New 3-3 invasion joseki by ELF

[Gomoto]

EDIT: This is not joseki, but a move for special cases and depends heavily on the whole board position

I discovered this joseki today while reviewing with ELF.

The joseki up to move 5 is very rare in my pro database and move 6 is the new joseki move. This looks rather interesting in my opinion. What do you think?

$$Bc19
$$ --------------------+
$$ . . . . . . . . . . |
$$ . . . . . 3 2 4 . . |
$$ . . . . 6 . X O . . |
$$ , . . . . . X O . . |
$$ . . . . . . 5 . . . |
$$ . . . . . . . 1 . . |
$$ . . . . . . . . . . |
$$ . . . . . . . . . . |
$$ . . . . . . . . . . |
$$ , . . . . . , . . . |

Click Here To Show Diagram Code

[go]$$Bc19
$$ --------------------+
$$ . . . . . . . . . . |
$$ . . . . . 3 2 4 . . |
$$ . . . . 6 . X O . . |
$$ , . . . . . X O . . |
$$ . . . . . . 5 . . . |
$$ . . . . . . . 1 . . |
$$ . . . . . . . . . . |
$$ . . . . . . . . . . |
$$ . . . . . . . . . . |
$$ , . . . . . , . . . |[/go]

For the joseki up to move 5 see also:

Lee Sedol 9p - Park Junghwan 9p, B+Resign (Komi 6.5)
Korean Hwaseong Go Festival special game, 2017-11-09

Ishida Yoshio 9p - Ishida Akira 9p, B+4.5 (Komi 5.5)
36th Japanese NHK Cup, round 2, 1988-08-28

[Gomoto]

This is what ELF thinks:

[Gomoto]

I like that white stops following blacks plan here and even reaches out for sente:

[Gomoto]

This continuation is even better for white:

[EdLee]

Hi Gomoto,

Lizzie has a pretty board and stones.

When the current move marker is , how to interpret the percentages ?
( are they winrates for White, since it's White's turn ? )

Same question for the opposite:
When the current move marker is , how to interpret the percentages ?

( Wish there's some way they can improve the UI so the % are obvious which color they represent... -- I still haven't used any of these bots yet. )

[Gomoto]

This is not your everyday joseki, the new move is only for special cases.

I try to investigate the relation to the whole board situation.

[Gomoto]

@EdLee

The percentages show the winning percentage for the player that is going to make the next move.

[Gomoto]

Some interesting evaluations by ELF:

eval1.jpg (163.93 KiB) Viewed 14862 times

eval2.jpg (159.68 KiB) Viewed 14861 times

eval3.jpg (159.97 KiB) Viewed 14861 times

[EdLee]

Hi Gomoto,

Thanks.

What are B's and W's winrates before ? ( 7.5 komi ? 6.5 komi ? Chinese rule ? )

Out of curiosity -- for this board:

IMG_0665.JPG (177.43 KiB) Viewed 14853 times

What was the previous board position ( which W move ) where W dropped below 50% ?

[Gomoto]

The hane (move 2) is a very random move for ELF. In the second picture beneath ELF misses the hane in this run on the same board configuration. It still thinks it is good if I force it. (It seems to be worth considering it from time to time but the reading is above my head.)

[dfan]

The percentages show the winning percentage for the player that is going to make the next move.

By the way, for anyone who doesn't like this system, I added an option in lizzie.properties ("win-rate-always-black") that always displays the win rates from Black's point of view. I prefer this because it's easier for me to see what's going on when the win rates on consecutive moves go something like 42-40-39-38-41 rather than 42-60-39-62-41.

[EdLee]

Hi Gomoto, thanks.

I meant in the absolute sense: on an empty board, what are ELF's winrates for B and W ? I seem to recall AZ thought W +50% with 7.5 komi --
I'm curious if ELF also uses 7.5 komi with Chinese rules,
and whether ELF thinks W is ahead before . Thanks.

[EdLee]

Hi Dan, are you the developer of Lizzie ?

[dfan]

I am not the main developer, but it's open source, so I have contributed a few features.

[Gomoto]

ELF empty board:

49,3% (4-4)
47,8% (3-4)