[gnugo-devel] Territorial value exaggerated

[SP LEE]

I have let gnugo to replay as white a game I played before and posted on 
http://mail.gnu.org/archive/html/gnugo-devel/2003-03/msg00113.html.
This is to see how gnugo thinks different than me.

Move 1 (white): GNU Go plays C6 (75.00) - Game move F17 (74.00)
Move 3 (white): GNU Go plays F15 (39.59) - Game move O17 (24.13)
Move 5 (white): GNU Go plays R14 (45.66) - Game move O16 (41.94)
Move 7 (white): GNU Go plays D18 (35.00) - Game move R12
Move 9 (white): GNU Go plays D18 (35.00) - Game move P12 (30.74)
Move 11 (white): GNU Go plays D18 (35.00) - Game move O3 (24.13)
Move 13 (white): GNU Go plays D18 (35.00) - Game move C6 (24.13)
Move 15 (white): GNU Go plays E6 (35.41) - Game move R3 (10.57)
Move 17 (white): GNU Go plays E6 (35.39) - Game move Q3 (35.00)
Move 19 (white): GNU Go plays D18 (35.00) - Game move P4 (21.01)
Move 21 (white): GNU Go plays B6 (39.01) - Game move R7
Move 23 (white): GNU Go plays B6 (39.05) - Game move R8
Move 25 (white): GNU Go plays B6 (39.09) - Game move Q8
Move 27 (white): GNU Go plays B6 (39.10) - Game move R6
Move 29 (white): GNU Go plays B6 (39.28) - Game move P9
Move 31 (white): GNU Go plays B6 (39.20) - Game move O8
Move 33 (white): GNU Go plays B6 (38.45) - Game move S4 (23.91)
Move 35 (white): GNU Go plays D18 (35.00) - Game move R5 (2.61)
Move 37 (white): GNU Go plays D18 (35.00) - Game move P10 (13.98)
Move 39 (white): GNU Go plays O10 (51.95) - Game move Q11 (45.64)
Move 41 (white): GNU Go plays D18 (21.07) - Game move H16 (14.91)
Move 43 (white): GNU Go plays D18 (20.45) - Game move J15 (1.60)
Move 45 (white): GNU Go plays G16 (27.77) - Game move J16 (20.78)
Move 47 (white): GNU Go plays N18 (26.72) - Game move M15
Move 49 (white): GNU Go plays J13 (22.64) - Game move M16
Move 51 (white): GNU Go plays N17 (48.38) - Game move H18 (3.42)
Move 53 (white): GNU Go plays K15 (37.04) - Game move N17 (25.13)
Move 55 (white): GNU Go plays M18 (48.96) - Game move P16 (5.17)
Move 57 (white): GNU Go plays M18 (50.95) - Game move O18 (1.26)
Move 59 (white): GNU Go plays L14 (60.01) - Game move N14 (13.29)
Move 61 (white): GNU Go plays O13 (27.51) - Game move N19 (11.92)
Move 63 (white): GNU Go plays M18 (49.58) - Game move M18 (49.58)
Move 65 (white): GNU Go plays L19 (50.47) - Game move L17 (48.56)
Move 67 (white): GNU Go plays Q5 (12.15) - Game move E7
Move 69 (white): GNU Go plays Q5 (12.15) - Game move F6
Move 71 (white): GNU Go plays Q5 (12.15) - Game move E8
Move 73 (white): GNU Go plays Q5 (12.15) - Game move D8
Move 75 (white): GNU Go plays G7 (44.93) - Game move P11 (37.69)
Move 77 (white): GNU Go plays G7 (45.43) - Game move C10
Move 79 (white): GNU Go plays G7 (44.81) - Game move C9
Move 81 (white): GNU Go plays G7 (48.71) - Game move C7
Move 83 (white): GNU Go plays G7 (51.17) - Game move B8
Move 85 (white): GNU Go plays G7 (50.13) - Game move F9
Move 87 (white): GNU Go plays J14 (17.28) - Game move H9
Move 89 (white): GNU Go plays J14 (17.19) - Game move G7
Move 91 (white): GNU Go plays B11 (68.37) - Game move F5 (13.48)
Move 93 (white): GNU Go plays J14 (16.44) - Game move H8
Move 95 (white): GNU Go plays J14 (16.46) - Game move J9
Move 97 (white): GNU Go plays H11 (79.06) - Game move J7

Global score: 1222.71 / 1841.28

The first impression is that many moves are not considered by gnugo. Then I 
think gnugo put high value on joseki's and may forget locally urgent moves. 
From move 21 to 33 it insists to play on B6, which claims a territorial value 
of 29.27. This really makes me puzzled. Even after B6, there are only 2 white 
stones locally, how could they make 29.27 points? I have also seen similar 
cases in other games played by gnugo, where territorial values were exaggerated 
at beginning stage of the game. Can anyone give me a clue about this?

SP Lee