Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

winrate and komi #2384

Closed
palusijf opened this issue May 12, 2019 · 2 comments
Closed

winrate and komi #2384

palusijf opened this issue May 12, 2019 · 2 comments

Comments

@palusijf
Copy link

HI,

I don’t understand why the win rate is still the same whatever the value of the Komi.
For the first move with LZ network (#225) :

  • with Komi = 7 .5 : WR = 43.6
  • with Komi = 0 : WR = 43.6
@wpstmxhs
Copy link
Contributor

wpstmxhs commented May 12, 2019

Winrate is from output of neural network's value head.
But neural network doesn't have a input for komi.
Therefore, regardless of GTP komi settings, winrate is calculated based on komi of 7.5 points.
Because weights are trained from komi 7.5 self-play games and value head output is estimation of its game result.

Simply speaking, you can't change komi from 7.5 points.

@roy7 roy7 closed this as completed May 12, 2019
@palusijf
Copy link
Author

Ok, thanks for the answer.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants