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

[Nils meeple]

Hi everyone,
I'd like to propose an idea to the moderators and get the community's thoughts on it: what if Elo rating only applied to Arena mode, and was completely removed from Play mode (games with friends)?
The main reasoning behind this is that Play mode currently allows for artificial Elo inflation — players can boost their rating by playing (and losing on purpose, or arranging easy wins) against friends. This doesn't happen often, thankfully, but it does happen, and it undermines the integrity of the rating system for everyone else.
There's also a practical side to this: moderators already have limited time and resources to investigate and act on this kind of abuse. If Elo simply didn't apply to Play mode at all, there would be nothing to exploit in the first place — which means less manual policing needed, and moderators could focus their attention elsewhere.
To be clear, this wouldn't remove Play mode's stats or functionality — just decouple it from the Elo rating, which would remain fully tied to Arena mode only.
Curious to hear what everyone thinks — does this seem like a reasonable change? Any downsides I might be missing?

[Jontia]

The ELO system is currently used to gate-keep access to the Arena, so you'd need to replace that system.

You'd also need some system to let people who dislike the current Arena settings track their progress. While ELO and indeed Arena ratings are not any kind of perfect ranking system, and getting over invested in them is probably unwise, it is still good (and fun) to watch those changes over-time. But there are several games where the Arena settings are stuck in a rut, with choices that some players don't like. Those players should still have an ELO to track to let them see how they are getting on.

Finally, Simple game and play with friends are two different options as far as I know. I'd be surprised if play-with-friends actually affected ELO at all. Certainly I've never seen an ELO adjustment screen following a game I've played with friends, but maybe if additional seats on the table are opened up to anyone, then it does effect ELO?

Ultimately I don't think this is workable.

[Nils meeple]

In Mountain Goats, there are two modes: Arena and Play. In Play mode, you can either invite friends directly, or wait for automatic matchmaking — but even with matchmaking, before the game starts you can see your opponent's name and choose to either accept or decline the match. In practice, this isn't much different from playing with friends directly.
In other words, Play mode gives players full control over who they end up playing against, and yet every single game in this mode still affects Elo rating. This essentially turns the rating into a fake metric — one that no longer carries the meaning we'd normally expect from the word "rating."
Here's an example of what rating-boosting games can look like (taken from match statistics between two players, names omitted since this topic isn't about calling out specific players, just illustrating the pattern): 18 wins in 18 games (100%).

[Ceaseless]

Why not just have rated games disallow players to choose their opponent? I'm not a fan of playing Arena for two reasons that are tied to each other. One is the rating resets each format, which I don't want to do on a game-to-game basis given how many different games I play. Another is the league system, which is terrible and as a result makes Arena's league filters greatly inferior to regular rating's filters.

[Nils meeple]

Why not just have rated games disallow players to choose their opponent? I'm not a fan of playing Arena for two reasons that are tied to each other. One is the rating resets each format, which I don't want to do on a game-to-game basis given how many different games I play. Another is the league system, which is terrible and as a result makes Arena's league filters greatly inferior to regular rating's filters.

I agree with your point — it wouldn't be fair to strip Elo rating away entirely from players who don't want to play Arena. I think you're right: if Play mode simply hid the opponent's name before accepting a match (the same way Arena does), and invite-only games with friends didn't count toward rating at all, then it would actually be a real rating — honest and consistent for everyone, regardless of whether they're playing Arena or Play mode.

[Nils meeple]

I don't know if this is the case in all games, or if it's just an imperfection in Mountain Goats' rating system, but here the rating is calculated across all games indiscriminately — with no distinction between invite-only matches and games where the opponent was matched automatically. And on top of that, as mentioned above, there isn't really any "automatic" matchmaking in Play mode to begin with, since you can see your opponent's name and choose to accept or decline anyway.

This is a really interesting discussion. Once different opinions have been shared, we can put them together to figure out exactly how this system could be improved — so that it genuinely gets better without stepping on anyone's interests (except cheaters), and then post the final, refined version in the "Bugs & Suggestions" section.

[cigma]

I'd like to propose an idea to the moderators and get the community's thoughts on it: what if Elo rating only applied to Arena mode, and was completely removed from Play mode (games with friends)?

I don't agree with your suggestion. I like to play Elo-ranked normal games with my friends or other people in order to see who plays better and how I improve.

I don't play arena mode, so I might be wrong here. But as far as I know, the arena games use a different kind of ranking (leagues). So I wonder why Elo would affect arena games at all (besides the 100 Elo you need for entering the arena), not to even think about "undermine the integrity of the rating system"?

[Nils meeple]

Cigma, you're right that ELO rating doesn't affect the arena results. But the thing is, all players participate in the ELO rating - both arena players and play mode players - but under different conditions: arena players can't choose their opponents, while play mode players can. That's why, in my opinion, either the conditions should be the same for both, or the ELO rating should only exist for one of the groups - either arena or play mode players.
Or there should be two separate ELO ratings, one for each of the groups.

[Jontia]

Cigma, you're right that ELO rating doesn't affect the arena results. But the thing is, all players participate in the ELO rating - both arena players and play mode players - but under different conditions: arena players can't choose their opponents, while play mode players can. That's why, in my opinion, either the conditions should be the same for both, or the ELO rating should only exist for one of the groups - either arena or play mode players.
Or there should be two separate ELO ratings, one for each of the groups.

Outside of mountain goats, there are much more significant differences in terms of gameplay options available. Like Catan base Vs Catan C&K, which are in any real sense different games. 2-player vs 5-player Ticket to ride is a hugely different game in terms of how to play and play well. Dice Hospital, which I do pretty well at with the base game, but have never won a game with the expansions added in any of the 4? 6? variations. Even in terms of all games, Turn-Based vs Real-Time have very different player bases. When I switch to my rare real-time game, I almost never seen the players I'm regularly matched against for Turn-based games.

All these settings run under the same ELO rules. For the same reasons that it isn't worth seperating Arean off from play in terms of ELO.

1) Shrinking the player pool makes ELO varitations more extreme.
2) Unless you play the same people over and over and you play the same amount of games, the relative ELOs are driven as much by volume as skill.
3) It's just for fun, and it's just for you. No one is looking to see how well you're doing at Mountain Goats to recruit you to fight the galactic Empire. The only real thing about ELO is if your own is going up or down. It's relation to other players is of limited value at best.

[Nils meeple]

Fair point on variance across game types — that's true and probably unavoidable. But that's a different issue from what I raised. Complexity/player-count variance is just natural noise you can't game. Opponent selection is different — it's a mechanic someone could exploit to inflate their Elo, not just random variation. That's the part I think is worth addressing separately.