Variance-to-Mean Ratio
The ratio of variance (standard deviation squared) to the mean. When this ratio equals 1, data fits a Poisson distribution; when greater than 1, the data is overdispersed.
Like checking if a coin is fair: if heads and tails aren't roughly 50/50, you need a different model than "fair coin."
Why it matters
It determines which probability model to use for a given stat — Poisson works when the ratio is near 1, but overdispersed stats need a different approach for accurate probabilities.
How DMP uses this
DMP checks variance-to-mean ratios to decide whether Poisson or an alternative distribution is appropriate for modeling a player's stat, especially for low-count props like blocks and steals.
Common mistake
Blindly applying Poisson to every counting stat without checking if the variance actually matches the mean.