First principles:
1 no detriment
2 student performance not that of cohort or socio-ec circs should be relevant
1/n
4 if estimation errors, assume underestimation is more damaging to prospects
2/
1 whatever modelling produces, if lower than prelims, use prelim mark
2 model exam mark on basis of
a) student's perf across all subjects last 2 years
b) perf in same or similar subjects last 2 years
c) teacher estimated mark
d) available coursework marks
3/
As an aside, I know PISA scores aren't perfect but it's clear pre-covid approach problematic and entrenching inequality
4/
5/
This is a very (tho not uniquely) British problem but layering that onto modelling assumptions that entrench inequality across classes should make one weep
6/6