Predictions clearly are hugely inflated, in general, as only around 40% of students make or exceed their predictions (although it is probably a bit more than this at the very top end of the spectrum).
PS writing companies make good money - and there were a lot of them around, and prior to the GMC's recommendation that they were to be less relied upon, would tailor what they wrote to the med school of choice.
Some GMC backed research concluded
-Overall, research evidence suggests that autobiographical submissions (PS) have limited validity in relation to medical student selection.
-Research evidence suggests that autobiographical submissions are more susceptible to contamination and input from third parties than many other common selection methods, which disadvantages applicants from lower socio-economic groups who are less likely to have the appropriate networks and resources to provide this.
Hence the recommendation they were given less weighting.
The same research concluded, about admissions tests in general
-Widening access is an aim of aptitude tests, and there is preliminary evidence that it has a positive effect.
-There appears to be a small impact of preparation on performance in UKCAT, but is less than the standard error of measurement.
(I think this may be a bit out of date now, as the preparation tools have expanded and adapted, but bursaries have moved forward in line with this).
Educational performance, ie GCSEs and A levels is far more affected by contextual factors, and those with WP flags will perform comparatively worse in school/exam testing than they will in the UCAT. As the research concluded
-there is clear evidence (in a UK context) that use of academic records introduces a significant socio-economic class bias.
So evidence suggests practice, preparation and support have a far bigger impact on exam results than on UCAT/admissions tests, so using the latter should make it a more level playing field, not a less level one. Med schools continue to be encouraged to try and level it further and many have made massive strides, with, for example, clear contextual points at Birmingham, Bristol, Kings, HYMS, etc etc
So I am not sure there is a fair-to-all way of selecting medical students, and there are winners and losers however it is done, but medical schools, on the whole, try to stick to that for which there is an evidence base and which disadvantages the disadvantaged the least, so UCAT and exam results, with an allowance for those with WP flags.