But, how much of the Oilers bigger PP could be due to McDavid getting better at it, the Oilers PP percentage was abnormal that year, not a league wide phenomenom that need some explanation. What would this be telling us here ?
It seems to a bit exclude the fact teams when the elite player is scoring that much more than a good first liner replacement their team score more, their percentage is one of an inflated by them denominator, i.e. they are hurt by how much they help their team scoring.
Take the Penguins of Lemieux for example, if we take the 91-92 and 92-93 season together we have a rare case of a superstar playing a lot but missing a lot of games at the same time to have some idea of what a superstar do to a team.
The Penguins with Lemieux scored 4.64 goals a game, without Lemieux they score 3.38 goal a game (1.26 goals or 37% team offense boost).
If Lemieux has 2.2 pts per game he has 47% of his team points, but would he had added no actual real offense and scored 1.7 pts a game he would have look better scoring over 50% of his team points. Would he have been better at hockey and added more goals but in ways that do not lead to him getting points (by playing better defense, screening the goaltender and what not) he would have looked worse.
Instead of thinking of it as PPG, think of it as offensive share. Let's take that 91-92 Penguins team, and break down their 80 games into 4 chunks of 20.
| | Games | GF | GF/G | GA | GA/G |
10/4/91 | 11/18/91 | 20 | 79 | 3.95 | 80 | 4 |
11/20/91 | 1/2/92 | 20 | 107 | 5.35 | 69 | 3.45 |
1/4/92 | 2/22/92 | 20 | 69 | 3.45 | 80 | 4 |
2/25/92 | 4/16/92 | 20 | 88 | 4.4 | 79 | 3.95 |
| Total | 80 | 343 | 4.29 | 308 | 3.85 |
For the season, they averaged 4.29 goals per game, but were widely variable in each 20 game segment. Now let's look at their 2 highest scorers, Lemieux and Stevens and their point totals in each of those segments. Stevens played all the games, while Lemieux missed 16 games. Here are the numbers:
| | Games | Goals | Assists | Points | PPG | GF | P% | | GF/G |
10/4/91 | 11/18/91 | 17 | 13 | 16 | 29 | | 72 | 0.403 | L P | 4.24 |
10/4/91 | 11/18/91 | 17 | 13 | 16 | 29 | 1.71 | 79 | 0.367 | L A | |
10/4/91 | 11/18/91 | 20 | 15 | 18 | 33 | 1.65 | 79 | 0.418 | S | 3.95 |
11/20/91 | 1/2/92 | 20 | 15 | 30 | 45 | 2.25 | 107 | 0.421 | L A | 5.35 |
11/20/91 | 1/2/92 | 20 | 15 | 21 | 36 | 1.80 | 107 | 0.336 | S | 5.35 |
1/4/92 | 2/22/92 | 10 | 1 | 13 | 14 | | 33 | 0.424 | L P | 3.30 |
1/4/92 | 2/22/92 | 10 | 1 | 13 | 14 | 1.40 | 69 | 0.203 | L A | |
1/4/92 | 2/22/92 | 20 | 9 | 14 | 23 | 1.15 | 69 | 0.333 | S | 3.45 |
2/25/92 | 4/16/92 | 17 | 15 | 28 | 43 | | 77 | 0.558 | L P | 4.53 |
2/25/92 | 4/16/92 | 17 | 15 | 28 | 43 | 2.53 | 88 | 0.489 | L A | |
2/25/92 | 4/16/92 | 20 | 15 | 16 | 31 | 1.55 | 88 | 0.352 | S | 4.40 |
| | 64 | 44 | 87 | 131 | | 289 | 0.453 | L P | 4.52 |
| | 64 | 44 | 87 | 131 | 2.05 | 343 | 0.382 | L A | |
| | 80 | 54 | 69 | 123 | 1.54 | 343 | 0.359 | S | 4.29 |
I've divided Lemieux chunks into 2 lines, L P and L A, standing for Lemieux Precise and Lemieux Actual. The Precise lines have removed the goals for totals from the game Lemieux missed in each chunk, while the Actual lines are the totals in full. As Stevens missed no games, his Precise and Actual totals are the same. The first segment I want to point to is the 2nd and 3rd ones - Pittsburgh went from averaging 5.35 goals/game to 3.45 goals/game, Lemieux's PPG dropped by 2.25 to 1.40 while Stevens PPG dropped from 1.8 to 1.15, yet in terms of P%, for both players, it is virtually identical at 42% and 33%. That's why I'm saying scoring levels don't matter. A player's offensive share tends to remain his offensive share, regardless of how few or how many goals his team scores.
Another segment of note is to compare Lemieux's first 3 precise chunks to his final one. He's cruising along at that 42% range, and suddenly he spikes to 56%. I'm guessing if we had his advanced stats for on-ice shooting percentage and IPP for both even strength and power play, that segment would be much higher than the other 3. Without them, it just is a big spike in his counting stats that we have to acknowledge.
The contention that I think you're having is that you end up with all this noise surrounding point totals, and you're trying to derive some sort of signal. Spikes are essentially just random, and as likely to be variance as they are player skill. I will make an admission though, you are somewhat on the right track when you say higher scoring leads to a lower P% - it seems that on teams way above league average (30% or more), it is less likely that top scorers hit that 45-50% P% threshold, it tends to be more in that 35-40% range. Again though, we're talking about a sample of maybe 75 player-seasons since WW2. It might be as simple as in a 5-2 game in the 3rd period, Gretzky got a regular shift while Esposito did not. Picking out that trend from all the noise is imputing way too much weight to a tiny sample.
[Also, because I'm sort of theorycrafting on the fly, I can admit that I'm surprised how neatly the 91-92 season fit into my hypothesis. I basically wanted to compare Stevens and Lemieux, hit on using 20 game chunks, worked out the stats, and came up with this presentation. I know my hypothesis is theoretically sound, but being able to use almost any sample to illustrate my thesis is really helpful. Finally, this back and forth is also probably not the best fit for this thread, given we're arguing more about statistical interpretations rather than hot takes.]
[[Also also, refreshing this thread and the 300 points Gretzky annoys me in a way no other hot take does, because it is just mathematically wrong. We have 100 years of data, including 80s Gretzky, that shows players top out at 50% P%. A team will not score 600 goals in an 82 game season.]]