Players who improve in playoffs vs get worse in playoffs stats.

[Bad Gatorade]

So, I don't actually think PER is a horrible measure to use, even though I don't think it's a great stat on the whole - PER being exempt from team performance means that for an individual player in an individual season, PER does a decent job of comparing productivity between the regular season and the postseason. Yes, PER is bad at assessing defence, and it's bad at comparing players in different roles, but because the same player is likely to have the same role between the regular and postseason, I don't think it's all bad to use in this case. Certainly not all encompassing, but it's a nice starter for this purpose, especially once we make allowances for defence.

In other words, we're comparing Player A in Role B to... Player A in Role B, so it's better than most conventional uses of PER, IMO, where PER itself is generally inferior to other, more encompassing stats.

Interesting to see how this would fare with something like BPM too, however, although I do think BPM maintains a level of bias with regard to teammate defensive performance (e.g. when Draymond suddenly decides to become the world's best defender in the playoffs again each season, his entire team benefits due to the DRTG adjustment).

But yes...

There is one change I'd make, and it's to calculate the change in PER, rather than the differences in overall PER. For example, to make a very crudely demonstrated point, Giannis' per increases from 24.9 to 25.6, but he also became a 30 PER guy in 2019. The last 5 years constitute 45% of his career minutes, but 75% of his career playoff minutes. In the last 5 years, Giannis has a 30.6 regular season PER and a 27.3 playoff PER. A more correct way would be to do something like this -

2019: 514 * (26.5 - 30.9) +
2020: 277 * (31.3 - 31.9) +
...

and so on. Basically, summing Minutes * (Playoff PER - Regular PER) and then dividing by total playoff Minutes.

Now, this yields a change of -3.3 (identical to the "crude" calculation above) and is infinitely more onerous than the method outlined in the OP, but it is also more informative, IMO.

FWIW, I think most playoff drop offs are overstated - taking Giannis as an example, dropping in PER by 3.3 points doesn't necessarily mean that he's a "true" 27.3 PER guy rather than a 30.6 PER guy. A stat like PER both doesn't take into account opposition (ergo, drop offs are likely to be overstated on account of better opponents) and it doesn't take into account that a more "successful" defence of a player might result in their PER dropping, but the same may not be said for his teammates - for example, Harden's postseason TS% drops are well documented, but IIRC, his plus-minus actually improves in the postseason. His impact is still there, because he's still James freaking Harden, and the man forces a defence to commit to him, which means that he's still generating shots for himself and his teammates. A stat like, say, BPM (which calibrates shooting efficiency + frequency to impact) may not successfully account for this. How well PER accounts for this, I don't know, but I wouldn't be shocked if some of the effect was still there.

I'm actually more interested in playoff increases than decreases - someone like LeBron post-2013 clearly ramps up his effort in the postseason, and I'm more inclined to believe something like that is "real" than players miraculously getting worse over a large sample size.