Proposal: Elo rating should be Arena-only, Play Mode-only, split into separate ratings, or apply under equal conditions

[Ceaseless]

the relative ELOs are driven as much by volume as skill.

What do you mean by this? What impact are you thinking of for volume?

[Jontia]

the relative ELOs are driven as much by volume as skill.

What do you mean by this? What impact are you thinking of for volume?

Imagine two people who win every game they play. One plays one game a week, the other one game a day. Which has the highest ELO?

The same is true for two players of any positive win rate. If you play more, your ELO will be higher over time.

[LPV123]

My instinct is to agree with you. That said, I could use some additional clarification about the purpose/benefit of higher ELO. Is it just a question of the being happier with a higher number? Does it affect who we are randomly paired with in regular play mode? Does it impact the Arena rankings? I just don’t know how it all works.

I tend to think play mode and arena and tournaments should be entirely separate (no shared ELO), both for the reasons originally mentioned and others, some of which have already been raised, such as the often-dramatic difference in game elements between/among play mode, arena, and tournaments. Yahtzee is one example of a game where earning wins can be much easier depending on the rules, which are locked in for arena but vary in play mode. A player might choose to skip a difficult arena season altogether because the risk to ELO, for instance.

[Ceaseless]

the relative ELOs are driven as much by volume as skill.

What do you mean by this? What impact are you thinking of for volume?

Imagine two people who win every game they play. One plays one game a week, the other one game a day. Which has the highest ELO?

The same is true for two players of any positive win rate. If you play more, your ELO will be higher over time.

No, the same is not true of players with any positive win rate. Elo doesn't go up just because your win rate is positive, nor does it go down just because your win rate is negative. If you beat a player 3 times for every 1 time you lose, then the elo will set itself so that you lose 3 times as much for a loss as you gain for a win.

[Ceaseless]

My instinct is to agree with you. That said, I could use some additional clarification about the purpose/benefit of higher ELO.

Elo is like a weight around your neck, the higher the number the heavier it gets. The higher your elo, the more points you lose for losses and the fewer points you gain for wins. In other words, the higher the elo, the more effort it takes to keep it where it is when playing. So if a 300 played certain set of games with a different opponent each game with no rating change, then a 400 somehow replaced the 300 for that exact same set of results vs the same players at that time instead, the 400 would lose rating. If a 200 did it instead of the 300, they would gain rating. It doesn't matter if the results had been 6 wins and 3 losses or 3 wins and 6 losses.

Players carrying roughly the same weight can play each other are expected to get roughly even results. The larger the gap, the less even the results. It gives players a ballpark of who they can play to get reasonable competition, or I guess a player can use it the other way around and target easy wins or impossible odds, though their opponents have access to the same information as they do.

Yahtzee is one example of a game where earning wins can be much easier depending on the rules, which are locked in for arena but vary in play mode.

As an aside, how difficult a ruleset is for elo is a player-by-player thing. Ruleset A may be easier for some players and harder for others compared to Ruleset B, and vis versa. The players will have different ratings for each ruleset. If one player exclusively played under ruleset A, then they'll probably do worse in ruleset B, and vis versa.

[Jellby]

Collusion to game the elo system is a punishable offense. If you see it, report it. No amount of tweaks will eliminate it.

[Nils meeple]

Collusion to game the elo system is a punishable offense. If you see it, report it. No amount of tweaks will eliminate it.

That's the theory, but in practice a lot of obvious cheating reports go unanswered and unrelieved for months, if ever. That's exactly the point of this proposal — make it structurally harder to manipulate Elo in the first place, so the system doesn't rely so heavily on manual reporting and moderation, since moderators clearly don't have the resources to keep up with it.

[Jontia]

What do you mean by this? What impact are you thinking of for volume?

Imagine two people who win every game they play. One plays one game a week, the other one game a day. Which has the highest ELO?

The same is true for two players of any positive win rate. If you play more, your ELO will be higher over time.

No, the same is not true of players with any positive win rate. Elo doesn't go up just because your win rate is positive, nor does it go down just because your win rate is negative. If you beat a player 3 times for every 1 time you lose, then the elo will set itself so that you lose 3 times as much for a loss as you gain for a win.

Eventually, yes. But fresh players are always coming in and there is always more ELO to gain. The more you play, the higher your ELO will go compared to someone just as good as you that doesn't play much. Say in your example we play each other 4 times, you win 3 and I win one, but those are my only games whereas you've been playing against other people and are winning all those games. Who's ELO has gone up over the period?