OR  Zero-K Name:    Password:   

Can we please address ladder decay?

5 posts, 380 views
Post comment
Filter:    Player:  

12 months ago
I find myself disincentivised to play.

Ladder decay has introduced additional noise to ladder. Normally it's an already loose approximation of skill. Now the effect to noise ratio is so large as to be meaningless. Even if I play every day (as if I could find that many games), and remove all the noise from my own score, I still could not make inferences about my own skill relative to others due to not knowing whether their score has decayed.

There are some arguments for having decay, but 400 is far too large. Rust will never make a 400 elo difference. Maybe 100 tops, and that would be down to near nothing within 10-20 games.

DeinFreund was working on an alternative system. But that seems to not have been implemented yet?

I know elo isn't that big of a deal, but personally I like seeing whether I'm improving or not.
+4 / -0

12 months ago
(edited 12 months ago)

CHrankAdminDeinFreund's WHR system has been coded but has not been deployed (yet). You can voice your concerns in the ticket: https://github.com/ZeroK-RTS/Zero-K-Infrastructure/pull/1663

I agree it should not reach 400 since that is the uncertainty of a completely new player. I don't know what the cap should be though (if we stick to regular Elo and not WHR). 100 sounds like an okay ballpark figure.

+2 / -0

12 months ago
A while ago I saw my 1v1 elo had dropped from about 1800 to about 1450. Over one weekend I played about 20 games, and only won something like 4 games, and my elo grew back into 1800s anyway.

I agree something should be done to incentivize playing 1v1 ranked games. Maybe the rust "penalty" should be shown separately.

Maybe "rust" isn't a proper word for it. A possibility would be a requirement of a few games per month to be shown on the ranked ladder and showing the last game date and an "INACTIVE" tag near the 1v1 elo for players who don't meet that criteria.

+0 / -0

12 months ago
WHR also includes a kind of Elo decay. This was required as to not be lacking any of the features existing in the current ratings. It's slightly different because it has a sqrt(x) decay function. This means it'll take a very long time for maximum decay to occur. If you didn't actually change much in rating it'll also return your old rating much quicker.

Personally I don't think decay makes much sense unless you want to replace ratings by lower bound estimations. What I'd rather do is, if this encourages more games, is removing people from the ladders quicker.
+0 / -0

12 months ago
In Go i lost around 2 ranks (roughly 200) elo after not playing at all for 5 years. So i guess this would be a reasonable amount?
+0 / -0