Outliers in red. I wonder if I should delete extreme outliers from my crowd estimate? Gut answer: Yes, if the prediction is stale (more than say 10 days old). NOTE: I’m not ignoring outliers now, but may ignore them in next version of this post, after watching Homeland. The outliers marked here are 1 standard deviation away from the median. I may play with this and go for 1.5 SDs or 2 SDs, depending on my mood.