I think (2) is in large part what people want. (1) was proposed and rejected in [poll] Balancing the new AI score estimator
(3) is important in a weak score estimator because it helps you mentally calibrate which parts of the estimate are wrong.
2 Likes