| 1 |
Rendered notebook here (for ease of including plots and formulas):
|
1 |
Rendered notebook here (for ease of including plots and formulas):
|
| 2 |
\n
|
2 |
\n
|
| 3 |
*
https://jklw.
github.
io/WHR-ZeroK/posting1.
html
|
3 |
*
https://jklw.
github.
io/WHR-ZeroK/posting1_v2.
html
|
| 4 |
\n
|
4 |
\n
|
| 5 |
Summary: Uneven team games should probably be balanced such that the smaller team has about 50-200 higher average Elo than the larger team (depending on the specifics). Also a little background on how WHR (which I had to reimplement here) works mathematically.
|
5 |
Summary: Uneven team games should probably be balanced such that the smaller team has about 50-200 higher average Elo than the larger team (depending on the specifics). Also a little background on how WHR (which I had to reimplement here) works mathematically.
|
| 6 |
\n
|
6 |
\n
|
| 7 |
The code is [url=https://github.com/jklw/WHR-ZeroK]here[/url], though parts of the notebooks other than posting1.ipynb have already bitrotten, and I didn't include the input data (not sure if there are any privacy issues), nor the sampling traces (which are needed to run posting1) because they are about 16 GB :) You can recreate these by running the initial (still working) parts of whr_cumsum_matrix_model.ipynb and then whr_fixed_rating_model.ipynb though (after downloading malric's data).
|
7 |
The code is [url=https://github.com/jklw/WHR-ZeroK]here[/url], though parts of the notebooks other than posting1.ipynb have already bitrotten, and I didn't include the input data (not sure if there are any privacy issues), nor the sampling traces (which are needed to run posting1) because they are about 16 GB :) You can recreate these by running the initial (still working) parts of whr_cumsum_matrix_model.ipynb and then whr_fixed_rating_model.ipynb though (after downloading malric's data).
|
| 8 |
\n
|
8 |
\n
|
| 9 |
I used @malric's data from [url=http://zerok-local-analysis.s3-website.eu-west-3.amazonaws.com/release2023-06-11-02/]Zero-K Local Analysis[/url] (thanks!); if there's interest by the admins, I could re-run this stuff on a DB dump (of game dates, teams and outcomes) as a check.
|
9 |
I used @malric's data from [url=http://zerok-local-analysis.s3-website.eu-west-3.amazonaws.com/release2023-06-11-02/]Zero-K Local Analysis[/url] (thanks!); if there's interest by the admins, I could re-run this stuff on a DB dump (of game dates, teams and outcomes) as a check.
|
Balancing uneven team games - An attempt at quantifying the advantage of the larger team