Re: [gnugo-devel] look at regress file

[kevin yong]

Hi, Evan:

  thank you again for your helpful reply.

  i did some detailed research on this case and the
following are what i found so far:

  1. how did gnugo found J11:
  J11 was found by gnugo in shapes(). it present J11
with 3 move_reasons:
    Move at J11 expands territory
    Move at J11 strategically attacks L7
    Move at J11 strategically defends J10

    J11: 9.97 - change in territory
    J11: 13.52 - strategic effect on J10
    J11:   7.07 - total followup value, added 7.07 
      as territorial followup
    J11: 3.53 - added due to followup (7.07) and 
      reverse followup values (0.00)
    J11: 1.35 - shape (shape values +1.00(1) -0.00(0),

      shape factor 1.050)
    J11: 0.41 - connects strings (connect value 4, 
      shape factor 0.082)

  2. based on the info about J11, i did some simple
experiements with L9: without going through a real
move-generate analysis, i arbitary add L9 into move[]
list with the same 3 reasons as J11 (i did it by
modify the source code of genmoves()), then run test
suite to see how gnugo reacts:
    Move at L9 expands territory
    Move at L9 strategically attacks L7
    Move at L9 strategically defends J10
    L9: 1.00 - change in territory
    L9: 13.52 - strategic effect on J10
    L9:   11.10 - total followup value, added 11.10 
      as territorial followup
    L9: 7.05 - added due to followup (11.10) and 
      reverse followup values (3.00)
    L9: 0.63 - connects strings (connect value 6, 
      shape factor 0.126)
    Top moves:
    1. J11 28.80
    2. L9  22.20
    3. M4  14.54
    ...

  it's interesting to see if somehow L9 was found as
move candidate, gnugo value_moves would give it a
weight quite close to J11. if we look at the detail of
how the points are assigned to J11 & L9:
  a: J11 got more points on territory gain than L9
(9:1). by look at the board situation manually, i
think this valuation is quite good.
  b: L9 got more points followup-value than J11 (7/4).
by look at the board situation manually, i think it is
in the right direction, but quantitatively not good
enough. as you said: W move at L9 is sente. that's the
focus of the entire story: gnugo should assign higher
followup-value to the move with sente.

  3. i found that there is no difference between if L9
is presented with 2 move_reasons:
    Move at L9 expands territory
    Move at L9 strategically defends J10
or with 3 move_reasons by add:
    Move at L9 strategically attacks L7
in terms of final valuation results.
Then, i dig into the source code of value_moves.c, and
found that there is no place the 'strategically
attacks' move_reason was valuated and add a value.
>From the manual judgement of the board situation, the
sente of L9 is because it's strategically attacks L7
if it get followup-move at M9. So, i think this is the
key problem of gnugo in handling this specific case.

the solution: i recommand to make some enhancement to
the value_moves.c to add handling 'strategically
attacks' move_reason, especially with sente.

how difficult it is and how well it will be done, that
depends, but in principle, that's something needs to
be done. i like to give a try and also like to co-work
with anyone who likes to join.

i really like to hear more comments from you and
anyone interested, especially the guy who is in charge
of value_moves.c

best regards.

kevin.



______________________________________________________________________ 
Post your free ad now! http://personals.yahoo.ca