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

very serious blind spot from AI competition 2024 #1003

Open
yehu3d opened this issue Nov 23, 2024 · 7 comments
Open

very serious blind spot from AI competition 2024 #1003

yehu3d opened this issue Nov 23, 2024 · 7 comments

Comments

@yehu3d
Copy link

yehu3d commented Nov 23, 2024

捕获
捕获
chinese rules komi 7.0 button=true.

@yehu3d
Copy link
Author

yehu3d commented Nov 23, 2024

i cant upload sgf format ,so i upload txt ,you can download it and rename it from 2.txt to 2.sfg
2.txt

@yehu3d
Copy link
Author

yehu3d commented Nov 23, 2024

i also collected winrate information from player, white got 90+winrate during this game under 20m visits,but finally they lost.

@michito744
Copy link

Results verified with Chinese rules
2024-11-24

Results verified with Japanese rules
2024-11-24 (1)

Conclusion: Could not reproduce

@yehu3d
Copy link
Author

yehu3d commented Nov 24, 2024

@michito744 Chinese rules komi 7.0 button=true,AI competition dont use chinese rules or japanese rules.

@michito744
Copy link

michito744 commented Nov 25, 2024

Oh sorry,

Very inefficient and a waste of energy, but it can converge on the right conclusion.
2024-11-25 (11)

If it cannot, config is not properly set up.

@yehu3d
Copy link
Author

yehu3d commented Nov 25, 2024

@michito744 Yes, starting from this position, if you continue to calculate, you can arrive at the correct answer. However, if you start calculating from an earlier position (for example, 10 moves ago), it might require significantly more computations to reach the correct answer. This could be considered a "blind spot."

@michito744
Copy link

@yehu3d Black is only winning by earning 0.5 points by “first pass” in a situation where he has the advantage of ko, so there is no other way to increase the percentage of “pass” explored, which are rarely explored in normal circumstances.

(And its very very inefficient and a waste of energy.)

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

2 participants