I mentioned two methods of getting a baseline rate for top end scorers, fixed tiers (e.g. an average of the 7th-24th scorers each year) and proportional tiers (e.g. using the 7th-12th scorers in O6 and the 31st-60th scorers today), and then suggested combining the two.
Another possibility would be to use an exponent < 1 (like a square root) of the number of teams and using this to determine which tier to use. Here is an example:
N = number of teams in league
first player in tier = [(6N)^0.5]+1
last player in tier = 3*[(6N)^0.5]
In the O6, the first player is #7 and last player is #18, so would take the 7th-18th scorers and average them.
Today, the first player is #14.4 and last player #28.4, so if rounded would take the 14th-28th scorers and average them.
Since most believe the number of teams is less of an influence than the total talent pool, it might be better to use an exponent < 0.5. If you used an exponent of 0.25:
First player = [(216N)^(0.25)]+1
Last player = 3*[(216N)^(0.25)]
This still suggests the 7th-18th scorers in O6, but today would suggest the 10th-27th scorers. Of course, there are endless variations on this method.
The simple fixed and proportional tiers are mostly in agreement about adjusted points of near-top scorers for most eras since WWII:
O6: sideways to down, bottoming out in mid-late 60's
Mid-70s to Mid-80s: steadily down, bottoming in mid-late '80s
Glory Years: significant increase in late 80's and early 90's
2000's: basically sideways
The two methods are really in conflict during two periods.
The first is the O6 expansion, when the fixed tiers show a mammoth increase while the proportional tiers show a small but significant increase. This is no surprise given that the number of teams immediately doubled with further expansion soon after. The fact that even the proportional tiers (e.g. #7-12 in O6 vs. #13-24 after) is clear evidence that top scorers had a much easier time of it after expansion, which is not close to being addressed by present methods of adjusting points.
The other is during the decrease in league scoring from the early 90's until about 2000. Proportional tiers mostly peak in '93 at levels at or below the previous few decades and drift down to sideways from there. Fixed tiers hit post-WWII peaks in '93 and '96, decrease and then level off from there.
Finally, there is another method that has been used in some capacity before which could be quite useful and perhaps most fair. Using individual players as the constant, with the variable being the season. By looking at changes in PPG from season to season of a large group of top players, we may have further insight into how adjusted statistics can be further adjusted. For example, I looked at 44 of the best players from '67 to '68 and by various metrics calculated the increase in adjusted PPG to be ~13-17%.
It's best to select a large sample of top players (at least 40-50 in each two consecutive seasons compared) to help offset the many factors that can cause changes in adjusted PPG (age, injury, team, linemates, luck, etc.). Also, I looked at medians, such as the middle third or middle half of players in terms of % change in PPG.
So for '67 to '68:
Player --- % change in adjusted PPG
=======================
McKenzie 84.0%
Cournoyer 74.3%
Provost 73.5%
Beliveau 72.5%
Bathgate 66.0%
Duff 66.0%
Hodge 60.0%
Gilbert 58.6%
Ingarfield 56.8%
TremblayG 49.6%
Hadfield 48.2%
Esposito 38.0%
Prentice 30.2%
Delvecchio 29.7%
Nevin 28.2%
Bucyk 26.6%
Howe 26.3%
Armstrong 25.0%
MahovlichF 21.3%
Goldsworthy 17.4%
BackstromR 16.2%
Marshall 14.9%
Goyette 9.6%
WilliamsTo 9.4%
Wharram 8.2%
Pulford 8.2%
Orr 7.9%
Ullman 5.7%
Pappin 4.1%
Ellis 3.5%
Westfall 3.1%
Rousseau 2.2%
Nesterenko 0.5%
Oliver -0.4%
MartinP -0.4%
Keon -1.7%
HullBo -6.3%
Mikita -6.4%
Mohns -10.9%
HendersonP -13.4%
HullD -20.1%
Pilote -27.1%
RichardH -33.9%
Larose -44.0%
So one method was to use the median half (since their were 44 players, this would be the middle 22 or players #12-33) and either sum and average their % change in adjusted PPG (this was +15.3%) or sum their games and points and then calculate the difference in adjusted PPG (this was +15.1%).
Alternatively, you could use simple PPG instead of adjusted PPG, then factor out the change in league scoring.
Perhaps someone with a complete database and the appropriate statistical/computer knowledge can further one of these ideas.