Average vs. Median

[flsaders85]

When tackling poisson props, which is better and why? I've run some analysis on players scoring in NBA and it looks like average is a better predictor. And typically, the median will give you a lower number than the average and the actual result. But, how does this translate into analyzing poisson props?

Right now, my thinking is to use average and normalize it according to opponent and the spread/total. Thoughts?

[TomG]

using the median makes sense if the distribution has outliers with leverage that skews the mean. for most props this isn't the case just use the mean

[flsaders85]

I like the idea of throwing out games where their minutes played is 1 or 1.5 standard deviations away from their average. This may skew the average. Or maybe just use median in this instance.

[roasthawg]

Mean has outperformed the median in any tests that I've done... haven't spent a bunch of time on it however.