Trying to get to everyone's comments:
Wow, Bobby Orr the best by a lot on defense. How about that?
That's why I do this - to provide these earth-shattering insights.
I am a bit surprised to not see Lindros-LeClair in there (maybe the minimum games requirement kept them out)
Lindros only played >80% of a season five times. (I knew it wasn't a lot of years, but I would've guessed he had a few more). He didn't meet the PK requirement in three of those seasons, so he's not shown here. (If I drop the PK requirement, his result is +13.1%, which is incredibly strong - well within the top ten).
LeClair, of course, had plenty more healthy seasons. But he very rarely played on the PK.
Kariya without the puck? No way. Fix the glitch (his great defensive pivot might've helped?). To wit, love Rucchin mentioned here! (though his offense was all about getting the puck and giving it to Selanne or Kariya).
I was surprised to see Kariya here too. I figured that the PK requirement would have resulted in his exclusion. It looks like he was deployed on the penalty kill a bit more than I remembered. The method is agnostic to the means - so it doesn't care that Kariya is all offense, as long as he drove his team's goal differential when he was on the ice, that's all the matters.
Out of curiosity, why does data only go up to 2020? Also, didn't someone here find a way to estimate Howe pre 1960 or am I misremembering?
Pre 1960, the data was estimated for Howe alone. From 2021 onwards - I have the data in a different format. (My starting point was overpass's spreadsheet). It would have been a huge undertaking to get everything set up properly for 2021 onwards. I might do that at some point in the future.
How much of this is from the late 70s Penguins being awful?
Is it a case of like 35% GF while the off is 25% GF?
That seems to be the reason. 52.1% on the ice vs 40.9% off the ice. (That number might not sound too bad, but it's actually the 5th worst off-ice GF% of any player here - the bottom five are Stackhouse, Gordie Howe, Harry Howell, Dennis Hextall, and Walt McKechnie. The top five are Doug Mohns, Guy Lapointe, Bill Barber, Serge Savard, and Rick MacLeish).
I'm the most surprised by how low Brian Trottier is. I figured he would be up there with Clarke.
Trottier was only deployed sometimes on the penalty kill, so that eliminated several of his top seasons. (Three of his top five seasons were years where he didn't get a ton of PK ice time). Excluding that requirement, he's at +14.5%, which is an outstanding result (as we'd expect from a player with his reputation).
Not sure how Mccabe, Wesley, Olausson made the list at all. Never thought much of any of those guys.
McCabe was a big surprise. Everything I wrote about Jovanovski also applies to him - all the tools, no toolbox.
I thought Wesley was steady - quiet, unassuming, but always reliable. His best season (by this metric) was when he was paired with Bourque.
Wondering where Craig Ramsay ranked here.
Thought he would have done well in this area.
He got filtered out by the powerplay criteria. Excluding that he'd be at +11.0%, which is a very strong result.
Dallas had pretty good strength through the line-up in the peak years of Modano and Lehtinen, so them standing out so much is a bit of surprise. Best two-way ES pair ever?
They would be way up there. Two of my all-time favourites.
Can you explain how this is calculated? I'm reading the words but in my mind can interpret this many different ways. Are you saying 15% of a player's time should be SH and 15% PP? Or that he should play 15% of his team's overall SH minutes? And is it 15% of both PP + SH, or combined?
I guess what I'm especially curious about is - did this parameter end up excluding any significant seasons for certain players? Example - Gretzky ranked 4th - but were some of his peak years excluded beacause of this SH/PP threshold? Same question for Lemieux and many others.
It means that (for a forward) he had to play at least 15% of his team's PP
and he had to play at least 15% of his team's PK time. For defensemen, it's 20% for each. If a player misses either criteria, the season is discarded outright.
Obviously the 15% and 20% are subjective. I wanted to filter it out so that players had at least some responsibility in both areas (otherwise, at least in my mind, they're not really a complete or two-way player). Sometimes it can eliminate a great season (Forsberg was dominant at ES in 2003 but got hardly any icetime on the PK, so that season got excluded).
The only seasons that were excluded for Gretzky were 1998 and 1999. Lemieux had a bunch of seasons excluded due to the >80% of the schedule rule. After that, he lost out on 1985, 1986 and 2003 (none of which were among his best, based on this metric).
Ovechkin didn't have a single season that qualifies. He had barely any PK responsibilities during his career. Otherwise, his results are what we'd expect - a dominant performer at ES at the start of his career, and then below average the rest of the way. (Keep in mind this doesn't look at powerplay production - he's one of the best all-time in that category). Crosby received more PK ice time than Ovechkin, but still didn't have a single year that qualified. He also had a great start, but didn't fall off to nearly the same extent later in his career.
Also - you mentioned Paul Coffey looked decent but didn't mention where he ranks, and now I'm curious. where did he end up?
He's at +3.9%. Still a good result, but not as strong as we'd expect.
One thing that I'll say in Coffey's defense (and this applies to Gretzky and Lemieux as well) - GF% probably doesn't fully capture their impact. I'd describe them as high-volume players. The extent to which Coffey increases his team's GF% is (ballpark) comparable to Rod Langway's. But Coffey is the more valuable player because although the percentage is similar, the volume of goals being scored when Coffey is on the ice (for and against) is so much higher. In other words - when Coffey's on the ice, a lot of stuff is happening, and the majority of it is good. Langway will give you a similar percentage, but on much lower volume, and that's not as good an outcome.