Love the dive into data you are doing but I was wondering if you are saying that raw point totals need to be valued more or less based on R-ON/R-OFF data?
My general view is that points are points whether at ES or on the PP and only under obvious factors do you move a player up or down to the level of a player with clearly superior or inferior offensive production. When you are talking about generational talents like Ovechkin (at his peak especially) and Crosby (moreso throughout his career), they get their points (or just goals in OV's case) regardless and one does not need to qualify them. They have proven to be able to produce either at ES or on the PP with any level of linemate.
That being said, factors like quality of linemates, offensive support from linemates/rest of the team, and defensive play can certainly give an edge to a player over another player with similar offensive production.
Generally speaking, Crosby throughout his career had objectively less talented linemates and was responsible, if not effective, in his own zone. He also carried his team's offensive load a bit more as Malkin, clearly the best support player between the two missed a lot of time.
I am not a fan of the "ES points are better" mantra. You seemingly devalue Crosby's 06/07 season due to a lower R-ON/R-OFF percentage but yet he lead his team to an historic points jump. I think the R-ON/R-OFF percentage doesn't consider that his ES time may have been (1) lowered due to his high PP time, and effectiveness on it and (2) playing more of his ES time in a less offensive mindset as his team was ahead after scoring on the PP.
And the R-ON/R-OFF percentage is essentially 60% to 70% affected by something that is almost 100% completely out of a player's control, how the rest of his team performs and what lines the coach makes up. And of course the other 30% to 40% is affected by who is on the ice with the player i.e. a loaded #1 line will do better than a balanced Top 6.
All that being said, your chart does reflect a reasonable positive for Crosby and maybe should be given more value given the quality of ES linemates over their careers.
Good question. No, we definitely shouldn't discard the traditional stats (goals and assists), but I think this helps give added context.
It isn't a case of whether a player is producing at even-strength or the powerplay. A point is a point, after all. But it's about giving some context to how the numbers were achieved. The 2015 season is a good example. Crosby and Ovechkin had similar production (brace yourself for the incoming rant about secondary assists). There wasn't even a noticeable difference between ES vs PP scoring. But Crosby's team was much more effective at outscoring their opponents when he was on the ice (compared to when he wasn't), both in the absolute sense, and in the relative sense (ie compared to Ovechkin). Therefore I think it's appropriate, in that case, to conclude that Crosby contributed more to his team, despite similar scoring stats. About three-quarters of the game is played at even strength, and simply looking at goals and assists doesn't tell whether the offense was produced "responsibly" (ie while minimizing the opponents' chances).
It's true that these numbers aren't solely the result of an individual player. But, as I said in another recent thread, this is as close to a controlled experiment as we can get. Ideally, we can make a team play 1,000 games against a variety of opponents without Crosby (or whoever). Then we can make them play 1,000 games against those same opponents, with Crosby. That would tell us how impactful the player really is. Obviously, that isn't close to being feasible. But looking at how the team performs when a player is and isn't on the ice, despite it being a much smaller sample size, is the closest we can get.
Over the course of a single season, it's obvious that there's luck. (FYI - there's luck in basic scoring stats too; if someone dismisses these types of stats because there's randomness, they'd have to dismiss goals and assists too, and then we're left with nothing). That's why we shouldn't place undue emphasis on a single data point. But that's why presenting full data over the course of their career helps (as it illustrates trends).
I think there's some value in looking at PDO, which is probably indicative of how lucky a player was. (It shows that from 2008 through today - December 30th, 2021 - Crosby and Ovechkin have precisely the same PDO, implying that neither has been particularly lucky or unlucky relative to the other). There are, of course, variations from season to season (Ovechkin appears to have had really bad luck in 2014, which probably explains his terrible R-ON/OFF result for that season; it also shows that his results in 2022 are unsustainable). But, again, looking at trends over time are more useful than looking at a single data point.
A lot of times the players with the best R-ON/OFF ratios are goons and bottom-pair defensemen. Bob Boughner (a solid defensive blueliner, but far from an all-time great) has some ridiculously good R-ON/OFF numbers. That's because players like that generally get easier matchups. It would likely be misleading to compare, say, Peter Worrell to Pavel Bure (because the competition they faced would have been very different). But comparing Bure to Jagr, Sakic, Selanne, etc., is probably appropriate.
I've looked at the R-ON/OFF numbers quite a bit over the past several years. The numbers generally make sense (once you take the disclaimers about matchups into account). See the quote from Bill James above. Hockey stats have been fixated on scoring for far too long (plus/minus is poorly constructed, and is half influenced by offense anyway), so it's interesting to try to make progress in trying to quantify other aspects.
