What is 'rating decay'?
'Rating decay' system is a variant rating system, which gradually decreases player ratings naturally over time.
Seasonal rating resets may be seen as a form of rating decay too.
Plenty of games implemented rating decay, and alternative rating systems like Glicko address rating decay by itself.
But I'm not arguing BGA should shift to another rating system, so I'll not mention them more in this article.
I'm against rating decay in Arena mode due to Arena's seasonal resets. Adding decay in Arena EAS is redundant.
However normal ELO needs to introduce rating decay for numerous reasons.
Reasons to introduce rating decay
One of the fundamental problem in ELO rating is 'ELO protection' which heavily stagnates and disencourages player activities.
This problem has been around for a long time in BGA too, and still exists now.
Check out this post from Oct. 2016: https://boardgamearena.com/forum/viewto ... 120#p32655
Most of top players just sit out and try to preserve ratings and ranks by playing once or twice a month.
This happens more frequently in luck-based or multiplayer games.
'Top players' pull off lucky streaks for first 20 games and get K-factor boosts, then stop playing to preserve their ELO.
Also, ELO protection discourages other players' motivation.
Without decay ELO 1st~20th prestiges become practically impossible goals in many games.
Veteran players got huge advantage from 2017 rating reform, receiving nearly 200 rating boost compared to newer players.
(Reform announcement: viewtopic.php?t=10974)
And many of those players still protect rating and ranks for years without playing games much.
In the past prestiges were not permanent so this behavior was understandable, to keep trophies.
But now they are, so losing current rank does not mean losing trophies too.
Adding rating decay rewards active playerbase highly, which is favorable for the entire site.
In addition, rating decay helps appropriate matchmaking.
Skill of most (if not all) players returned from long hiatus is not the same as before.
However current BGA rating system can't reflect this, so players suffering 'skill rust' are assured to lose some rating.
Thus these players also stop playing if they want to protect own ratings.
Plus, few lucky streaks from multiplayer games will not guarantee a safe spot anymore if rating decay is introduced.
This phenomenon is apparent in many games, so some games add activity bonus instead of rating decay.
However there are a lot of games with small playerbase in BGA.
So, adding activity bonus in BGA will inflate ratings, especially of few diehard players.
In contrast, rating decay can keep rating inflation in check by decreasing overall rating sum.
Adding rating decay can solve these problems altogether by forcing players actually play games to preserve ratings.
Good rating system should help matchmaking and encouraging players to play, not doing the opposite.
Which algorithm to use?
I think linear decay (flat -X per day) fits in BGA's environment most.
Consecutive inactivity is hard to measure, because in BGA both real-time and turn-based games use the same rating.
Linear decay without any condition related to playtime seems to be fairest method then.
-0.5 ~ -0.7 per day might be proper. It is -15 ~ -21 per month.
For reference, winning a single 1:1 game vs a player with -400 rating gives +2.
I can't say this is the best value cause I don't have statistic samples large enough though.
Setting decay cap
If BGA introduces rating decay, there should be a decay cap to prevent over-reducing ratings.
Rating decay must not put ratings below 200, which is a borderline of 'good players'.
Further decay seems to be meaningless and overly harsh for new players trying to stay above 100.
Decay cap prevents former strong or higher-rated players being classified as average players or beginners as well.
Also, rating decay should not punish high-rated players too much.
So my proposal is, '200 or half of the maximum rating record, which is higher'.
For example, max. strong players' cap stays in 200, and masters' cap becomes at least 350.
In this fashion higher rated players can retain some kind of advantage.
Feel free to add comments or opinions!
'Rating decay' system is a variant rating system, which gradually decreases player ratings naturally over time.
Seasonal rating resets may be seen as a form of rating decay too.
Plenty of games implemented rating decay, and alternative rating systems like Glicko address rating decay by itself.
But I'm not arguing BGA should shift to another rating system, so I'll not mention them more in this article.
I'm against rating decay in Arena mode due to Arena's seasonal resets. Adding decay in Arena EAS is redundant.
However normal ELO needs to introduce rating decay for numerous reasons.
Reasons to introduce rating decay
One of the fundamental problem in ELO rating is 'ELO protection' which heavily stagnates and disencourages player activities.
This problem has been around for a long time in BGA too, and still exists now.
Check out this post from Oct. 2016: https://boardgamearena.com/forum/viewto ... 120#p32655
Most of top players just sit out and try to preserve ratings and ranks by playing once or twice a month.
This happens more frequently in luck-based or multiplayer games.
'Top players' pull off lucky streaks for first 20 games and get K-factor boosts, then stop playing to preserve their ELO.
Also, ELO protection discourages other players' motivation.
Without decay ELO 1st~20th prestiges become practically impossible goals in many games.
Veteran players got huge advantage from 2017 rating reform, receiving nearly 200 rating boost compared to newer players.
(Reform announcement: viewtopic.php?t=10974)
And many of those players still protect rating and ranks for years without playing games much.
In the past prestiges were not permanent so this behavior was understandable, to keep trophies.
But now they are, so losing current rank does not mean losing trophies too.
Adding rating decay rewards active playerbase highly, which is favorable for the entire site.
In addition, rating decay helps appropriate matchmaking.
Skill of most (if not all) players returned from long hiatus is not the same as before.
However current BGA rating system can't reflect this, so players suffering 'skill rust' are assured to lose some rating.
Thus these players also stop playing if they want to protect own ratings.
Plus, few lucky streaks from multiplayer games will not guarantee a safe spot anymore if rating decay is introduced.
This phenomenon is apparent in many games, so some games add activity bonus instead of rating decay.
However there are a lot of games with small playerbase in BGA.
So, adding activity bonus in BGA will inflate ratings, especially of few diehard players.
In contrast, rating decay can keep rating inflation in check by decreasing overall rating sum.
Adding rating decay can solve these problems altogether by forcing players actually play games to preserve ratings.
Good rating system should help matchmaking and encouraging players to play, not doing the opposite.
Which algorithm to use?
I think linear decay (flat -X per day) fits in BGA's environment most.
Consecutive inactivity is hard to measure, because in BGA both real-time and turn-based games use the same rating.
Linear decay without any condition related to playtime seems to be fairest method then.
-0.5 ~ -0.7 per day might be proper. It is -15 ~ -21 per month.
For reference, winning a single 1:1 game vs a player with -400 rating gives +2.
I can't say this is the best value cause I don't have statistic samples large enough though.
Setting decay cap
If BGA introduces rating decay, there should be a decay cap to prevent over-reducing ratings.
Rating decay must not put ratings below 200, which is a borderline of 'good players'.
Further decay seems to be meaningless and overly harsh for new players trying to stay above 100.
Decay cap prevents former strong or higher-rated players being classified as average players or beginners as well.
Also, rating decay should not punish high-rated players too much.
So my proposal is, '200 or half of the maximum rating record, which is higher'.
For example, max. strong players' cap stays in 200, and masters' cap becomes at least 350.
In this fashion higher rated players can retain some kind of advantage.
Feel free to add comments or opinions!
