To me it seemed that it's inversely proportional to the time (lap time, etc.). But despite there being a good amount of data to work on, I haven't since it would be so dependent on track and all the other variables.
It was based on HP and weight with some other factors, maybe Torque ? Some kind of aero factor? I haven’t seen an exact algorithm but an empirical formula is =759.22*(HP/kg) ^0.3416.
GT5 was similar but definitely had an aero part as you could see the PP change when adjusting downforce.