Player types: Conclusion, and in defense of the “low-event” superstar



Well, after a couple of weeks of research and organizing and sorting, we got six basic standard “player types” that are sorted between “high- and low-event” players that I’ll be using for a little bit more of my analysis on the Nations.

If you missed it, those six player types are:

The Two-Way Forward
The Defensive Forward
The Offensive Liability
The No-Way Forward
The Defensive Liability
The Offensive Forward

Pigeon-holing hockey players into groups wasn’t particularly the purpose of the series, but it’s nice to know that about 75% of the players listed were probably the ones who fit into a certain category. The purpose of the series is more to be able to discuss hockey players in more qualities than “good” and “bad” but more in a sense of what they can do right.

Certain players are useless with the puck but are tremendous at preventing shots. Since one goal for is equal to one goal against, this makes a player who is bad at defense but an offensive juggernaut the yang to the more defensive one.

I’ve often used the phrase “handicapped by superstars”. Certain General Managers are under pressure to pick up established hockey players. Two recent examples of this come from Philadelphia and the New York Rangers:

In Philadelphia’s case, under pressure to lock down a star goaltender, Paul Holmgren gutted the remainder of the Flyers’ core including their key players who played tough minutes to be able to sign Ilya Bryzgalov. While the team is left with an impressive slew of offensive performers, none are battle-tested in their own end. The Flyers could have been more efficient with their ressources by seeing that Brian Boucher or Sergei Bobrovsky actually had similar statistics to Bryzgalov, and the cheaper improvement would have been Tomas Vokoun who signed a cheap deal in Washington.

The Folly of Sather

For the Rangers, much was made of their big-name signings in the pre-lockout era to forwards such as Eric Lindros, Bobby Holik, Jaromir Jagr, Pavel Bure, Alex Kovalev and Martin Rucinsky. Despite the narrative ideal that states “you can’t just sign superstars to win a championship” the Rangers were actually quite good at scoring goals throughout that period, but their inability to find players to play defense hurt the team and they consistently allowed well more than they scored and missed postseasons as a result. The Rangers have returned to this tenet, and instead of bolstering the defensive unit around Henrik Lundqvist and finding cheap alternatives for the forward group, they went out and signed Brad Richards who is not a player with strong enough underlying numbers to justify his signing at the price they paid.

“Win with the nobodies and the fans showed up, and the nobodies became stars,” wrote Michael Lewis in the 2003 classic Moneyball. “Lose with stars … and the stars became nobodies.” Efficiency is key in assembling a hockey team and what’s important is finding exactly which guys to use in the right situation. In my bizarro fantasy world of four lines, knowing that it’s virtually impossible to get 12 plus-players every night, you have to separate. Give your top line easy minutes sheltered by a third line that can eat up tough minutes. The second line should be the best combination of two-way players and hopefully they can break even. The fourth line is your group of small minutes players who will keep their heads above water defensively so you at least don’t give anything up. They don’t have to be stars. You can make the playoffs and win a playoff round with Sergei Kostitsyn as your leading scorer, provided you also have David Legwand and Jerrod Smithson eating up tough minutes. Is it flashy? No. Is it effective? I would say so, and teams have to work to find the players who will succeed in those types of roles.

One of the issues with my recent analysis, or any statistical analysis, is that there are still a lot of variables that are left unaccounted for. Offensive shot quality leads to more goals, but we do a shot differential plus/minus. We stack players from different teams up against one-another, but we have no easy way of telling the effect of either the player in the system or account for the bias from that team’s home scorekeeper. Detroit, for instance, doesn’t record many missed or blocked shots. Tampa Bay block a lot of shots, which warp their Fenwick numbers. New Jersey have a tight defensive system that leaves Brian Rolston and Henrik Tallinder as apparently two of the top defensive players in the NHL.

Is that quantifiable? The one sure measurement we have in hockey is that of the goal. The shot, block and missed shot are subjective measurements. New Jersey was the ninth best team in goals allowed last season, and, by watching Martin Brodeur, we can know that he wasn’t particularly terrific. It’s safe to say that New Jersey is still a good defensive team. Are Rolston and Tallinder the best at defensive play within the system? Yes, the question is whether they are the best overall, independent of the system.

I’m not sure a relative measure would attack that conundrum. It makes no sense to punish a player like Manny Malhotra because Ryan Kesler is also a good hockey player, just how it makes little sense to champion Mikhail Grabovski because of the poor play of Tyler Bozak. Instead, I’m attempting to get around this, breaking apart the numbers and seeing what happened on the ice when a player was on: where the puck was and what pace was the game. From there, we can determine whether a player was successful in that role.

Splitting the Corsi or the Fenwick number apart makes sense, because every shot for or against will be submitting itself to the same biases. A shot for in Detroit may mean a different thing than a shot for in Anaheim, but I have to think that a shot for in Detroit is equal to a shot against in Detroit. Splitting the Corsi number to the parts that make it up allow us to determine what happens when a player is on the ice. If they have a high amount of offensive and defensive chances while they’re on, they’re more likely to be a “high-event offensive” player. The reverse means that they’re more likely to be a “low-event defensive” player.

Consider that many of the high octane scorers in the league ended up as either “defensive liabilities” or “offensive forwards” in this analysis and the ones who didn’t, in the “two-way forward” group or “no-way forward”, were still on the “high events” side of the graph. It is not until R.J. Umberger, tied for 49th in goals, does a “low event” player make it onto the list of top goal scorers last season.

Tracking the Invisible

The un-tracked “goals prevented” statistic would therefore fill up with more “low event” guys. We can make a reasonable case for Brian Sutherby or Blair Betts as strong players, or Boyd Gordon, David Steckel, David Clarkson and Derek Dorsett as having legitimate value. These are all cheap players who don’t give up many shots via the standard data pulled from The more you adjust for the variables, the more you can discover about these hockey players, and they’re so good because you so don’t see them and what they do well. For a low-event guy, you never see what he does well, because it doesn’t tangibly exist.

Take two defensemen. Let’s assume that one gives up two scoring chances and four outside shots against, while the other gives up two scoring chances and two outside shots against. The second defensemen had the stronger game, but he didn’t “keep the majority of the shots to the outside” like the first one did. Without expressly counting, we can’t tell how many actual shots were taken against whom and deal with only our impression of the game.

So, cheers to the low-event players and the unsung heroes who play for near-free and soak up the tough minutes for the stars. Grab a Jarrod Smithson and turn Sergei Kostitsyn into a superstar. Alex Ovechkin’s most important teammate this season may not be a linemate, but quite probably the low-event two-way forward in Joel Ward. Adam Hall, Nate Thompson and Ryan Malone were key to preventing goals in Tampa Bay last season and they responded by turning from one of the worst in the conference to one of the best in a span of a season. Meanwhile, two lone bright spots in Calgary last season were Tim Jackman and Curtis Glencross, low-event guys who opened up ice, but the Flames forwards were swallowed up by variance at the wrong times.


Superstar protection can come in the form of a goon, but, in the new, numbers-oriented NHL, it may become the low-key guy who eats up the tough minutes without allowing much against. Keep an eye out to see how these players are deployed throughout the season, and hopefully we’ll be able to tack on more information as coaches start to give more offensive zone time to the star, high-event forwards who can do something with it.

  • SmellOfVictory

    great read. i have two thoughts:

    one, as an avid fantasy hockey participant, i’d love to see a list of these strong, “low event”, tough minutes guys so that i could perhaps target their scoring teammates who may be flying under the radar (finding this year’s Sergei Kostitsyn, perhaps).

    two, one thing you said concerns me. my understanding of what makes corsi and fenwick so desirable as statistics is that the greater pool of events make them more effective than simply using goals. more events = reduced noise, right?

    but then we are limiting events to only while a particular player is on the ice, and maybe only during even strength, and heck, maybe only when the score is close. if we further narrow our analysis by the rink they’re playing in, are we still ensuring that we maintain that advantage of more events?

    just a thought. mostly i’d love to see that list!

    • SmellOfVictory

      Although part of the draw of advanced stats is that they do provide larger event pools, the major draw is that they provide a more comprehensive representation of how good a player is. This is done via the removal of a great deal of randomness from the stat (which both goals and +/- contain to a significant extent) as well as taking greater context into account.

      This is just supposition on my part, but I think it would be difficult, to say the least, to filter corsi/fenwick events to such an extent that you no longer had a reliable sample from which to draw over the course of, say, a season.