Re: [gnugo-devel] A terrible game

[SP LEE]

> 
> Currently I think the GNU Go engine is stronger than it's
> ever been. Yet here is a game in which 3.3.23 allowed two
> groups to die, both very unnecessarily.
> 
> This is a level 9 game. (The only difference between levels 9
> and 10 is that at level 9 the breakin code is turned off.)
> 
> Dan
> 


Again at move 77, white is checking the sequence after playing T3, I got the 
following debug output:
(loadsgf bad.sgf 77)
(owl_defend N8)
 ...
    Pattern D720 found at T5 with value 55
      Trying white T5. Escape = 0. Current stack: W:T3 B:Q5 W:T5 
      eye space at O8 of type 5252
      eye space at S6 of type 310
      eye space at T4 of type 202
      vital move at S5, score 50 (eye at T4, value 0011, pessimistic_min 0)
      Pattern A718 found at O5 with value 50
        Trying black O5. Escape = 0. Current stack: W:T3 B:Q5 W:T5 B:O5 

So although gnugo found S5, it checked O5 instead and I'm confused.

After white played T3, the gnugo checked the sequence of Q5 and I got the 
following debug output:
(loadsgf bad.sgf 78)
(owl_attack N8)
 ...
  Pattern D720 found at T5 with value 55
    Trying white T5. Escape = 0. Current stack: B:Q5 W:T5 
    eye space at O8 of type 5252
    eye space at S6 of type 310
    eye space at T4 of type 202
    vital move at S5, score 50 (eye at T4, value 0011, pessimistic_min 0)
      Trying black S5. Escape = 0. Current stack: B:Q5 W:T5 B:S5 
      eye space at O8 of type 5252

So can anybody explain the difference of these two output, i.e. the vital point 
S5 was not considered in the first case? I suppose they sould be the same. The 
only difference is that in the second case T3 was actually played. Is this a 
cache problem?

SP Lee