|
Post by Disaster Area on Sept 29, 2020 17:37:51 GMT
I was having a little think about how we could adapt the 2k20 rating system (which is more or less the same as Pokemon Perfect's) to generate a single rating score for players on the site, that could maybe be used in team tournaments to, for example, restrict the teams that can be formed or auto-generate a limited form of player rankings.
My first idea is something like this:
30% all time points divided by the highest all time score 30% player rankings points divded by the highest points in the player rankings 40% the highest of those two numbers
then multiply it by 1000
the idea would be to reward both impressive long term records and significant shorter term achievement.
A player with the highest all time score and not participating in any seasons counting to the latest player rankings would have a score of 700, and the #1 player ranked in the player rankings would have a score of at least 700. There's a fixed maximum score of 1000.
One good thing is that for good players it wouldn't take an extraordinary amount of climb to a good points level once the system has been in place for some years, but all time achievements aren't erased either.
|
|
|
Post by pumpyaarony on Sept 30, 2020 20:34:44 GMT
It sounds a bit complicated, but im sure we’ll make more sense of it as we go along. Let’s do it!
|
|
|
Post by Ortheore on Oct 4, 2020 3:59:27 GMT
Honestly I think it's hard to say how useful this would be without some practical examples, which obviously won't come until we start running seasons (unless we use PP data?). I think we should go with it, and try a couple different formulas and alter based on feedback.
I am curious as to how we go about handling long term players, because I believe there's serious potential for things to go askew. I guess the two scenarios I'm thinking of are: - Legacy player simply isn't as good as they used to be
- Legacy player isn't as active as they used to be
In the former scenario, the old player could potentially be overvalued by the rating system. Who cares what they did 5 years ago if they're not good now? Although lifetime achievements are a really cool thing to note and should definitely be recognised in some way, if we're using this for a team tournament then it might not be appropriate
In the latter scenario, I'm not even sure what the best approach would be. There's just too many variables imo. Have they been entirely afk? Have they been following the meta? Have they maybe been dabbling here and there, but perhaps not seriously grinding tournaments? Their ability could span a broad range, and different players would respond differently as well. Could there be more ways of deconstructing it? Idk, I'm just spouting ideas and questions atm
|
|
|
Post by Disaster Area on Oct 4, 2020 12:41:03 GMT
They're good questions to ask Orth. Let's both have a think about it...
I guess what we could do to answer them in some way is:
- making assumptions when players are inactive, about what happens to their skill over time - deciding how to evaluate new information for players with longer records
I think the current system I proposed basically assumes no loss of skill even if new results are poorer, and gives a lot of credence to players when they have a recent uptick.
Given the timescales involved, I think giving a lot of credence to players when they have a significant uptick in the points they earn is a good and fair idea. But maybe we need to evaluate player's performance over the long term differently.
|
|
|
Post by Disaster Area on Oct 5, 2020 21:25:37 GMT
I created a discord channel on this topic if anybody wants in let me know. Firstly discussing how a rating system could be stored and how we might implement GXE
|
|