Lost ELO and Arena ranking in a [timed-out] game that I would have won

[zippzip]

Playing spacebase, I was ahead on the last turn with no chance for anyone else to win. The person in 4th place went unresponsive and timed out.

I have a high ELO, so naturally I stand to lose a lot if I place 3rd or 4th, and typically lose a little ELO even for 2nd place. But in this game, that I would have won, I lost ELO after the person timed out.

To me, something is wrong with the formula for calculating ELO in timed-out games if anyone can lose ELO (other than the problematic player). I get that it's not going to factor in the game state, but at worst a player should stay the same, no?

*

*Moderation Edit: please do not use the Forums to call out specific tables or players by name, this is what the Moderation Reports are for. The Forums should remain a general discussion.

[GTSchemer]

It's a known problem that a lot of people seem unhappy about. I don't know if it's because of the new elo system or what (see the massive thread at https://boardgamearena.com/forum/viewto ... 59#p171559).

As far as I understand, if someone is winning, another player (e.g. the person already in last place) can choose to time out or quit, causing the person who would have gained elo to stay the same or even, as in your case, lose elo.

It is hard for me to imagine that this behavior is intended, since it essentially gives a ragequitter power to tank an opponent's rating, but I don't know any BGA response on the matter so all we can do is speculate.

[RobertBr]

It's a known problem that a lot of people seem unhappy about. I don't know if it's because of the new elo system or what (see the massive thread at https://boardgamearena.com/forum/viewto ... 59#p171559).

Nothing to do with it. The issue with multi-player games has been apparent for a long long time. The problem is simple - no-one knows how to fix it, every extension of timing to more than two players runs into a problem of how to resolve issues of a time out unrelated to one of the players.

[ErikLevin]

The formula is wrong and the solution is extremely simple - non-quit players should not exchange Elo with one another.

The same error exists for all team games (where you will erroneously exchange Elo with your teammates) and all games that have only a win/lose state and no ranking (where you will erroneously exchange Elo with fellow winners/losers).

The errors have been known for years and no fix has been made by the BGA team, while they do find the time for other tweaks to the Elo system that nobody has asked for. So the chances of getting this fixed seem slim. I think stoicism is the best approach, personally - just playing and not focusing too much on Elo rating.

[GTSchemer]

It's a known problem that a lot of people seem unhappy about. I don't know if it's because of the new elo system or what (see the massive thread at https://boardgamearena.com/forum/viewto ... 59#p171559).

Nothing to do with it. The issue with multi-player games has been apparent for a long long time. The problem is simple - no-one knows how to fix it, every extension of timing to more than two players runs into a problem of how to resolve issues of a time out unrelated to one of the players.

I'm talking about a very specific case:

* Player A is going to win
* Players B, C, D are going to lose


As I understand it, if player D quits, sometimes player A will lose elo rather than gaining >= 0 elo.


The thing I do not understand is what abuse case could occur if they just added an if ( ) check to prevent losing elo only for the winning player if someone quits.

If I am mistaken and the winner can still lose elo even with no quitters, then that's another matter entirely.

[Gulchen]

@ GTschemer ​ :

Most likely, the same abuse case as for something else it's not clear how might be attempted.
How would such a check be coded?

[GTSchemer]

I realize I must be missing some detail, but if I were to naively program such a check it would be, using pseudo-code:


gainedElo = someCalculation();

if (isSingleWinner && wasResultCancelled) {
gainedElo = max(0, gainedElo);
}


That would mean the single winner of the game earns 0 elo if gainedElo calculates to negative in a cancelled game situation.

[Gulchen]

Since your earlier post used ​ ​ ​ "going to win" ​ , ​ ​ ​ I had thought you were including ​ ​ ​ "is WAY ahead" ​ .


From your reply to me, I now belive your idea is that ​ isSingleWinner ​ would
only be true in situations corresponding to ​ ​ ​ already crossed the finish line ​ .

If this is correct, then: ​ Yes, that would be an improvement, though it would
require letting game-code indicate to BGA when a player has done so.


however: ​ Are there any games other than Heat which continue after that point?

[Jellby]

There is no single winner if the result is cancelled. Unless it was a 2-player game, but then the elo change should always be positive. With more than 2 players, there are n-1 winners (everyone but the expelled player).

[GTSchemer]

Of course, but the code itself is still aware of the scores before they are cancelled. I'm referring to those scores.

And yes, Heat is probably one of the worst cases where this can happen, or maybe other racing games like Formula D. Challengers might have had it too, but it was specifically implemented to allow players to leave the game voluntarily.