A "Hall of Stats" Approach to Hockey

Baseball isn't everything.

Well, maybe it is. But bear with me. There are some other sports out there and I really don't spend much time following them. These sports also have Halls of Fame. Obviously, I spend a lot of time thinking about the Baseball Hall of Fame. But lately, I've been thinking about these other, quieter Halls.

Originally, I was thinking it might be neat to make Halls of Stats for other sports from advanced metrics, but I figured I'd save my time and go right to the list people care about—who's outside the Hall and (perhaps) should be in?

I'm starting with hockey. Because hockey's back! (And because hockey is the only other sport I know anything about.)

Picking a Metric

With the Hall of Stats, I started with WAR from Baseball-Reference. For hockey, I'm turning to Hockey-Reference. H-R has a stat called Point Shares that combines the offensive, defensive, and goaltending contributions of a player.

Sounds a bit like offense, defense, and pitching, right?

When I started following hockey in the early 90s, 100-point seasons were expected from the top scorers. We weren't far off from 150-point seasons being attainable (and even Gretzky's 200-point seasons). By the time my hockey fandom faded, even a point-per-game pace was something reserved for only the elite.

Sounds like variable run environments, right?

During that time, we saw huge variations in a goalie's GAA (goals-against average). We went from Grant Fuhr's "eff it, you score four and we'll score eight" approach* to the league average GAA looking like a Koufax ERA. Some was due to changes in scoring environments. Some was simply a team neglecting defense for more offense. How do we really know how good the goalie was?

Sounds like DIPS theory, right?

* Wait—was Grant Fuhr goaltending to the score, like our friend Jack Morris supposedly pitched to it? Perhaps this warrants another study.

Point Shares were eye-opening to me. Like before WAR and similar predecessors, I had never compared hockey players from different positions and eras before. Point Shares takes all of these contributions, context-adjusts them, and gives you one number that represents a player's total value.

Sounds like WAR, right?

The Makeup of the Hall

Going into this article, I knew nothing about the Hockey Hall of Fame. Here are some numbers about the Hall of Famers listed in the Hockey-Reference Play Index (which is free, by the way):

• 217 Hall of Famers with NHL stats (some Hall of Famers played before the NHL existed)
• 207 Hall of Famers with NHL stats who were inducted as players—179 skaters and 29 goaltenders
• 171 Hall of Famers with NHL stats who were inducted as players who appeared in 200+ games (eliminates players who primarily played before the NHL existed)—156 skaters and 25 goaltenders

These numbers are important, because only 13.8% of all Hall of Famers are goaltenders. Point Shares don't seem to work when comparing skaters to goalies. 143 players in history have accumulated 100 point shares, for example. 29.4% are goalies. So, this could mean a couple things:

1. Goalies accumulate more point shares than skaters and/or
2. There are too few goalies in the Hall of Fame

I think there's a little bit of both going on. I mean, there are six positions. Goalies take up about 16.7% of the playing time. So you'd think they'd be at least that high. That said, goalies tend to play then entire game and have a more prominent role in the game than many individual skaters. I wouldn't be surprised if the typical goalie was actually more valuable than the typical skater. I also wouldn't be surprised if quarterbacks were worth more than your typical football player (though I haven't looked into this yet).

For the most part, baseball positions are on a level playing field in terms of WAR. Of the 52 players with 80 WAR, 20 are pitchers (about 38%). Seems reasonable to me. Two exceptions are catchers (who get a boost for their time spent on the field, but not the time lost due to the rigors of the position) and relief pitchers (who just don't play enough to contribute as much WAR as a starter, typically).

The Snubs

Who are the best eligible non-Hall of Famers, according to Point Shares? I'm going to list the skaters with 100+ point shares (there are 18) and the goalies with 120+ point shares (there are six).

The Skaters

1. Phil Housley (170.7 Point Shares): I was absolutely shocked to see Housley isn't in the Hall of Fame. He's in the US Hockey Hall of Fame and the International Hockey Hall of Fame. But not the Hockey Hall of Fame. I suppose he was seen as a weak defender. 56.7% of his point shares were on offense, but he was so good that his 73.9 defensive point shares still rank 22nd all time (he's 41st in offensive point shares and ridiculous 15th overall). Furthermore, it's hard to imagine a defenseman with 1,232 points not making the Hall of Fame. Alas, Housley never won a Stanley Cup or a Norris Trophy. Given Housley's ranking in Point Shares, that'd be like Pete Alexander (who is 15h in WAR) not being in the Baseball Hall of Fame.
2. Brendan Shanahan (147.9): A shocking snub from the last Hall election, Shanahan owns 656 goals, 1,354 points, and a hair under 2,500 penalty minutes. 80.1% of his point share are offensive. He's 35th all time in point shares, 17th all time in point shares of the offensive variety. His defensive point shares are actually incredibly impressive—for a forward (ranking 7th all time among forwards).
3. Pierre Turgeon (134.3): A similar point total to Shanahan, but about 2,000 fewer penalty minutes. I remember Turgeon winning the Lady Byng Trophy, which sounds like the worst thing that could happen to a hockey player. 82.2% of his point shares are on offense. He's 52nd all time in point shares (23rd in offensive). Not surprisingly, he doesn't rank in penalty minutes.
4. Jeremy Roenick (125.8): I haven't followed hockey much over the last few years, but Roenick always struck me as the Curt Schilling of hockey. 78% of his point shares are offensive.
5. Dave Andreychuk (124.6): Another 600-goal player with a very high percentage of his value on offense (79.3%). In fact, his offensive point shares rank 32nd all time.
6. Gary Suter (121.7): Another player I was drawn to as a kid, Suter didn't reach 1,000 points (844), but he did score 200 goals and contribute 50.9% of his point shares on offense. He's 61st all time in defensive point shares and 78th all time.
7. Eric Desjardins (114.5): The first of the defense-first players on this list, Desjardins collected 575 points but also 74 defensive point shares (64.6% of his total and 21st all time).
8. Steve Duchesne (112.7): Similar numbers to Suter, with 752 points and 53.1% of his point shares on offense. Suter is 78th all time in point shares while Duchesne is 97th.
9. Alexander Mogilny (111.5): A whopping 84.6% of Mogilny's contributions came on offense. He scored 473 goals and cleared 1,000 points while only spending 432 minutes in the box. He's 46th all time in offensive point shares.
10. Doug Wilson (110.5): The only player on this list who didn't retire in the 2000s (he retired in 1993). Another defenseman who split his contribution with over 800 points and 51.1% of his point shares on offense.
11. Theoren Fleury (110.3): Like Mogilny, he cleared 1,000 points and had a super high percentage of his contribution on offense (82.7%). Unlike Mogilny, he was a bit of a 5'6" punk, finishing 63rd all time in penalty minutes.
12. Kevin Hatcher (110.2): Over half (54.1%) of Hatcher's point shares were on defense. He's the only player on this list with a negative +/– rating.
13. Teppo Numminen (108.9): More of a stay-at-home type defenseman, 63.6% of his value came on defense (28th all time in defensive point shares).
14. Vincent Damphousse (105.5): A 1,200-point man with over 80% of his contribution on offense.
15. James Patrick (104.9): 63.7% of Patrick's point shares were defensive. Teammate Brian Leetch, by comparison, was 58.7% offense. Patrick ranks 34th all time in defensive point shares.
16. Peter Bondra (104.3): Bondra picked up a hair under 900 points and provided 80% of his value on offense.
17. Glen Wesley (104.1): 73.8% of his point shares were defensive—tops on this list by a wide margin. In fact, he is 17th all time in defensive point shares. He scored 128 goals in 21 years.
18. Eric Lindros (101.6): He had 865 points—but in just 760 games. HIs point shares per game are on par with the likes of Pavel Bure and Paul Coffey. 80.4% of his value was on offense.

The Goalies

I'll write a bit more about the goalies since there are not as many and… well, I love goalies.

1. Curtis Joseph (167.2): Joseph is fourth all time in wins (454) and tied for second all time in losses (352). He's also a remarkable seventh all time in goalie point shares. But he was passed over for the Hall in his first chance. To me, he feels like the Bert Blyleven of the NHL—not appreciated when he played, but when you look back at his numbers… wow! Joseph ranks 7th all time in goalie point shares.
2. John Vanbiesbrouck (152.6): Beezer has been eligible for a while, but has been passed over thus far. His ranking of eleventh in point shares surprised me. I'm sure it would surprise many Hall voters as well. He strikes me more as the Rick Reuschel of the NHL. Certainly not on the Blyleven level, but so much better than he is remembered.
3. Rogie Vachon (144.5): I'll admit—I had never heard of him. He has excellent credentials, however (the wins, the Vezina trophy, the Cups, the Canada Cup, etc.). There are many articles out there passionately fighting for his induction, as he's been eligible for a long time. He ranks 12th in goalie point shares, which seems clearly Hall-worthy to me.
4. Sean Burke (141.9): I was pretty shocked to see Burke here. Maybe he's here because he's 12th all time in games played. Point Shares, like WAR, seems to like playing time (and Burke ranks 13th). Burke is "just" 21st in wins and 6th in losses, which probably clouded my view. Burke certainly wouldn't be the first sub-.500 goalie in the Hall. There are four who played in 200 or more NHL games. Burke has more point shares than all of them (Gump Worsley is close behind).
5. Tom Barrasso (127.8): While "Tom Barrasshole" is a convenient nickname, apparently there was something to it. In baseball, we've seen borderline-to-solid Hall cases miss out because they had strained relationships with the media. Looks like that's happening here. Barrasso reminded me a lot of Grant Fuhr—a goalie on an incredible offensive team who had a higher goals against average, but was probably better than the numbers showed. The Point Shares similarity scores have Barrasso as Fuhr's #1 comp. Fuhr, of course, was elected in his first year of eligibility.
6. Olaf Kolzig (122.2): Kolzig is another near-.500 goalie who played for a lot of lousy teams (only 45 playoff games—21 of those during the 1998 Cup run). He's brand new to the Hall of Fame ballot, only being passed over once on a stacked ballot. I've long been a fan of Kolzig and his work with Athletes Against Autism (founded with Byron Dafoe and Scott Mellanby) makes him all the more admirable. Kolzig is 22nd all time among goaltenders by Point Shares. If he was 22nd all time by a baseball position, he'd be borderline. But hockey has fewer positions and goalies are a pretty important one. That implies that Kolzig (and those above him) just might belong.

During this process, I made a few connections between baseball and hockey.

• Offense is easier to quantify than defense. Offensive point shares are based on playing time, a positional adjustment, and "marginal goals", which is based on "goals created". Goals created is basically based on goals and assists related to the team's total goals and assists.
• Defense is hard. I'm guessing that DRS, Total Zone, or UZR is a similar can of worms. There are a billion inputs based on time on ice, positional adjustments, marginal goals allowed, plus-minus adjustments, etc. I don't pretend to understand it.
• Goaltending is pitching. Goalies allow goals. Pitchers allow runs. But the goalie's defense is also somewhat responsible for those runs. Pitchers' defenses are also somewhat responsible for runs scored. Where's the separation? That's the secret sauce.
• The more recent the data, the more precise. Baseball value stats, of course, get more precise as we get more specific data (like play-by-play, pitch f/x, etc.). The same is true in hockey, as "time on ice", "shots against", "plus-minus", and other data only goes back so far.
• The biggest "snubs" are recent stars. This is true in baseball as well, at least according to the Hall of Stats' biggest snubs (14 were on this past ballot, 13 will return next year). Of course, this is naturally true because recent stars haven't waited as long to be inducted. History shows many of them (in both baseball and hockey) will eventually get in.

Here's one more key observation I made:

The Hockey Hall of Fame is for Great Players… and "Winners"

I made a list of the Hockey Hall of Famers since 1950 (which is approximately when the seasons became a "modern" length) who appeared in 700+ games. Then I added Bobby Orr, who just missed the cut. That gave me a list of 87 players. The median from that group is 112 Point Shares. The range spans from Wayne Gretzky (251 point shares) down to Bob Gainey (26.8 point shares).

When looking into Gainey, I wondered what was up. He was hailed as a great defensive forward, winning four Selke Trophies. About two-thirds of his value was on defense, so Point Shares seems to reflect that. But it also seems that his offensive and defensive point shares simply don't match up with his reputation.

What I did notice is that Gainey won five Stanley Cups. That prompted me to look at the "most overrated" players in the Hockey Hall, by Point Shares.

21 of the 87 players on my list had fewer than 90 Point Shares. Those 21 averaged 3.4 Stanley Cups between them. They were winners. So, that leads me to a pair of hypotheses:

1. Point Shares isn't doing a good enough job of dividing the value of winning teams. Perhaps there are more "intangibles" in hockey than baseball.
2. Winners are overrated, regardless of their individual contribution.

I'm guessing the truth lies somewhere in the middle.

In closing, I'm a "former" hockey fan and incredibly new to any type of advanced hockey stats. What type of stats are out there that could be used for this type of analysis? Am I using the best one? For exercises like this, I need numbers that span the entire sport's history. Point Shares, luckily, does that. Are there others I'm missing?

If you're familiar with these players, what are your reactions to their placement on this list?

And most importantly—what is Point Shares missing about Bob Gainey?