Team games -- Equal placement=equal ELO?

[Jest Phulin]

This is primarily to create a place for a discussion in another thread, so that thread doesn't get hijacked.

Assume a team-v-team game where players are A, B, C, and D.
A and B are on the same team, and win the game.
C and D are on the same team, and lose the game.

ELOs have been established, and are a fair representation of player skill.
A has an initial ELO of 300.
B's is 100
C and D are at 190 and 210 respectively.

The question becomes, should A and B both gain the same ELO for the win?

If yes, then A is effectively gaining more ELO, because it will be (roughly) calculated on a team average ELO 200 vs 200, not a 300 vs 200.
If no, then B is getting rewarded for skill they do not actually possess -- they are using A's skill to win the game.

Aaaaand, discuss.

[Caffe Latte]

If no, then B is getting rewarded for skill they do not actually possess -- they are using A's skill to win the game.

Ahemm, B skill was used too to win, since we call them team. Are there many games one "with no skill possessing" can just dovetail high eloed one and "win"? And masters, knowing that, still play with low elos in team?

Name some, I will go to collect ELO there

[Silene]

The ELO you gain or lose depend on the outcome and on the winning-propability you had in the game. Winning-propability is calculated from ELO-difference of the players (or in team game: opposing teams).

When 2 players form a team, they have exactly the same winning propability because they can only win or lose together.

So they should gain/lose the same amount of elo if they win/lose.

[RicardoRix]

Silene is correct.
If the K factor is the same, then there is no reason why there should be an unequal share of points.

[eoc]

Except if you keep dishing out identical ELO gain to all team members, you might never end up with Elo rank as fair representation of player skill. It's an interesting discussion where I could reasonably support both sides, so here's why I think I'm in favor of identical rewards/penalties across all team members: BGA already conflates Elo rank and, to a lesser degree, victory percentage / trophies etc. between the many supported player counts per game. It's unquestionably easier to win 1:1 games than it is with four players on the table, but stuff like the "Most victories" trophies don't care about that. In similar fashion, some games become a lot less unpredictable or at least lose part of their non-strategic elements as well as the at times fairly random element of another human being inserted into your plans, if played with only two players. And yet other games play completely differently at various player counts, with certain options enabled or disabled, or are frankly a literally different game (Saboteur 2 implementation as option for Saboteur 1 comes to mind).
So the 1:1 setup serves to tie actual game outcomes more closely to player skill and thus ELO point rankings computed from those outcomes to player skill, but there's no way to tell from a player's ELO points which of those were gained from which set of options, player counts, team sizes – how could we defend "no" as answer to the original question under these circumstances? "Yes" is perceived as more fair, won't hurt the mathematical representation, and people looking to game the system will have to be addressed site-wide anyways with competitive team games, it's not just a matter of ELO.

Lots of words to basically say: Elo rank is already meaningless enough that doing the "right" thing won't further reduce its meaning.

[vigorousRoll]

If equal ELO points are being handed out to winners of a team game, a very high rated player will be gaining several points more per win than he should be. He might start seeking out team games exclusively to milk this imbalance. What you want is a site-wide rating system where the players are indifferent, from an ELO standpoint, between whether they choose team or individual games.

A good rating system gets players to an equilibrium rating that reflects their true ability quickly. If we're under-rewarding a low rated player for a victory in a team game because we're handing out equal points for a win, it's just going to take that much longer for his rating to reach its equilibrium value. And by over-rewarding a high rated player, we'll be disturbing the equilibrium that his rating may have already reached.

[vigorousRoll]

The question becomes, should A and B both gain the same ELO for the win?.....


If no, then B is getting rewarded for skill they do not actually possess -- they are using A's skill to win the game.

I've been in a handful of Innovation 2v2 team games where I've played the part of the "player rated 300" while my partner has been the "guy rated at 100". For each turn that I take, my low-rated partner will be taking his turn two moves later (after one of our opponents has had his say.). So we've been distributed an equal amount of tempos with which to influence the game. Sometimes all that is necessary for the win is for my partner not to blunder so badly that it mucks up our chances. Other times there will be a critical move that he has to make and he often finds it. And on a couple of occasions, the cards have smiled so sweetly upon my low rated partner that I simply played the role of the point guard while he was hitting 3's and slam-dunking his way to victory.

If I would characterize why me and my low rated partner have won some of these 2v2 Innovation games, it's because our 200-rated opponents made some oversights (not prioritizing defense at a critical moment, say) or passed up opportunities that came their way while we capitalized on ours. [Edit: I think what I was trying to say in this 2nd paragraph is that my low rated partner made less mistakes than our medium-rated opponents or that his mistakes didn't wield as much damaging impact and that thereby he's worthy of receiving an ELO boost ...because mistakes can mean the difference in winning or losing every bit as much as 'making good moves' can.]

[RicardoRix]

As your ELO partner rating goes down (depending on who you partner ELO400, ELO300 ,ELO200,ELO100 - all different potential partners), your chances of winning also decreases.

So in order to address the balance, when you partner ELO100, the ELO gain you need to get for a win needs to be higher simply because the chance of you winning is lower.

If not, then high ELO players (or in fact everyone) will simply NOT want to partner lower rated players and seek out the mostly highly rated players.

The argument about who goes first in the team is mute, on average you'll go first 50% of the time in 2 player teams and this evens itself out in the long run.

[Schtitn]

ELOs have been established, and are a fair representation of player skill.

I think all the matter is in this first hypothesis.
If it's true, the game is tied, Gods of games flip a coin to decide who win and ranged gaming has no sense.
To say it in another way : if it's true, let's have this game a million time, elo of each player are the same than at the beginning. So we must don't care if calculation is fair for one game because the "system" will correct it by time.
So Elo is always a false representation of skills, showing a picture of a moving dynamic that is players's progress.
And a new game between some players is a new false picture of skills of players.


So to answer your question : yes each players win or lose elo regarding their own elo at the precise moment of the game.
So yes B is using skills of A. It is just what team games is about.
He become 120 maybe. And a next game will correct it to 140... or to 100 back again. Because no matter who you playing with Elo is an average of your level based on past,
Elo is more false when introducing team gaming and personnal ranking can't fairly show skills at multiple player game.
Let's have a team ranking !

[Tulivuori]

Assume a team-v-team game where players are A, B, C, and D.
A and B are on the same team, and win the game.
C and D are on the same team, and lose the game.

ELOs have been established, and are a fair representation of player skill.
A has an initial ELO of 300.
B's is 100
C and D are at 190 and 210 respectively.

If we assume that chances of win are 50-50, after all the sum of ELO's are equal, anything else than the same gain/lose for both in team would be irrational.

However, ELOs are not a fair representation at least for new players, so it might have unwanted consequences. However, with the way it is now (a win/lose has equal effect regardless of with whom of those they play with) is ridiculous.