Introducing Rating Decay

[ufm]

Something like

daily_reduction(x) = ln(x-250)/10

so that the rate of decay slows down as you get closer to 250, but is about 0.5 for most ratings.

I can't collect enough game stats to determine the best formula so really not sure about which is better.

ELO is a mark of skill, not how often you play. You could play once a year and beat the #1 guy everytime - your ELO should not be 200, or whatever other artificial lower limit you place.

You are making the mistake of blaming ELO, when in fact it's the leader-boards that you have a problem with.
You do drop off the leaderboards after 6 weeks, so it sounds like you should propose to drop of the leader boards in a shorter time.

Arena is (apparently) the solution to the problem you're trying to 'fix'.

It's nice to see the ELO on a players profile even if they're not active in that game - just like yours - very impressive.

People that have that ELO rating have usually earned it. Stop trying to take it away from them.

The case is quite extreme. Even pros keep on training (in this case, playing games) during offseason to prevent rust.
It is highly probable if somebody plays frequently outside BGA and visits here once a year, but how many such games are here?
In addition, if that imaginary player plays 1/year the rating cannot reflect that player's skill correctly anyway. Achieving high ratings takes time.
What I can think of is a player went up to expert or master in BGA, left BGA for some reason,
keeps playing in somewhere else while returning to BGA 1/year. This would be exceedingly rare.

I don't know why you're blaming leaderboards. Rating decay is not limited to few top players, but it is for every player.
Plus, player inactivity and rust after hiatus have been and still are both well-known problem.
These are the issues rating decay tries to solve, and ELO does not deal with them inherently.
Some huge game servers already implemented Glicko-2 instead of ELO and it works better.
Though I'm not for switching the entire BGA rating system as I mentioned, pretending the problem does not exist never helps anyone.

A considerable number of players do not play in Arena mode for various reasons.
Existence of Arena mode is not really a convincing excuse of leaving Normal mode as it is.

[RicardoRix]

First and foremost:
ELO is a mark of skill, not how often you play.

The one-a-year scenario I used is simply to show that rating decay would be wrong.

A rating decay system is used in Tennis, where it is a single constant pursuit of professional players - this now seems to fit the bill better. (note this is the complete opposite of BGA). Even so, you get examples of players like Serena Williams that takes time off due to injury and/or pregnancy and sharply falls down the ratings. Comes back whoops everyone MUCH higher rated than them, and everyone starts to blame the ratings because it's so badly wrong.

If ELO is just a number to you, and you don't have a problem with the leader-boards, then you simply don't have a problem. What do you care what someone's number is?

There's been quite a few previous threads on this kind of subject. The main theme is, come up with something better than ELO....... This isn't it.

And as I've previously stated, and if you do care about this kind of thing, then play Arena mode. You don't need 2 different systems that will then be even closer linked. What's the point of having both now?

IMO: There is nothing wrong with ELO, is has strong mathematical foundations and works very well in almost all cases, which is extremely hard to try and achieve. Artificial balances with these calculations is normally wrong, the best you can have is the K-factor, which is used as some kind of 'time' adjuster.

[ufm]

First and foremost:
ELO is a mark of skill, not how often you play.

The one-a-year scenario I used is simply to show that rating decay would be wrong.

A rating decay system is used in Tennis, where it is a single constant pursuit of professional players - this now seems to fit the bill better. (note this is the complete opposite of BGA). Even so, you get examples of players like Serena Williams that takes time off due to injury and/or pregnancy and sharply falls down the ratings. Comes back whoops everyone MUCH higher rated than them, and everyone starts to blame the ratings because it's so badly wrong.

If ELO is just a number to you, and you don't have a problem with the leader-boards, then you simply don't have a problem. What do you care what someone's number is?

There's been quite a few previous threads on this kind of subject. The main theme is, come up with something better than ELO....... This isn't it.

And as I've previously stated, and if you do care about this kind of thing, then play Arena mode. You don't need 2 different systems that will then be even closer linked. What's the point of having both now?

IMO: There is nothing wrong with ELO, is has strong mathematical foundations and works very well in almost all cases, which is extremely hard to try and achieve. Artificial balances with these calculations is normally wrong, the best you can have is the K-factor, which is used as some kind of 'time' adjuster.

Do not generalize far extreme cases. We are talking about overall system, not a specific person.
Legendary players can pull off such feats and thus earned their fame, but it is not the case for most of players.
Even if it matters a little, those extraordinary players very quickly recover own ratings regardless of decay.
Ultimately rating is an expectation, not some kind of absolute 'power level' meter.

Yes, ELO has mathematical foundation, however Arpad Elo intentionally simplified complex calculations due to computing power limitations.
It was created and established during 60~70s so simplification was necessary.
But now the situation has changed. The math behind ELO is not 'strong' anymore.
If there is nothing wrong with ELO, why revised systems are designed and gaining popularity over time?
Why gaming organizations add incentive for active playing or modifier for staying away instead leaving ELO as it is?

'What's the point of having both now?'
I really don't understand why BGA admins left ELO untouched while introducing Arena mode, so please ask to authorities.
I already said I'd rather choose Arena than current ELO if I were to decide.

[sprockitz]

Why gaming organizations add incentive for active playing or modifier for staying away instead leaving ELO as it is?

This is because they want people to play. People playing equals money for them.

If you get worse because you aren't playing, then your Elo will drop when you start playing again. Artificial decay is bad. If you aren't playing then you won't be ranked so the Elo doesn't matter to anyone.

[ufm]

This is because they want people to play. People playing equals money for them.

Chess.com and Lichess, the largest and the second largest online chess communities, use Glicko-2 over ELO. Yes, chess communities.
Lichess doesn't even have ads and runs on donation. The list continues and is growing over time. Australian Chess Federation, FICS...
Of course people playing equals money for every community, but it is not the only reason for implementing rating decay.

If you get worse because you aren't playing, then your Elo will drop when you start playing again. Artificial decay is bad. If you aren't playing then you won't be ranked so the Elo doesn't matter to anyone.

Why do you think decay is artificial and ELO is not?
ELO itself is an arbitrary system created to predict relative skill levels by definition. It is not set in stone.

'If you aren't playing then you won't be ranked so the Elo doesn't matter to anyone.'
If it be true decay also does not matter to anyone, including the person having decay, for the same reason. Then why does decay bother you?
One of the purpose is automatically reflecting possible rust. That argument is missing the point.