[gnugo-devel] Semeai:60 traced

[SP lee]

I have traced semeai:60 after applied Paul's patch for filling bulky eyes http://mail.gnu.org/archive/html/gnugo-devel/2004-01/msg00065.html and my patch for ignoring certain lunches http://mail.gnu.org/archive/html/gnugo-devel/2004-01/msg00111.html.

 

I could overlook something, but I come to some fundamental parts again. Here is what happened. First 3 moves white and black tried are correct: W:H5 B:B8 W:H4

 

   A B C D E F G H J
 9 . X X X X X X X X 9
 8 X X O . . . . . . 8
 7 X O O O O O O O O 7
 6 . X . O X X X X X 6
 5 . . X X O X . O . 5
 4 . . . . O X X O . 4
 3 . . X X O . X X X 3
 2 . . X O . O O O O 2     WHITE (O) has captured 0 stones
 1 . . . . O . . . . 1     BLACK (X) has captured 0 stones
   A B C D E F G H J
Then black tried F3 (!) which was claimed as winning move by tactical reading, yet of course made black losing semeai.

 

In detail, there is something wrong when do_owl_analyze_semeai was trying to find tactical defense of E6:

 
 if (attack(semeai_worms[sworm], NULL)
     && find_defense(semeai_worms[sworm], &upos)) {
   critical_semeai_worms[sworm] = 1;
   owl_add_move(moves, upos, 85, "defend semeai worm", 1, 0, NO_MOVE,
         MAX_SEMEAI_MOVES);
   TRACE("Added %1m %d (0)\n", upos, 85);
 }
The tried moves of find_defense were: B:F3, W:G5, B:D4 (breaking chain), W:E2 (in attack2).

Then black tried both J5 and J4 and losed. Yet white has to try attack2 with black tenuki. When white tried J5, black captured white with J4 in fast_defense in defend1 and claimed victory.

 

Maybe F3 is just a move to be tried, but the semeai module still think it's the best move for black at that moment and didn't try anything else.

 

So I come to the fundamental part, i.e. fast_defense. Is it good to draw conclusion there? I think I have to wait the above to be confirmed before trying to make any solution.

 

 

SP Lee

semeai60trace.txt
Description: Text document