My guess is that if you were a statistician, you'd adapt your new methodology to yield results similar to the last results of the old methodology until you could measure new performance. So maybe the new methodology just tries to match overall 2020 results.