Elo broken?

Forum rules
Please DO NOT POST BUGS on this forum. Please report (and vote) bugs on : https://boardgamearena.com/#!bugs
Post Reply
User avatar
MasN
Posts: 27
Joined: 26 December 2015, 23:45

Elo broken?

Post by MasN »

I believe, if you win a game as hunted, it counts as a tie against the other hunted. This means that you can lose elo even when winning.
User avatar
Ossido
Posts: 3
Joined: 08 June 2018, 16:16

Re: Elo broken?

Post by Ossido »

There seems to be a problem with the elo score, but I'd point out that it is not so much an issue with joint placement, but with the way team games are implemented. At the end of the game you can see that (with a mouseover) that the algorithm that updates the elo score matches you against all players, included your teammates, and computes your expected performance against them with the real one. This should only happen between the monster and each hunted (or in general between members of one team and the other), but not between teammates, because the possible outcome between teamates can only be a tie.
User avatar
RicardoRix
Posts: 1757
Joined: 29 April 2012, 23:43

Re: Elo broken?

Post by RicardoRix »

same difference.

The problem is the same with Lewis & Clark where everyone always finishes joint 2nd if they didn't win.

When comparing each other with the same place would be considered a draw between those 2 players and if they have differing ELO rank then the ELO score will be given to the player with the lower ELO from the higher ELO player, that's not correct in ANY joint placement case.
User avatar
Ossido
Posts: 3
Joined: 08 June 2018, 16:16

Re: Elo broken?

Post by Ossido »

I disagree that it is the same difference (or it's the same difference to a different extent, if you prefer).
The issue at heart is the same: using the elo score, which works for two player matches, in multiplayer games.
The way it is implemented, the multiplayer match is treated as a lot of two-player games between all possible pairs of players. This works better in games with victory points, in which all rankings are likely to appear (e.g. 7 wonders).
In a game like Lewis & Clark not all rankings are likely to appear but any given pair of players may end up in any position, just not in a way which is independent of the other pairs. So what fails here is that the basic and simplistic assumption of the elo score (that the matches are all independent and won according to a probability a priori which is a function of the difference of the scores) breaks down.
In the case of team games this failure is extreme, in the sense that not only the results among pairs of players are independent, but given a pair of players (after teams have been drawn) not all possible outcomes are even possible.

The reason I want to point down this difference is that, while it seems difficult to come up with a fix for the general problem (do you have suggestions?), it seems easy to fix the problem with team games (or at least to bring it to the level of the other multiplayer games) by discarding the matches among members of the same teams.
In fact a game like Not Alone is a match between two teams, and elo score should work great for it.
User avatar
tanaraerdoth
Posts: 1
Joined: 02 August 2017, 00:36

Re: Elo broken?

Post by tanaraerdoth »

Yeah, it should be that winning with a lower ranked team means you did a better job.
Post Reply

Return to “Not Alone”