If I'm understanding the concern correctly, it sounds like we're worried that games with lower K-factors will automatically have smaller Elo spreads and lower Elos for top players. I don't think that this is true to a significant degree, at least in the long run.ExaltedAngel wrote: ↑15 January 2026, 05:01 It makes sense. The only argument against it is that the Elo ratings of the top players in every game are currently a fairly good indicator of how much luck impacts the game, and by doing this we would lose that signal in favor of an arbitrary 1-5 rating that inevitably will be imperfect, if not plain wrong, for many games
The K-factor should only effect the rate at which Elos move, not the number they're moving toward (your "true strength"), if that makes any sense.
One way to see why this is is to observe that the K-factor does not appear in the formula for predicting win-rates between players of different Elos:
your predicted win-rate against a player rated x Elo points lower than you is 10^(x/400) to 1.
So for example, if my long-run win rate against 200 rated players is 10 to 1, and I play exclusively against 200 rated players, my Elo will eventually settle to around 600.
A lower K-factor means it will take more plays for my Elo to get to around 600, and that once it's there it will vary less from 600, but it doesn't mean that my Elo would instead settle to around 400, or something like that.
So I still think the Elo ratings of top players would be a reasonable (though imperfect for a few other reasons) proxy for "how much luck impacts the game," as you put it, provided that the game in question has been popular enough for long enough.
I think some of this math is fuzzied by bga's hard Elo floor at 100, but I don't see why the principle wouldn't hold.
