If only there were some way to produce some kind of "interval" of scores where you were confident that the actual score sat, and then had some way of comparing

If only there were some way to produce some kind of "interval" of scores where you were confident that the actual score sat, and then had some way of comparing these intervals between the different models...