hehe, here's a thought.
modified score formula, idea: re-weight each person's vote by giving less weight to groups who vote more often than minority groups.
to completely automate this, a first iteration could go something like this. group voters into nintendo/xbox/ps/sega whatever "supporters". count voters in each category (they can overlap). reweight votes by say a log function. example:
1000 nintendo supporters, 100 ps supporters, 50 xbox supporters.
average score for gran taurismo:
nintendo supporters ratings: 5.0
ps supporters ratings: 9.0
xbox supporters ratings: 7.0
as it stands, the average score would be 5000+900+350/1150 = 5.44
if you re-weight (or "normalize") them:
5.0 * log 1000 + 9.0 * log 100 + 7.0 * log 50 / (log 1000 + log 100 + log 50) = 6.7
taking instead of base 10 take natural log.
which, is more respectable.
of course, if given a good nintendo game, the score would go down much faster if PS fans decides to hate vote.
the hard part would be the "categorization" part. but it's still pretty easy for a cursory implementation.
so, not that hard to implement, and the results would be very, very interesting. you can even derive a "bias" measure(ment) using the difference above. the greater the difference between re-weighted score and raw score, the greater tha bias.
the Wii is an epidemic.
