yet the position is about even. The reason is that Black can play A, threatening B or C.

If for instance White prevents the connection with B and Black plays C,

Then if White cuts with 4 then Black 5 lives since White cannot protect A and B at the same time.

If instead of cutting White plays hane, Black can live with B or capture the whole group with A in the diagram below:

The last position of interest is this one:

Since the ladder didn’t work I thought I would squeeze with N7. However Katago prefers playing out the non-working ladder:

If White A then Black B. White’s dragon is not alive so Black can make profit by forcing a capturing race, using the fact that C can extend Black’s liberties in the forner.

If White B then Black A. The white central group has an eye but one eye is not enough to live so Black can continue to make profit by attacking either white group.

[FinrodFelagund]

Thanks for the review of our game. I think you covered the key encounters of the game pretty well. This format of game might be useful to me because it seemed to emphasis a weakness I’ve noticed before. When both sides are playing well and no obvious fighting has started, I have a tendency to get impatient and nervous. Both C14 and my first mistake F3. I sorta knew from reading the position that both moves were unproductive, but I played them anyway, almost for lack of better ideas.

I think I need to focus on being patient and taking big points when I don’t see any good opportunities.

[jlt]

My counting is really bad. In this position, who is ahead, and approximately by how many points?

[Groin]

Black like 65
White like 68 +komi
?

[jlt]

Komi was 6.5. The game was this one: Tournament Game: Fast Correspondence Title Tournament 2022 (82076) R:1 (AA8 vs jlt)

AI evaluation

Black+25.5

[Groin]

Then obviously i missed something

I only counted what was sure points, and quickly like playing a game. Didn’t study the remaining points to share in late yose too.

[Groin]

Maybe white UR corner is weak

[Lys]

If I was in game I would’ve counted this way:

Which is white ahead by 5pts + komi.
But the red circled groups look very unsettled to me, so I didn’t count points for them and killing one of them would completely change the balance

[jlt]

Both of you count as badly as me. A few moves later, all groups were completely settled and Black was still leading by 19 points according to AI (i.e. 25 points minus komi).

[jlt]

Counting like this I get
Black = 63+2 prisoners = 65
White = 53+6.5 komi = 59.5
so Black seems to lead by 5.5. However Katago estimates that B+25.5 so there are 20 invisible points somewhere, without killing any group. Endgame is a mystery.

[shinuito]

I’d say it’s not just that black has missing points. White is probably being given too many on the one hand, and on the other maybe black can eek out a few more while reducing white. Where they end up might depend on the sequence chosen. (It’s not very easy to picture though if true)

I think I could imagine B+14 maybe on the board, but B+20 including komi is hard to see.

I think white has a lot of weaknesses in the middle left, between the cut or clamp of C7, B8 and some kinds of attachments F10 or C10. I don’t know the variations, and I’d have to probably study it with katago at least to speed things up.

Maybe even both J4 and K5 get captured if white doesn’t get time to save them

Edit: oh Black also has 5 extra prisoners right?

[shinuito]

Maybe an interesting idea could be to fork the game from that point and challenge one of the site katago bots and see how it plays out.

Let’s say let Black be katago so they can demonstrate where the extra points are?

[shinuito]

If you’re wondering about it, it’s a minefield trying to play that endgame as white against katago.

I just keep dying randomly at my level

Endgame practice

Endgame practice

[claire_yang]

Some of the missing territory probably should belong to black, with a total about 15 points lead (after komi)

And since black has sente and gets to play the next move, the extra points I feel are the power of sente.

Like this, would gain extra 7 points, which adds to 22 points lead, added the 2 captures, is pretty close to the AI estimated 25 points lead.

[FinrodFelagund]

I think this sorta explains why katago estimates the score so differently. White’s territory has a lot of shape defects that end up shrinking its potential with proper play. If you try to hold onto it all, white gets cut up.

[Groin]

From @shinuito using katago, i see that i overestimated white solidity. it’s amazing.
black has a promising yose thanks to his lack of weakness compared to white’s thiness

[Groin]

Makes me think of games with quite strong players good in yose. During the game i feel i am ok afterall with some moves i regret a bit or some thiness, bad aji but i feel good because he didn’t really crush me and looked even cautious… And finally when the big yose starts, you get crushed, he takes all and leaves you nothing. Aaargh so depriming.

[claire_yang]

Many players think yose is just something to calculate after mid-game, but it’s actually linked to how mid-game is settled.

There is a reason why Lee Changho at his prime was said to be invincible in his yose, he is so thick and strong everywhere, sometimes there is literally no sente yose for his opponent to play.

[jlt]

Before going to bed, just a quick review of a game that I lost against Sofiam. I was Black.

First position: resist the urge to capture the cutting stone! Extend at B instead!

Second position:

I was pretty unhappy with the beginning of the game and really felt behind but actually AI says I am only 3 points behind at that stage. I really need to work on self-confidence and not be distracted by negative thoughts.

I was afraid of White’s moyo on the right but AI suggest to play calmly. The following sequence builds the bottom while helping the weak central group.

In the following position, again I thought that the position was really bad for me but AI says that I was about 5 points behind and suggests this sequence, which shows that the right side is not so big. White does make some points, but Black reduces while enlarging his corner so that’s fair.

White can try to resist, and the capturing race is difficult to read. It seems that Black can capture White by ko but it’s hard to judge for an ordinary human player like me. In any case, Black has an insurance at A in case the corner is in danger.

[jlt]

A quick review of this game that I would probably have lost against a 4k if he had protected his lower right corner towards the end of the game.

I will focus on two positions at the beginning of the game.

The following sequence is suggested by AI.

It makes sense to save M15 since this weakens J17. I am usually afraid to let White get Q14 but AI thinks the corner is fine. If White later plays S15, then according to AI Black should play P14 and the fight may get complicated, but I should probably try that variation one day.

The second position is a joseki that I repeatedly get wrong, so I want to record it here. It starts from this attachment:

After White pushes at D16 and D16, Black pushes at C14 and the following sequence is expected

If however White plays 1 at 5, then Black can respond like this:

(Not clear if the exchange 4-5 is good or not.)

If White plays 3 at 5 then Black 4 is directly at 6.