Quantifying the impact of players making league minimum on team possession

In this post, we establish a baseline against which player performance can be judged.

For a long time we've been limited to using phrases such as "one of the better players in the league at driving play" to discuss a player's Corsi value to his team. We know that having good players who drive possession will result in high goal differentials in the long run, that's been well studied, but finally this summer we're beginning to quantify that effect.

The trouble with using Corsi has always been context. There are a lot of factors that influence a player's Corsi that may inflate or drop his numbers that are (somewhat) out of his control. I've always tried to look at Corsi in the context of ice time, zone starts, quality of teammates and quality of competition, and there's been some progress made on quantifying those effects as well. But we're still largely limited by the dataset we have to work with.

The challenge in the NHL is that teams can shelter their replacement-level players. They're not going to see the same ice time as the player they're replacing but their inclusion in the lineup does impact the roster. Moves must be made to accommodate them and (presumably) this will have an impact on long-term team possession ability.

Why do we care how well a player making the NHL minimum drives possession? Because it provides the freely available talent possession floor. It's the reference level all other players can be judged against. So instead of looking at how players making the NHL minimum performed on the ice, we can take a look at how their team did as a whole when they're in the lineup to gauge what kind of impact their presence had.

Value of NHL Minimum Salary Players
Group Avg Games Played Avg TOI% Corsi For% Corsi Rel% Fenwick Close% Differential Range
Forwards 17.2 19% 47% -3% -0.52% -0.39 to -0.66
Defensemen 13.9 25% 48% -1% -0.92% -0.7 to -1.14

Fenwick Close% Differential: The difference between the Fenwick Close% with the player in the lineup minus the Fenwick Close% with the player out of the lineup. Range represents error estimates.

I think the most important column to look at is the Fenwick Close% differential, which is the Fenwick Close% performance of the team as a whole when the player was in the lineup compared to when he was out. This removes zone starts, QoC, QoT, all sheltering effects, and leaves the raw results of the roster change. We can see that on average, players who are paid the NHL minimum salary drag down their teams' Fenwick Close% when in the lineup by 0.52% for forwards and 0.92% for defensemen..

That is, if we took a league-average team, took out a league-average player and replaced him with a player making the league minimum, we would likely see that team's Fenwick Close% drop from 50% to 49.5% if that player is a forward, and 50% to 49% if he is a defenseman.

In the next edition of this series, we'll look at how this changes as players gain more ice time.

Notes on Methods:

I grabbed Fenwick Close data from Extra Skater (when it was still around) for every game from the 2007-08 to the 2013-14 season (excluding the lockout shortened 12-13, and using different datasets prior to 11-12), as well as the rosters from those games. For each player, I analyzed his team's Fenwick Close% for games in which that player was in the lineup compared to when he was not (but still part of the team; players playing for multiple teams per season were excluded from the data). I then grabbed salary information from NHL Numbers to bin all players who made the NHL minimum salary. This resulted in approximately 866 forwards and 442 defensemen.

I did not include player bonuses when calculating player salary so, for example, Tommy Wingels was considered a league-minimum player for 2011-12 because his salary without bonuses was the NHL minimum in that season. As you can imagine the sample sizes for this type of study remain small and the conclusions are relatively weak as a result of the lack of certainty.