Ideas/comments on how this score should affect rating changes?

[fbosnic]

No description provided.

[EduardEdiJerkovic]

I think that the impact factor should not have a high variance even if the score difference is large. I would suggest that factor should not be higher than 1.5 than the base one because it would introduce high volatility in the Elo rating.

[fbosnic]

Would you prefer to use no impact factor at all?

The initial thinking was something like:
If a team wins by a margin of 3 goals then that team was probably better and we can be confident about the rating changes that will be made.
On the other hand, if they win by only one goal then it might have been just luck so we cautiously update the rankings by a small amount.

The current formula is:
min(3, abs(home_goals - away_goals))
we could change it to:

d = abs(home_goals - away_goals)
if d == 0: return 0
elif d == 1: return 1
elif d == 2: return 1.5
elif d == 3: return 2

or something of the sort