Originally Posted by Redway (Post 5814753)

Yeah, extreme cases tend to be pulled back towards the average. So if you get a very low score (70-) they'll add a lot of points and if you get a very high score (120/130+), they tend to subtract a lot of points. :/

It's entirely symmetrical for average scores based around 100 but it's a bit uneven before and after that.

Originally Posted by Joseline (Post 5814759)

Yeah, positive and negative skews and all that. Someone should come up with a transformations of the log-normal distribution to model the the CI size for scores above 100.