It's worth noting that when you're doing multiple significance tests, the threshold for significance goes up dramatically. Naively, if you're doing 20 tests, yo

It's worth noting that when you're doing multiple significance tests, the threshold for significance goes up dramatically. Naively, if you're doing 20 tests, you should expect one to be statistically significant, even if the null hypothesis is true. That 3 were statistically significant from a sample of 25 is, itself, not statistically significant and any attempt to explain them defaults back to just-so stories. An interesting test to run would be to see whether the same outliers held when new movies are added to the page which shift the position of the existing movies.