Re: [gnugo-devel] endgame module for GNU Go

[Gunnar Farnebäck]

I wrote:
> I would suggest starting with some very basic examples (of go
> endgame positions) to get some familiarity from both sides with how
> traditional go reasoning would apply and how planning theory might
> be used.

Below is an example which I would say is simple, but complex enough to
be interesting. It also features a few very typical local endgame
positions. Territory scoring is assumed, no prisoners, no komi. An sgf
file is appended at the end of the message.

   A B C D E F G H J
 9 . . . . . . . X . 9
 8 . . . O X . X . X 8
 7 O O + O X X X X X 7
 6 X O O O X . X . . 6
 5 X X X O O . X X X 5
 4 . X X X O O O O O 4
 3 . X + X O O + O . 3
 2 X X . X O . O O . 2
 1 . X . . . . . O . 1
   A B C D E F G H J

All stones currently on the board are unconditionally alive
(invincible) so there is no potential for life and death
complications. There are three independent local endgame positions:

1. Around D9/E9.
2. Around D1/E1.
3. At F5.

"Traditional" analysis says:
1. Black can play at D9 for three points in sente.
   White can plat at E9 for three points in reverse sente.
2. Black can play at E1 for two points in gote.
   White can play at D1 for two points in gote.
3. Black can play at F5 for one point in gote.
   White can play at F5 for one point in gote.

Regardless who plays first, 1 is more important than 2, which is more
important than 3.

If black plays first, ideal play is

   A B C D E F G H J
 9 . . 2 1 3 . . X . 9
 8 . . 4 O X . X . X 8
 7 O O + O X X X X X 7
 6 X O O O X . X . . 6
 5 X X X O O 8 X X X 5
 4 . X X X O O O O O 4
 3 . X + X O O + O . 3
 2 X X . X O . O O . 2
 1 . X . 7 5 6 . O . 1
   A B C D E F G H J

resulting in

   A B C D E F G H J
 9 . . O X X . . X . 9
 8 . . O O X . X . X 8
 7 O O + O X X X X X 7
 6 X O O O X . X . . 6
 5 X X X O O O X X X 5
 4 . X X X O O O O O 4
 3 . X + X O O + O . 3
 2 X X . X O . O O . 2
 1 . X . X X O . O . 1
   A B C D E F G H J

Black gets 13 points of territory, white gets 11 points, B+2.

If white plays first, ideal play is

   A B C D E F G H J
 9 . . . 3 1 2 . X . 9
 8 . . . O X . X . X 8
 7 O O + O X X X X X 7
 6 X O O O X . X . . 6
 5 X X X O O 7 X X X 5
 4 . X X X O O O O O 4
 3 . X + X O O + O . 3
 2 X X . X O . O O . 2
 1 . X . 6 4 5 . O . 1
   A B C D E F G H J

resulting in

   A B C D E F G H J
 9 . . . O O X . X . 9
 8 . . . O X . X . X 8
 7 O O + O X X X X X 7
 6 X O O O X . X . . 6
 5 X X X O O O X X X 5
 4 . X X X O O O O O 4
 3 . X + X O O + O . 3
 2 X X . X O . O O . 2
 1 . X . X X O . O . 1
   A B C D E F G H J

Black gets 12 points of territory, white gets 13 points, W+1.


Now, what might a planner approach help with in this example and what
kind of computations (presumably exploring some parts of the variation
trees for the local games) would be involved? Assume that the problem
with splitting into local games has already been solved.

/Gunnar

(;GM[1]FF[4]
SZ[9]HA[0]KM[0]
AW[db][dc][ac][bc][bd][cd][dd][de][ee][ef][ff][gf][hf][if][eg][fg][hg][eh][gh][hh][hi]
AB[ad][ha][eb][gb][ib][ec][fc][gc][hc][ic][ed][gd][ae][be][ce][ge][he][ie][bf][cf][df][bg][dg][ah][bh][dh][bi]
)