Page 1 of 2

ELO calculation for joint placements.

Posted: 13 June 2019, 12:14
by RicardoRix
https://boardgamearena.com/bug?id=12256

With 52 votes and no official answer or comment, what exactly is going on with this?

2 of the newest games, THE WEREWOLVES OF MILLERS HOLLOW and NOT ALONE, amongst others are effected.

It also seems like it might be an easy fix - don't calculate ELO changes between players that finished together...

https://boardgamearena.com/forum/viewto ... 770#p48770
https://boardgamearena.com/forum/viewto ... 08&t=13548
https://boardgamearena.com/forum/viewto ... 08&t=13551
https://boardgamearena.com/forum/viewto ... 71&t=13449

Re: ELO calculation for joint placements.

Posted: 14 June 2019, 01:38
by Silene
+1

This alone might be an easy fix. But there might be other stuff attached:

* Ties are not the only issue in ELO: i.e. balancing multiplayer (3+) player games, fixing resignation in coop-games (abandon in hanabi without penalty)
* ELO is not the only issue with ties: prestige-medals, gift-points, first-win-message are somewhat buggy too
* unfinished games are also somewhat concerned in both matters (abandon as well as when someone quit)

So maybe this small fix needs to be addressed on a larger scale...

Re: ELO calculation for joint placements.

Posted: 14 June 2019, 15:08
by sourisdudesert
Hello everyone,

We finally found the good way to implement this.

It concerns only games where the winners win as a team, and losers lose as a team. These games are easy to spot: they are the one where the "1st / 2nd / 3rd" are replaced by "Winners / Losers" in the table results. Examples: Werewolves, Not alone, Coup city state.

For these games, you cannot win ELO if you are a "loser", and you cannot lose ELO if you are a "winner".

Thanks for suggesting this.

Re: ELO calculation for joint placements.

Posted: 14 June 2019, 15:36
by eoc
Thanks a lot for taking a look, I'm sure many players will appreciate this change! :idea:

Re: ELO calculation for joint placements.

Posted: 14 June 2019, 16:11
by RicardoRix
Hi.

I see 2 potential problems with this.

1. Lewis & Clark doesn't have winners and losers, it has 1st, and 2nd, 2nd, 2nd.

2. People who win with high ELO should still potentially get points, not receive a bottom line of 0.

Re: ELO calculation for joint placements.

Posted: 14 June 2019, 16:17
by diamant
sourisdudesert wrote: 14 June 2019, 15:08It concerns only games where the winners win as a team, and losers lose as a team.
This is inaccurate. When a game is interrupted due to the departure of a player, the result is a team of winners and one loser. This case concerns all games on BGA with more than 2 players.

Re: ELO calculation for joint placements.

Posted: 15 June 2019, 05:16
by Liallan
diamant wrote: 14 June 2019, 16:17
sourisdudesert wrote: 14 June 2019, 15:08It concerns only games where the winners win as a team, and losers lose as a team.
This is inaccurate. When a game is interrupted due to the departure of a player, the result is a team of winners and one loser. This case concerns all games on BGA with more than 2 players.
+1

Re: ELO calculation for joint placements.

Posted: 15 June 2019, 12:36
by Silene
sourisdudesert wrote: 14 June 2019, 15:08 It concerns only games where the winners win as a team, and losers lose as a team. These games are easy to spot: they are the one where the "1st / 2nd / 3rd" are replaced by "Winners / Losers" in the table results. Examples: Werewolves, Not alone, Coup city state.

For these games, you cannot win ELO if you are a "loser", and you cannot lose ELO if you are a "winner".
Will there be the same amount of ELO given to all players in a team? I think there should (except of course factor for first x games will remain).

Not being able to win points when you're on loser side is one thing. But also I think all players shall lose the same amount (losses of weaker players shouldn't be negated by "ties" with other team-members (same goes on the winning side of course)

Re: ELO calculation for joint placements.

Posted: 15 June 2019, 13:00
by eoc
Silene wrote: 15 June 2019, 12:36 Will there be the same amount of ELO given to all players in a team?
Right now, the only modification in place is what you quoted. Aside from capping ELO loss at 0 in a win and ELO gain at 0 in a loss, nothing else has changed (yet).

Re: ELO calculation for joint placements.

Posted: 15 June 2019, 18:00
by vigorousRoll
sourisdudesert wrote: 14 June 2019, 15:08
We finally found the good way to implement this.

It concerns only games where the winners win as a team, and losers lose as a team. These games are easy to spot: they are the one where the "1st / 2nd / 3rd" are replaced by "Winners / Losers" in the table results. Examples: Werewolves, Not alone, Coup city state.

I'm heartened to hear that a satisfactory path has been spotted that can clear things up for these team games.