Well, I have no vested interest in the method of star ratings because I don’t submit to the forums but votes too far removed from the average can be filtered out by just changing the calculation that averages them.
There are equations to do this properly (I programmed some once but I’m damned if I can remember what they’re called and they’re probably OTT here) but a simple way to do it would probably be to take the average (after a certain thresshhold number of votes have been registered) then take the average again between the previous average minus (say) 1.4 and the previous average + 1.4 (tweak as needed*)
Basically this would eliminate values too far removed from the general consensus instead of making them part of the average.
Example - filtering out 1-star bombers…
Say a piece has votes of 5, 5, 5, 4, 4, 4, 1, 1
The average = 3.625, which has been brought down by the 1s
The filtered average = average of values between (3.625+1.4) and (3.625-1.4)
the average of values between 5.025 (5) and 2.225 =
average of 5, 5, 5, 4, 4, 4 = 4.5, which is more like it
And filtering “best friend” voting…
Say a piece has votes of 2, 3, 3, 3, 1, 3, 5, 5, 5
Average = 3.33, which is too high because of the unlikely 5s
Filtered average = average of values between 4.73 and 1.93 =
average of 2, 3, 3, 3, 3 = 2.8, which sounds about right
(it didn’t take notice of the miserable 1, either)
This geeky method might help if someone is thinking of programming solutions, as it’s only the average calculation that needs to be tweaked and not the entire process, and it does tend to filter out people who are likely voting the wrong things for the wrong reasons.
Of course, if the number of asshats outnumber the proper voters then the system won’t work, but then again neither will any other.
- Geeky note: you could make this dynamic based on the first average to give better filtering
" <---- is equally 
?