[gnugo-devel] trevor_1_16.3 & trevor_1_16.4

[Trevor Morris]

http://www.public32.com/games/go/trevor_1_16.3
 - split USAGE_DEBUG in half due to VC limitation
 - removed final VC warning in dfa.c

http://www.public32.com/games/go/trevor_1_16.4
 - tuning


The tuning patch breaks and fixes several the regressions, but in
whole I think is a net positive.

Here's my analysis of the FAILEDs it creates.  Notice that the comment
lines follow the problem.


The dragon status changed for these 3 dragons; I don't understand
why this happened.  The result is that the mystic strategic effect
changed.
strategy2:66 FAILED 
  #J11 dragon status changed from weak to alive...
strategy3:119 FAILED 
  #C6 dragon status changed from weakly alive to alive...
trevor:15 FAILED 
  # dragon A7 status changed from weakly alive to alive.


owl-node limit effect.  More-or-less random fluctuations in
owl-limited reading.
strategy:49 FAILED 
  # owl node limit effect on:
  #  gq -l games/strategy12.sgf -L 21 --decide-dragon C16 -o x.sgf
neurogo:11 FAILED 
  #loadsgf games/FSF-neurogo.sgf 148
  #owl_does_defend N5 M4
  #This result has now changed, and N5 doesn't really owl-defend M4
  #However, that's not really the whole point to this problem.


owl-threat changed.
nicklas1:2301 FAILED
  #owl tuning broke this, but seems a tad random.


Owl breakage, mostly due to improved attack or defense w/out
corresponding improved defense or attack.  All very simple to
fix.
niki:8 FAILED   
  #tough life & death problem.
niki:9 FAILED   
  #tough life & death problem.
trevor:401 FAILED 
  # improved owl attack, now needs new defense pattern
vie:19 FAILED 
  # simple failure - prefer to get patch in before fixing.


Bad problem:
owl:245 FAILED 
  #K7 is also a valid defense.



PASSED:

nicklas1:1213 PASSED 
nicklas1:1901 PASSED 
niki:14 PASSED 
strategy2:84 PASSED 
strategy3:105 PASSED 
strategy3:110 PASSED 
strategy3:140 PASSED 
strategy4:197 PASSED 
trevor:163 PASSED 
trevor:200 PASSED 
trevor:400 PASSED 
trevora:540 PASSED 
trevora:550 PASSED 

(I didn't analyse these as carefully ;-)