David J Wu

Looks like in #639 - you trained much longer and still it doesn't predict much better. What data are you training it from? You can also try asking for help...

Strings are marked as pass-dead in a region if the region is fully surrounded by pass-alive groups of the opponent and there is

In your first case, you are correct. White's group is not scored as dead because black's stone are not pass alive. And black's two stones at the bottom are not...

No, none of these settings are likely to help. * conservativePass only affects area scoring. It makes it so that if the opponent passes, KataGo won't try pass to win...

@hope366 Can you send me the SGF for testing so I can improve this in the future?

Thanks! > By the way, "fillDameBeforePass" doesn't seem to be found in the config file, where can I find this? There are parameters that are not present in the config...

@michito744 Thanks, interesting position. It doesn't seem related all to the problem described in this thread, but still interesting! The analysis is correct if you change the komi to -10,...

@michito744 - As you can see from hope366's screenshot, the winrate AND the score are both correct when komi is -10. It is only when komi = 0 and the...

Thanks for the report, will see if I can fix for next release coming soon along with the merge of the TensorRT stuff and the fix for non-ascii paths.

A fix is should be merged into master branch now, will be part of upcoming release.