Heavily flawed ELO calculation (?)

[clementia]

I noticed this earlier already, but my last game made it even more obvious and questionable, to say the least. I was drawn together with someone with ELO 1. We lost with terrible cards even if we somehow managed to reach 1 vs 1 where we were still kind of unlucky several times.

Anyway the result showed -20 which was astonishing for me. Looking at the details, I noticed that not only your result vs the other team counts, but also your relative difference to your partner. So it somehow counts any result apparently as a 'draw' with your partner, which means: You are much higher and win -> get a lower gain than p who plays with a (supposedly) more skillful player. You lose -> you lose more while expectedly your p contributed more to it.

This is kind of mind-boggling to me as it's reverse to what should be expected as a fair system. I also wonder if it's the same in other partner games, I have no history to conclude something from here.

https://boardgamearena.com/table?table=64253818

[sprockitz]

same problem for all partner games.

as well as for any game where someone quits

in both cases everyone is considered to have tied with their partner or with the other players who didn't quit.

At very least this should be a no result. other option is to use an average Elo for partner games, but may not be easy to conclude who partners are in a few select game situations.

[clementia]

What I haven't pointed out, though is naturally implied in the text: The higher your partner's ELO, not only you lose less / win more in any result, but you also in theory win more games in total. So there are multiple layers where it just doesn't work out. Every player would want to play with the best as partner, but not with the lower ones, and 2 persons (in a 4 p team game) will have to sustain that.

(Or if two similarly low encounter two similar high: the former ones would have preferred any other by all means, while the others are totally happy, all the more as they win very likely and with less 'subtraction' as mentioned before)