Re: [gnugo-devel] how to fix a problem?

[Arend Bayer]

Paul wrote:
> while looking through Trevor's test suites today i discovered trevora:610
> problem. a wonderful tesuji! it would be very nice if gnu go knew it.

Wouldn't "How to teach gnugo wonderful tesujis?" be a more motivating
subject line? ;)

>    A B C D E F G H J
>  9 . . . . . . . . . 9
>  8 . . . . . . O O O 8
>  7 O O O O . . O X X 7
>  6 X O X X O O X . X 6
>  5 X X . X + . X X . 5
>  4 . . X O O O O X . 4
>  3 . . X . . . X . . 3
>  2 . . . . X . . . . 2     WHITE (O) has captured 0 stones
>  1 . . . . . . . . . 1     BLACK (X) has captured 1 stones
>    A B C D E F G H J
>
> white D2 or F2 invades. this position must be rather common in games.

I just recently chatted with Gunnar about similar situations. Take e.g.:
   A B C D E F G H J
 9 . . . . . . . . . 9
 8 . . . . . . . . . 8
 7 . . + O X X + . . 7
 6 . . . O O O X . . 6
 5 . . . . + . . . . 5
 4 . . . . . . X . . 4
 3 . . + . . . + . . 3
 2 . . . . . . . . . 2     WHITE (O) has captured 0 stones
 1 . . . . . . . . . 1     BLACK (X) has captured 0 stones
   A B C D E F G H J

O can gain a lot here by playing G7 or G5. And I think GG could improve
a lot by finding such moves (moves breaking into opponents purposed
territory/moyo).

Possible ways to try to find these:

1. Try disconnect()s between the strings enclosing the moyo.
I fear this would give far too many false postives, where a
disconnection is only succesful in a readconnect sense, but the cutting
stones will get captured eventually.

2. Both examples above suggest looking aggressively for non-transitity
situations. Again, many false positives seem likely, but it's certainly
s.th. worth considering. (And s.th. we have to do at some point anyway.)

3. I have the crude idea that one could misuse the connection code
as follows:

   A B C D E F G H J
 9 . . . . . a a a a 9
 8 . . . . . a a a a 8
 7 . . + B X X a a a 7
 6 . . . B B B X a a 6
 5 . . . . + . . a a 5
 4 . . . . . . X . . 4
 3 . . + . . . + . . 3
 2 . . . . . . . . . 2     WHITE (O) has captured 0 stones
 1 . . . . . . . . . 1     BLACK (X) has captured 0 stones
   A B C D E F G H J

connect(a, B) is deemed successful if white can get any stone on a
connected to B.


If someone feels like trying out anything of the above, or, even better,
has better suggestions/ideas...


Succesfully implementing s.th. like that could be worth as much as
a full stone in strenght, I'd think.
It's also s.th. that seems to be implemented in stronger computer go
programs (Go4++, Haruka).


Arend