OR  Zero-K Name:    Password:   

Rating variance - does it converge?

3 posts, 136 views
Post comment
Filter:    Player:  
2 months ago
Is the rating variance (at https://zero-k.info/Charts/Ratings?RatingCategory=1&UserId=5295 the number that has +/- for the current rating) guaranteed to converge with increasing number of games played in a period? (and if yes, by how much)

I am playing rather rarely (once in 2 weeks for example) and I now have the impression that the variance remains high. This might be due to the fact that as I play in large team games (don't judge :-p) and the system is "undecided" forever if I actually play good or bad.

While I understand the objective of penalizing players that play rarely, and players that play randomly (which might look like for my case), I still think it would be nice/fair/desirable to have some ladder rating. If now I look at my "Ladder rating" it would be 2163 which would put me quite high. Of course, if there are 50 others players in my situation, then maybe I am much lower.

Would welcome any fix for this (for example an edit/dropdown to choose the variance at: https://zero-k.info/Ladders/Full)
+0 / -0

2 months ago
The rating variance will not converge past a certain point, this is because a single number can't store all information about your performance.

I'm working on a change to make the ladders more stable by averaging recent ratings. I think I'll replace the lower-bound estimate by the maximum likelihood estimate, so the variance has no more influence on your ladder rating.
+2 / -0
2 months ago
Great to hear you are working on a solution, thanks! Yes, indeed, I understand it can't converge to 0, but if it does not "converge" past the point of me having a rank (think you mentioned in another post 180), then I am a bit left in the dark.

Will wait for the next update then...
+1 / -0