Hockey Outsider
Registered User
- Jan 16, 2005
- 9,375
- 15,392
I wanted to see if we could identify the greatest two-way players in NHL history, using some basic data. This was inspired by @DitchMarner's recent thread in the polls forum.
The starting point for the analysis is GF%. (That's a relatively new stat, but it can easily be calculated using the plus/minus component data - which I took from @overpass's enormous spreadsheet). GF% looks at the percentage of the total goals that are scored by a player or team. (A simple example - if a team wins 3-2, their GF% is 60%, since they scored three of the five goals).
If we just look at GF% (which lots of hockey fans do), it doesn't tell us very much. GF% (like plus/minus) is heavily dependent on the quality of a player's team. It makes a big difference if someone is playing in front of Dominik Hasek or Dan Cloutier. Therefore this analysis focuses on the increase in GF% between when a player is on the ice, compared to when he isn't. The assumption is, if a player is doing better than the rest of his team, he's probably doing something right. (He might have an objectively poor GF% if he's playing in front of Cloutier, but that would also impact his teammates, so looking at this on a relative basis makes sense).
The analysis is done only for even strength. That's a necessary step. Otherwise, anyone who played primarily on the powerplay would look better than anyone who played primarily on the penalty kill.
There's a lot of context that goes into these numbers. I go into that in some detail in this post (I recommend reading it if you're having trouble falling asleep). I want to be clear that I'm not posting the following table as definitive lists - it's just to get the conversation started. Context and common sense need to be considered.
Parameters:
The starting point for the analysis is GF%. (That's a relatively new stat, but it can easily be calculated using the plus/minus component data - which I took from @overpass's enormous spreadsheet). GF% looks at the percentage of the total goals that are scored by a player or team. (A simple example - if a team wins 3-2, their GF% is 60%, since they scored three of the five goals).
If we just look at GF% (which lots of hockey fans do), it doesn't tell us very much. GF% (like plus/minus) is heavily dependent on the quality of a player's team. It makes a big difference if someone is playing in front of Dominik Hasek or Dan Cloutier. Therefore this analysis focuses on the increase in GF% between when a player is on the ice, compared to when he isn't. The assumption is, if a player is doing better than the rest of his team, he's probably doing something right. (He might have an objectively poor GF% if he's playing in front of Cloutier, but that would also impact his teammates, so looking at this on a relative basis makes sense).
The analysis is done only for even strength. That's a necessary step. Otherwise, anyone who played primarily on the powerplay would look better than anyone who played primarily on the penalty kill.
There's a lot of context that goes into these numbers. I go into that in some detail in this post (I recommend reading it if you're having trouble falling asleep). I want to be clear that I'm not posting the following table as definitive lists - it's just to get the conversation started. Context and common sense need to be considered.
Parameters:
- the tables are based on a player's best five seasons (why five? that's how many fingers I have)
- seasons don't need to be consecutive
- a player must appear in at least 80% of the games (if he appears in fewer than that, the season is disregarded entirely)
- forwards must play at least 20% of ES minutes, and defensemen must play at least 25% (this is done to filter out goons who get limited ice time and typically face weak opponents - if a season doesn't meet this criteria, it's discarded)
- forwards must play at least 15% of PP and SH minutes, and defensemen must play at least 20% (this is an indirect way of taking "two way" play into account - I figure that if someone is getting minimal deployment in either situation, they're not a complete player)
- the five year results are a simple average (ie not weighted for games played - that would improve the results, but it adds a level of complexity to the calculation - the impact should be minimal as seasons with less than 80% of the games played are ignored)
- the data starts in 1960, and only goes up to 2020 (therefore recent results from McDavid, Barkov, Matthews, etc are excluded)
Last edited:
