Secondly, I'm not sure the +GD/-GD split is conceptually the right comparison. Those choices are always tough because every time you bin data you run the risk of obscuring meaning, but if you don't group data at all then you tend to produce a single number that is difficult to interpret and obscures the underlying detail. Personally, I tend to care about how a player does against the very bad teams (say, bottom 20% or so), and the elite teams (top 20-25%), because the first is meaningless in that it doesn't really impact winning, while the latter is meaningful because those are the teams you are likely going to run into at the business end of the playoffs. In the middle there's lots of room for variance in terms of style of play, dedication to matching lines, etc. Perhaps most importantly, though, a team's record is only an approximation of its talent level, and it's very easy for a team's GD to swing by 10-20 goals just based on injuries or backup goaltending or an extended streak of running hot/cold by the percentages. As a result, to me it makes more sense to set lower and upper bounds where you have a higher degree of confidence that anybody who falls into those groups definitely deserves to be there (although of course I understand if others disagree).
And finally, are you making any adjustments between eras at all in your analysis? Just based on your Lafleur/Ovechkin comparison in terms of team goals from another thread I'm assuming you aren't, and I wonder if that might be helping/hurting particular players? Bobby Hull seems to me to be a player that is going to be particularly disadvantaged by a +GD/-GD analysis, given that he spent his mid-'60s goalscoring peak in a period where there tended to be four good teams and two awful teams. As a comparison, in the 1950s ('50-'59), 14 out of 40 playoff teams had a negative goal differential (35%). However, from 1960-67 just 6 out of 32 did (19%).
You raise good questions, and I'm going to try to answer some of them, but I'm still lacking some data based on how I started my research and started building my spreadsheets. I started with team splits, and I have them for most all of the top 100 players prior to that 1988 cutoff in hockey-reference. That was before the NHL website added the ability to do splits, so it was a lot of year over year transcribing of game logs. Then I moved deeper on specific players (like Hull), and left others (like Harvey) alone, so I have deep dives in some places. Then I went back and did the +GD/-GD stuff, but it was all manual for a bit. I've managed to create enough formulas that the splits come quickly now, but I still have to go back for a whole lot of players. Some players I looked at playoffs, other players not so much. Also, because it was all manually entered, there are undoubtedly a few bad entries where I mis-typed something.
I want to concentrate in particular on these two points, and endeavor to provide some answers to your points. I'll start with the second part, and say that I have adjusted in some places, but this particular analysis doesn't use it. The gap in PPG vs +GD/-GD teams is larger in higher-scoring eras, but I've moved to using team goals/game rather than league goals/game, because that's what really matters in looking at player point scoring. [That doesn't really answer your question, but consider this in the Hull/Ovechkin comparison - Chicago scored 3381 goals in 1082 games during Hull's career, 3.125 per game, while Washington has scored as I stated, 3367 goals in 1114 games, 3.022 per game. Do we need to adjust for the fact that Hull's teams were higher-scoring than Ovechkin's teams, or just accept that their eras are similar scoring. With the exception of some of the mentioned players, most everybody else in this discussion falls in that 2.9 to 3.2 goals per game window.]
With regards to Hull, if you breakdown his +GD/-GD splits from 59-60 through 66-67, he actually played 238 games against +GD teams, and 296 games against -GD teams (though a large portion of that gap is driven by the 66-67 season - it was 225 to 243 prior to that year, and since only Chicago and Montreal were +GD that year, the gap grew by 40 that year). Also with Hull playoffs, I have him playing 12 playoff series prior to expansion, 5 against Montreal (20 points in 28 games), 5 against Detroit (41 points in 31 games), and 2 against Toronto (14 points in 12 games).
To your first question, I started with just +GD/-GD, and I got into a Ted Kennedy kick that lead me to try and categorize teams based on if they were above or below 0.5 GD/G on both sides, essentially your quartering suggestion. So in a 70 game season, any team that was +35 or higher got placed into the highest bracket, between 0 and 35 the next one, and so forth. For clarity of labels, let's call those four brackets Great, Good, Poor, Abject. I have a bunch of numbers from that, though it is based on player age, rather than a full overview of years. I'm struggling with how I want to display these numbers (it's a full page summary sheet in a large spreadsheet), but we'll start with this table of wingers (and a center) [I also have another list of 10 centers (and a winger) that covers the same time period, but we'll stick with the Hull/Ovechkin comparison for now]:
| | | | | Team | Team | |
| | Games | Points | PPG | Goals For | Games | Tm GF/G |
Maurice Richard | 46-47/54-55 | 562 | 539 | 0.959 | 1606 | 600 | 2.677 |
Gordie Howe | 46-47/54-55 | 572 | 581 | 1.016 | 1869 | 600 | 3.115 |
Ted Lindsay | 46-47/54-55 | 564 | 524 | 0.929 | 1869 | 600 | 3.115 |
Bill Mosienko | 46-47/54-55 | 557 | 352 | 0.632 | 1556 | 600 | 2.593 |
Alex Ovechkin | 11-12/18-19 | 609 | 597 | 0.980 | 1865 | 622 | 2.998 |
Sidney Crosby | 11-12/18-19 | 530 | 644 | 1.215 | 1729 | 550 | 3.144 |
Bobby Hull | 64-65/71-72 | 554 | 686 | 1.238 | 2003 | 592 | 3.383 |
[TBODY]
[/TBODY]
I started with trying to cover this 46/47-54/55 era, and matched the more modern players by age after that, so that is Maurice Richard's age 25 to 33 seasons, Howe's 18 to 26, Lindsay's 21 to 29, Mosienko's 25 to 33, Ovechkin's 26 to 33, Crosby's 24 to 31 and Hull's 26 to 33. From there, we have two tables, one covering the split in +GD teams, the other the -GD teams. The names are missing from these tables, but each line corresponds to the previous one.
Great | | | | Team | Team | | | Good | | | | Team | Team | |
% of Games | Games | Points | PPG | Goals For | Games | Tm GF/G | | % of Games | Games | Points | PPG | Goals For | Games | Tm GF/G |
0.242 | 136 | 105 | 0.772 | 309 | 146 | 2.116 | | 0.132 | 74 | 68 | 0.919 | 193 | 80 | 2.413 |
0.131 | 75 | 55 | 0.733 | 191 | 78 | 2.449 | | 0.219 | 125 | 100 | 0.800 | 343 | 134 | 2.560 |
0.129 | 73 | 55 | 0.753 | 191 | 78 | 2.449 | | 0.222 | 125 | 93 | 0.744 | 343 | 134 | 2.560 |
0.300 | 167 | 93 | 0.557 | 423 | 186 | 2.274 | | 0.233 | 130 | 82 | 0.631 | 305 | 134 | 2.276 |
0.110 | 67 | 73 | 1.090 | 207 | 69 | 3.000 | | 0.429 | 261 | 245 | 0.939 | 747 | 268 | 2.787 |
0.104 | 55 | 57 | 1.036 | 178 | 58 | 3.069 | | 0.453 | 240 | 294 | 1.225 | 761 | 251 | 3.032 |
0.265 | 147 | 135 | 0.918 | 432 | 158 | 2.734 | | 0.213 | 118 | 142 | 1.203 | 384 | 124 | 3.097 |
[TBODY]
[/TBODY]
Abject | | | | Team | Team | | | Poor | | | | Team | Team | |
% of Games | Games | Points | PPG | Goals For | Games | Tm GF/G | | % of Games | Games | Points | PPG | Goals For | Games | Tm GF/G |
0.244 | 137 | 170 | 1.241 | 529 | 146 | 3.623 | | 0.383 | 215 | 196 | 0.912 | 575 | 228 | 2.522 |
0.238 | 136 | 163 | 1.199 | 584 | 146 | 4.000 | | 0.413 | 236 | 263 | 1.114 | 751 | 242 | 3.103 |
0.238 | 134 | 155 | 1.157 | 584 | 146 | 4.000 | | 0.411 | 232 | 221 | 0.953 | 751 | 242 | 3.103 |
0.066 | 37 | 25 | 0.676 | 115 | 40 | 2.875 | | 0.400 | 223 | 152 | 0.682 | 713 | 240 | 2.971 |
0.164 | 100 | 116 | 1.160 | 349 | 102 | 3.422 | | 0.297 | 181 | 163 | 0.901 | 562 | 183 | 3.071 |
0.160 | 85 | 124 | 1.459 | 326 | 88 | 3.705 | | 0.283 | 150 | 169 | 1.127 | 464 | 153 | 3.033 |
0.282 | 156 | 235 | 1.506 | 686 | 170 | 4.035 | | 0.240 | 133 | 174 | 1.308 | 501 | 140 | 3.579 |
[TBODY]
[/TBODY]
That's a lot of data to process, and I don't necessarily think it is a great argument to distinguish players, but the reason I went to all this trouble to paste this spreadsheet is to show the percentage of games each player had against each bracket, and the different scoring levels for the team. You'll notice that it is true that older players tended to face more "Great" teams as well as more "Abject" teams, but the modern players actually face more +GD teams, while the older players face more -GD teams. You can also see the limiting effects of a 6 team league, as both Howe and Lindsay never had to face themselves, while Mosienko never got to face himself.
This is getting way too long, and I'm not really satisfied with how I've tried to answer your points, but I'm going to post it anyway rather than spend any more time on it.