Comments (2)
Thanks! Yes, everything you're reporting is expected, there isn't anything wrong.
-
The 9x9 finetuned neural net is for 9x9 only. That's why it was advertised as a 9x9-finetuned neural net. It no longer has much experience with understanding how different board sizes behave. So it is completely normal if it gives worse evaluations on boards that are not 9x9. You should use the normal net instead.
-
None of the neural nets are trained on 6x6. So their score estimates are not guaranteed to be reliable on 6x6. However the normal net does have experience with all sizes from 7x7 to 19x19, so it can probably have some extrapolation to 6x6 even if it's not perfect.
-
It's well-known that the score estimates produced by MCTS are noisy, because maximizing score is only a secondary objective of the search. This is especially true when the score estimate is very far from 0. In this case, attempting to use the score estimate from MCTS is usually the wrong approach. Try instead adjusting komi and seeing which values of komi result in a game that is a draw, or a win, or a loss.
from katago.
Thank you for your detailed explanation.
I adjusted the value of komi to a value that would result in a tie.
It was komi=32. Originally it was supposed to be a draw with komi=35, but that didn't happen.
Next, I made a similar diagram using 7x7. The desired komi value for a draw is 48, but it was actually 47.
from katago.
Related Issues (20)
- Train a weight‘s weight, to predict the value of the next Katago weight HOT 3
- How to use Katago 1.14.1 with time cache? HOT 1
- Issue in training: low visit counts and strange initial conditions(board dimensions, komi settings, piece position) HOT 2
- kata-genmove_analyze with KEYVALUEPAIR not working HOT 1
- A problem building for CUDA HOT 1
- Why does lc0 ship cuda dlls with engine but katago doesn't?
- Minor document questions
- Error checks for kata-raw-human-nn HOT 1
- How to set rules to "twisted cross and eating" for beginner HOT 2
- Throwback HOT 3
- Lack of ability HOT 3
- Katago cannot give a definitive answer for the best move. HOT 2
- Report an error HOT 3
- allowResignation affects humanSL strength?? HOT 4
- Loose ladder
- katago 1.15.1 build failure
- Do you have a plan to create a GUI? HOT 3
- kata-analyze reports a non-human-like move for the preaz_20k human profile HOT 1
- Evaluate the value of moves HOT 2
- KataGo works with old libssl versions only HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from katago.