Board Game Arena

This is kind-of odd result:
https://boardgamearena.com/3/sevenwonde ... e=54548978
Getting negative ELO for the first place is, well, unexpected. Win is a win, so could BGA coders please either don't call it "the first place" (this implies winning), or fix the code to give at least 0.01 ELO to a winner

Thank you and good luck

One way to think of it is: Elo adjusts based on expectation v. reality. Since you went into the game with the highest Elo, you were 'expected' to win, thus a tie for first is a reality that is lower than the expectation, and you lose Elo. The only way for you to not lose Elo when you play with all lower-Elo players is to win alone.

Technically any bug with proper explanation can be turned into a feature

You came joint 1st. You could have done better, you could have won out-right. You drew with tatto17 which would give you the negative result.

6element wrote: ↑08 July 2019, 03:47 This is kind-of odd result:Win is a win, so could BGA coders please either don't call it "the first place" (this implies winning)

Actually, the recent framework change makes an explicit difference between placement and winning, as you can see in the comments here: https://boardgamearena.com/bug?id=12256 Since then, in games where there only are winners and losers (notably, no ties between the two) you can no longer lose Elo points if you were part of the winning team.

A first place may very well imply winning to you, but it certainly does not imply a sole win if more than one of them are awarded after the table ends. The way BGA implements Elo changes in such scenarios is what you're observing as (slight) loss: it can be called counter-intuitive or opaque or even unfair, but not a bug. That being said, as this topic pops up frequently, better communication to players and/or an improved handling of ties especially for first place seems necessary. Hopefully it's being looked into soon.

A way to avoid this to happen is to play with players on same level.

Silene wrote: ↑08 July 2019, 13:56 A way to avoid this to happen is to play with players on same level.

I'd wish to follow this advice, but there aren't enough 300+ 7 Wonders players, unfortunately. This would mean waiting for half an hour which is why I moved from BSW to BGA.

eoc wrote: ↑08 July 2019, 13:55

6element wrote: ↑08 July 2019, 03:47 This is kind-of odd result:Win is a win, so could BGA coders please either don't call it "the first place" (this implies winning)

Actually, the recent framework change makes an explicit difference between placement and winning, as you can see in the comments here: https://boardgamearena.com/bug?id=12256 Since then, in games where there only are winners and losers (notably, no ties between the two) you can no longer lose Elo points if you were part of the winning team.

A first place may very well imply winning to you, but it certainly does not imply a sole win if more than one of them are awarded after the table ends. The way BGA implements Elo changes in such scenarios is what you're observing as (slight) loss: it can be called counter-intuitive or opaque or even unfair, but not a bug. That being said, as this topic pops up frequently, better communication to players and/or an improved handling of ties especially for first place seems necessary. Hopefully it's being looked into soon.

I understand the explanations and I'm not really upset because of 1 point ELO loss, I'm just saying - winning is winning, and clearly 1st place is winning. In this game there have been two winners and one loser. A sign of ELO change (not the magnitude!) should reflect it.

A site which "rewards" winning with negative change in ranking, is a bit broken. It is a bug.

But, as I said - with proper explanation you can claim it is "expected behavior", "by design", a feature, or a minor bug not worth fixing. I actually code games for living, I made these arguments myself many times

You didn't win, you came joint 1st - there is a difference, the ELO calculation is correct.

You want to know more about a very similar situation that is wrong then:
https://boardgamearena.com/forum/viewto ... =3&t=13755

6element wrote: ↑08 July 2019, 23:59 I'm just saying - winning is winning, and clearly 1st place is winning. In this game there have been two winners and one loser.

I understand where you're coming from, and in many offline groups that would indeed be the interpretation of what happened Perhaps it helps to think about it this way:
Your "co-winner" performed better than expected, since they managed to tie with a player whose Elo rank was higher going into the game. That means their Elo rank should increase since they played better than expected (probability-wise)! And if there's Elo gain on the one side, it must come at the expense of another player – the "better" one, who was expected to put enough score difference between the two (again, expected w/r/t probability).

(And yeah, as mentioned earlier, the notion of winners vs losers only applies to team games such as Not Alone. Instead of placement results, BGA as of recently does indeed show "Winner" / "Loser" since they're well-defined for these games.)