[question] Typical value of resign threshold during training
y-ich opened this issue · comments
ICHIKAWA, Yuji commented
Hi.
In your paper, there is a description "As in Silver et al. (2017), we dynamically calibrate our resignation threshold to have a 5% false positive rate".
Could you tell me the typical value of the resignation threshold?
Thank you.
ICHIKAWA, Yuji commented
I checked selfplay SGFs.
It was near 16%。