The Geek's Guide To Karma
This is some new-age happy-clappy hippy trip right?
Wrong.
What karma isn't
Our karma rating is not a ratio of good to bad - 5/10 is not 50% (and more importantly 1/1 is not 100%). That might work for Amazon but it doesn't work for us.
It is not a mystical score based on divination, intervention of spirits or tea-leaves.
It is not a score made up by one or two 'experts'.
What karma is
Our karma rating is based on statistical confidence scoring.
The karma rating for an act will go up the more positive things that people say about it and down the more negative things that people say about it.
We take into account that lots of people saying the same thing is much more meaningful than just one or two. Nothing is "rig-proof" but we've done our best.
What karma is (the maths stuff)
Our karma rating is the lower bound of the Wilson interval computed for the ratio of positive reviews to the total of combined positive and negative reviews as contained in tweets mentioning Fringe acts. We've calculated this with a statistical power of 0.05 or, in other words, a 95% confidence level.
Come again?
The Wilson interval is a statistical tool used to determine a lower and upper bound for the expected value of a series of binomial samples. (A binomial sample is a "yes/no" or "heads/tails" sample).
In lay terms this formula computes two numbers that we expect the average to fall between - an optimistic bound and a pessimistic bound. The more samples we have the closer the bounds get and the more confident we are of our average.
For example, I have a coin that I think might be weighted. I flip the coin ten times and it lands heads seven of those times, tails three times - our interval says that the chance of flipping a heads on that coin is somewhere between 40% and 90%.
