Nice insight. Fully agree.Quote:
I'm disappointed that Winston doesn't include the statistical error, since it is generally known that the biggest problem with the adjusted +/- stat is that it is extremely "noisy". For a statistician this is a serious oversight.
With adjusted +/-, even with 2 year sample size you routinely see errors of 50, 100% or more, which means that the frequency of bogus results (called "outliers") increases rapidly with decreasing sample size. Half a season sample size is too small to have any confidence in the calculation.
Without getting into specifics, you basically need enough data with enough variation in lineups for & against to separate the effect of the player vs. teammates vs. opposition lineups. There just aren't enough minutes in enough scenarios to do this with any level of confidence with only half season worth of data.