UCT linear dynkomi: make it default, set val_scale=0 & sure_win_threshold=0.9
commitfed160da05a66a2a614fbd8b1f4fc3cf1fe98ee0
authorJean-loup Gailly <jloup@gailly.net>
Sat, 27 Aug 2011 10:24:03 +0000 (27 12:24 +0200)
committerJean-loup Gailly <jloup@gailly.net>
Sat, 27 Aug 2011 10:24:03 +0000 (27 12:24 +0200)
tree9a3304a69d5ceeb3e75eb097d5fcdf40da4bdbe7
parent22500eb54a36a2b5f72d084e4b1caefa9a1c9400
UCT linear dynkomi: make it default, set val_scale=0 & sure_win_threshold=0.9

The adaptive phase of linear dynkomi works well on KGS,
allowing wins by many points (val_scale > 0 wasn't performing
well for this). Also allow the adaptive phase to start earlier
(works better on KGS). sure_win_threshold raised to be > green_zone.
uct/dynkomi.c
uct/uct.c