As a more analytical mindset pervades discussions about the great game of hockey it’s become increasingly common to defend some players poor possession numbers because they play tough minutes.  “Tough minutes” are frequently defined as either playing against better competition (as measured by the average time on ice for the opposing skaters), because of a greater frequency of starting in the defensive zone, or a combination of the two.  While I’d agree that playing against the other teams better players is definitely a component of playing tough minutes, there are reasons to consider the emphasis on zone starts dubious.  

I am mainly writing this to draw attention work which suggests that the effect of zone starts on a player performance over an entire season is largely overstated (see here, here, here, and here) and I’m primarily interested in highlighting the explanation as to why.  Simply put, measuring zone starts and then analyzing their impact on player performance discards a lot of useful data about that player.  This is because zone starts, by their operational definition, are confounded with faceoffs and players do not start every shift with a faceoff.  Indeed, it appears that over half of player shifts may begin with an on-the-fly change (see here).

This is an important observation for a few reasons.  First, it suggests that zone starts are commonly interpreted incorrectly, as they are often presented as data accounting for where a player most commonly starts his shift.  Yet, it appears zone starts do not represent on average where a player starts his shift, rather they constitute data on how a player is deployed following a whistle.  Consider the following scenarios:

• Player A is sent out for a defensive zone faceoff.  His team loses the draw and concedes a shot against that is frozen by his goaltender. On the subsequent faceoff the puck winds up in the corner and a puck battle ensues (note: this can be the outcome regardless of who “won” the faceoff).  Player A’s team does not recover the puck and a shot attempt occurs but goes wide of the net.   The puck is then recovered by Player A who attempts to make a pass out of the zone.  The pass does not connect with a teammate and the puck goes down the ice resulting in an icing call.  A third defensive zone faceoff occurs, Player A again recovers the puck but his pass misfires and his team fails to get the puck out of the zone.  Two more shot attempts occur until the goaltender is able to freeze the puck.  The coach changes lines.
• Player B is sent out for a defensive zone faceoff.  His team loses the draw and concedes a shot against that is frozen by his goaltender. On the subsequent faceoff the puck winds up in the corner and Player B emerges with the puck.  He proceeds to pass the puck to a teammate and his team exits the zone.  A turnover in the neutral zone then occurs and the puck comes back in the offensive zone for another shot on goal.  Player B recovers the rebound and again make a pass that leads to a successful zone exit.  His team gets the puck over the red line and dumps it in for a line change.
• Player C changes on the fly while a defenseman on his team retrieves a dump-in and pauses behind the net to set up a breakout. Player C receives a pass from the defenseman and quickly moves it to another player on his team in the neutral zone.  Player C’s team enters the offensive zone and generates a shot attempt that goes wide.  Player C recovers it, moves it back to a point man who gets another shot off that is stopped by the opposing goaltender. The rebound is directed to the corner allowing the opposing team to collect the puck and exit the zone.  Once over the red line the opposing team dumps the puck in.  Player C changes.

Although these are three made up scenarios, they all represent a plausible sequence of events in the course of play.  Let’s break these sequences down a bit further with zone starts in mind.  While all three players started their shifts with the puck in their defensive zone, Player A and Player B would both be credited with defensive zone starts. Specifically, Player A would be credited with 3 defensive zone starts – the initial faceoff, the faceoff after his goalie froze the puck, and the faceoff following the icing he was partially responsible for. Player B would be credited with 2 defensive zone starts – the initial faceoff and then the faceoff after his goalie froze the puck. Player C would not be credited with a defensive zone start, since his shift did not begin with a faceoff.  In terms of shot attempts against, Player A was on the ice for a total of 4, Player B for a total of 2, and Player C for 0.

If we simply look at zone starts and shot attempts we could easily draw the conclusion that more defensive zone starts tend to lead to more shot attempts against.  Yet, consider the scenarios.  Player A is partially responsible for all of those defensive zone faceoffs because he had difficulty making a pass out of the zone, twice.  Player B had a much better shift than Player A, as he quickly got the puck out the zone twice.  Furthermore, the second shot attempt against was the result of a teammate turning the puck over in the neutral zone.  In terms of Player C, his data is essentially “lost” because even though he also started in the defensive and moved the play forward, he is not given a zone start since his shift began on the fly.

This reminds me of my frustration with using 5 on 5 close (when the game is tied or within 1 goal in the 1st and 2nd, and tied in the 3rd) to evaluate team possession stats.  Simply put, why discard all that data!  Thankfully, many have moved beyond 5 on 5 close statistics and now use 5 on 5 score adjusted statistics which do not discard data.  Something similar needs to happen with zone starts because we are losing valuable information by only considering where a player is typically deployed for a faceoff.  As the links above demonstrate (again see here, here, here, and here) it appears that the majority of a player’s shifts do not begin with a faceoff.  This begs the question of how informative zone starts really are.  Can we really judge that a player is given tough minutes on the basis of zone starts?  Are some of the players who are thought of as getting “tough minutes” somewhat responsible for those “tough minutes,” as in the case of Player A?

Essentially, I’m arguing that quality of competition, while it still has its flaws, is a much better indicator of tough minutes.  Zone starts can be informative, as they may clue us in as to how a coach views a player. For instance, there is certainly a reason John Hynes puts Andy Greene and Adam Larsson out for more defensive zone faceoffs than Damon Severson and John Moore, but those zone starts may only capture about 50% of their actual shift starts, skewing our perception of where each player starts their shift.  In other words, don’t discard data when you don’t have to.