Yeah I don’t know where this “average error rate” is coming from as I haven’t looked at this guys stuff. But if he’s saying he’s applied his model to previous seasons and ON AVERAGE his model predicts the point total within 10.4 points in either direction, then I’m not sure he should even be putting it out there. I could create a model with one single input; previous year point totals; and it will be more accurate. Assuming that’s an input for him (maybe it isn’t, which could be part of the problem), the rest of his data is actually hurting his accuracy.
|