Rating formula? Help apreciated!

[alex]

I need a more advanced formula than just an average for calculating
items rating.

I have:

raitng value is on scale 1 to 10.

s - sum of all ratings for an item
n - number of rates (votes)

scenario 1)
s=12
n=2
12/2 = 6

scenario 2)
s=24
n=4
24/4 = 6

same average value, however what i would like to do is take into
account number of votes for an item. item in scenario 2 should have
higher rating that item 1, because it is more popular (viewed and rated
more often).

what kind of formula whould do this?

thanks!

 

[Guest]

Hey Alex,
I think there's two main approaches:
1 - keep it as two separate values: Popularity and Rating. This is what
CodeProject and MSDN both do. I think this makes sense because the two items
tell very different things. One article with a 1000 0s should be rated lower
than another article with 10 5s.

2 - If you do need to combine them into a single value, then you need to
scale popularity. For example, you may say:
- Every 1000 hits = +1 Rating point. (i.e. s/n + n/k where k is your scalar).
- make it non-linear = (s + n/k)/n.

 

[alex]

Thanks Tim.

This approach althoug makes total sense will add another variable in
the formula.
Why 1000? Why not 10? How do I decide?

I just came across this:

"ShutterPoint has adopted the individual numeric rating normalization
method for calculation of the total rating for each photo. The method
is based on the true Bayesian estimate formula, commonly used in
statistics. The formula is:

Total rating (TR) = (V/(V+M)) * R + (M/(V+M)) * C

where:
R = mathematical average for the photo (mean);
V = number of received ratings for the photo;
M = minimum number of ratings required for calculation (1);
C = median of the rating scale (5.50)

This method allows photos that receive higher number of individual
numeric ratings to score a more representative total rating. Please
note that individual numeric ratings are not shown."

Looks liek there is a standart method in statistics for doing this type
of things. Will test both methods and see which results makes more
sence for my scenario.

Regards,
Alex