Rating similarity confidence

[Luminiscental]

If someone has only rated say 2 maps in common with me, and it happens I rated the first higher and they rated the second higher, the rating similarity probably shouldn't display as extreme as a literal -1.

Maybe there should be a minimum number of ratings in common before it displays?
Alternatively I think a way to do a kind of Bayesian thing would be to add C artificial identical 2.5 ratings when calculating the correlation coefficient, where C is a confidence value.

As a side note if there's division by zero in the correlation coefficient the result should be 0 or nothing at all, though I assume you're already handling that fine.