1 |
Very good! I really like the change to standard deviation, how ladder rating is calculated, map votes, that uneven custom games are excluded and afkers specced.
|
1 |
Very good! I really like the change to standard deviation, how ladder rating is calculated, map votes, that uneven custom games are excluded and afkers specced.
|
3 |
The following two things make me curious, though:[q]Ratings now have higher variance, meaning they will allow for bigger changes between days.[/q]How much higher? What is the new [url=https://zero-k.info/Forum/Thread/25943]rating variability[/url] w²? Previously, it was already 35/day where only 14/day is used in the [url=https://www.remi-coulom.fr/WHR/WHR.pdf]WHR paper[/url]. If it is this ω² shown at the rating charts, I'm wondering in which units it is. Looking at current ZK ratings, it seems like variance has increased by a factor of about 2000. I see not a sinlge player, not even @Firepluk, with a standard deviation of less than 100. This means that all variances are now bigger than 10000.
|
3 |
The following two things make me curious, though:[q]Ratings now have higher variance, meaning they will allow for bigger changes between days.[/q]How much higher? What is the new [url=https://zero-k.info/Forum/Thread/25943]rating variability[/url] w²? Previously, it was already 35/day where only 14/day is used in the [url=https://www.remi-coulom.fr/WHR/WHR.pdf]WHR paper[/url]. If it is this ω² shown at the rating charts, I'm wondering in which units it is. Looking at current ZK ratings, it seems like variance has increased by a factor of about 2000. I see not a sinlge player, not even @Firepluk, with a standard deviation of less than 100. This means that all variances are now bigger than 10000.
|