By default, the View Results section sorts thoughts using a weighted average for their ratings.
Where the Simple Average would be the total number of stars a thought received divided by the number of times it was rated, the Bayesian Average is weighted to reflect the idea that more ratings equals more confidence in the result. Put simply, it ensures that the top-rated thought was rated by more than a handful of people and thoughts that have a small number of ratings don't appear above ideas with wider-reaching support.
How do you calculate Bayesian Average?
When each new thought comes in we assign it five 3-star ratings (Only in the calculation. These ratings don’t actually appear on the ratings graph). We go with 3-stars because that is approximately the average star rating for any single thought across all the exchanges in our database (thousands of exchanges and millions of thoughts). Essentially this technique assumes that every new thought is of average importance and waits for participants to provide evidence otherwise. By assigning each thought 5 default ratings, the effect is most prominent on thoughts with few ratings, but becomes negligible on thoughts that have been rated by a lot of people.