Board Game Arena

Ant Stewart wrote: ↑04 September 2026, 16:34 Not really as the current ELO is also based on those settings.

Then your current elo was overestimated, probably by playing with with other players with underestimated elo.

The point is, if the difficulty score of some settings is 700 (say), then consistently reaching perfect score with those settings should lead to an elo of 700. Anything higher is an artifact and makes no sense (with those settings), from the point of view of what elo is supposed to measure.

Jellby wrote: ↑04 September 2026, 19:24 Then your current elo was overestimated

If I play with 3 experts, or 2 experts and one master, or even sometimes 2 masters, one expert - the ELO usually goes up slightly, maybe 1 point over one or two perfect games.

Therefore it is clear that my ELO is below that of a theoretically perfect player in my settings.

Whether or not I am a perfect player (spoiler - I'm not) is irrelevant - the issue is that a player that is perfect with my ELO would be underrated, and could still loose ELO for a perfect game.

So the statement that a negative ELO for a perfect game means you are overrated is not strictly true.


----


In short in all non-coop games with ELO - gaining ELO as you gain ranks is always possible but becomes exponentially harder the better you get. There are diminishing returns as you get stronger.

However, it is never meant to be the case that this flips, and the diminishing returns become negative.

As an extreme example - suppose you have a 3000 ELO player playing a 100 ELO player.
That is a massive difference in ELO - but it works out as an increase of 0.0000018 (for K=32).

That is how ELO is meant to work, and I think that capping negative ELO for perfect games gets you closer to that.