Can we fix team ELO please ? At least partially ? Something ?

[Phoxtrot]

There are many games where you cannot possibly gain ELO if you don't finish first.
For instance in Innovation, play a 3p or more game and only the winner can gain ELO because people "on the loosing team" cannot win ELO.
Now, games that are not actually not team games can get around this with the proper settings, this issue was solved for King of Tokyo for instance.

But...
The whole handling of team ELO in the first place is WRONG.

1) One thing that should change when you are in a team, losing or winning, is that you should not be compared to your teammate. To win or avoid the loss of some ELO because your teammate has an higher ELO than you is absurd (and vice versa) .

-------
Now, to make team ELO changes fully "fair" is difficult and probably subjective.
If I am ELO 100 in team with an ELO 300 and opposed to 2 ELO 200 guys and we win, the current situation is that I win ELO for drawing with my team mate (absurd, see point 1) but I also gain more for my matchups against my opponents than my teammate does and that is probably not right. This really mean that you have everything to gain and nothing to lose by teaming with an higher ELO.

2) So on balance, I think it would make more sense to compare the team average ELOs or something like that.

I'm sure that this is an issue that has been tackled in game theory somewhere by someone so ideas are welcome.

[tarongot]

There are many games where you cannot possibly gain ELO if you don't finish first.
For instance in Innovation, play a 3p or more game and only the winner can gain ELO because people "on the loosing team" cannot win ELO.
Now, games that are not actually not team games can get around this with the proper settings, this issue was solved for King of Tokyo for instance.

But...
The whole handling of team ELO in the first place is WRONG.

1) One thins that should change when you are in a team, losing or winning, is that you should not be compared to your teammate. To win or avoid the loss of some ELO because your teammate has an higher ELO than you is absurd (and vice versa) .

-------
Now, to make team ELO changes fully "fair" is difficult and probably subjective.
If I am ELO 100 in team with an ELO 300 and opposed to 2 ELO 200 guys and we win, the current situation is that I win ELO for drawing with my team mate (absurd, see point 1) but I also gain more for my matchups against my opponents than my teammate does and that is probably not right. This really mean that you have everything to gain and nothing to lose by teaming with an higher ELO.

2) So on balance, I think it would make more sense to compare the team average ELOs or something like that.

I'm sure that this is an issue that has been tackled in game theory somewhere by someone so ideas are welcome.

I would like that too, but chances are 0. I don't play team games, but I play a lot of cooperative games, that can be treated as team game. And in these games it is even worse . Most of them doesn't have ELO/ranking system at all. Those few who have, and are proof that regardless of some people and vocal game developers opinions it works, aren't free of bugs and faulty logic. Like when you LOSE ELO points for game that you WON .

[ChiefPointThief]

Is the fix that hard? Taking into account teammates elo their should be a rating per team instead of per player. Then even pts distributed for all members of each team.

[FrankJones]

Is the fix that hard? Taking into account teammates elo their should be a rating per team instead of per player. Then even pts distributed for all members of each team.

Agreed; this seems like an easy and obvious fix.

[yakko_]

It does seem like an easy an obvious fix, and the team rating problem has been especially frustrating in the one game I've played competitively in arena. (The team rating system problem affects arena ratings just as it does persistent Elo ratings, compressing the player range by effectively transferring points from the higher-rated to the lower-rated partner in each team every game.)
There are old, oft cited and still open bug/suggestions that seem to suggest the BGA team simply isn't interested in changing the rating system, but perhaps existing ones suffer from being tied to other complaints about ratings in multi-way games (which apparently have been changed from an even older approach, of treating teammates as opponents, to the current approach of ignoring them entirely, which is still wrong). Maybe it's worth creating a bug/suggestion specifically for team games.

[tbhp]

Is the fix that hard? Taking into account teammates elo their should be a rating per team instead of per player. Then even pts distributed for all members of each team.

It seems to me that this is what the current system achieves. If two players keep teaming up in all their games, then their Elos will eventually become one and the same. Since after all the system can't guess which player contributed more to the victories, that seems to be the normal behavior.

If even points were distributed, this wouldn't achieve a team rating, because the Elo gap between teammates would forever remain the same.

[yakko_]

The rating system is for individual players. The idea of rating teams, and expecting that both players on a team should have the same rating, might make sense if each team of two were tracked as a distinct entity and each player had many ratings with which they were associated, one for each partnership, but that is not going to happen.
In a game like spades, a player plays with many different teammates from game to game. Arena matchmaking in particular enforces this. What ChiefPointThief, Frank, and I and others would like to see is a rating system that assesses chances of winning for the two teams within each game and distributes rating points for the game result accordingly, consistently with the Elo system's basic principle for 1v1 games. Currently, if I'm at, say, 450, and get into a match vs. a pair of 300 players, the system expects the 450 player to win a certain % of the time, >50%, and gives out points after the game in accordance with that, supposedly so that we all maintain our ratings if we win and lose at that expected rate. But that says nothing about the 450 player's partner. If I'm paired with another 450 player, or if I'm paired with a 150 player, our chances of winning that game are going to change drastically, but the current system assigns my rating adjustment the same in both cases, as if my chance of winning depends only on my strength and that of my opponents.
Put another way, all members of a team win or lose each game together. It is not possible for one partner's chance of winning to be 75% in the same game that the other partner's chance is 25%. But the current system awards points as if it were.

[FrankJones]

The rating system is for individual players. The idea of rating teams, and expecting that both players on a team should have the same rating, might make sense if each team of two were tracked as a distinct entity and each player had many ratings with which they were associated, one for each partnership, but that is not going to happen.
In a game like spades, a player plays with many different teammates from game to game. Arena matchmaking in particular enforces this. What ChiefPointThief, Frank, and I and others would like to see is a rating system that assesses chances of winning for the two teams within each game and distributes rating points for the game result accordingly, consistently with the Elo system's basic principle for 1v1 games. Currently, if I'm at, say, 450, and get into a match vs. a pair of 300 players, the system expects the 450 player to win a certain % of the time, >50%, and gives out points after the game in accordance with that, supposedly so that we all maintain our ratings if we win and lose at that expected rate. But that says nothing about the 450 player's partner. If I'm paired with another 450 player, or if I'm paired with a 150 player, our chances of winning that game are going to change drastically, but the current system assigns my rating adjustment the same in both cases, as if my chance of winning depends only on my strength and that of my opponents.
Put another way, all members of a team win or lose each game together. It is not possible for one partner's chance of winning to be 75% in the same game that the other partner's chance is 25%. But the current system awards points as if it were.

Exactly. It's nonsense.

[tbhp]

What is the way it works for esports? Many games work in teams and have ranking systems that seem to work appropriately (they have to when money is on the line and you want a player's rank to be a reliable indicator of their skill). Why doesn't BGA use the same system that is used for those games?

[FrankJones]

What is the way it works for esports? Many games work in teams and have ranking systems that seem to work appropriately (they have to when money is on the line and you want a player's rank to be a reliable indicator of their skill). Why doesn't BGA use the same system that is used for those games?

Inertia.

It just takes too much effort, time, energy, etc, to change something that is already in place. It's likely on their length list of things to fix, in some mysterious order of priority.