>The algorithm is based on the assumption that performance of successive school years at a given school is roughly consistent, not only as an average but as a d

>The algorithm is based on the assumption that performance of successive school years at a given school is roughly consistent, not only as an average but as a distribution. And that assumption simply doesn't hold. // It seems likely to hold pretty consistently. What's the variation?