I was reading some discussion of the Ottawa Senators over on HFBoards (http://hfboards.com/showthread.php?t=360266&page=2
in case you're interested.)
Here's a comment:
"Kelly/Vermette have by far the most penalty killing minutes for our forwards and yet they are considered the worst? If Heatley is our best statistical PKer why does he have the least amount of PK time?And Volchenkov is the worst statistical PKer? Phillips has over 300 minutes on the PK...yet according to the +/- ranking we should be using Schubert."
It would be wrong to infer that a good +/- on the PK by itself means that a player should be playing the PK. You have to take into account who's on the opposing PP:
Volchenkov 279 minutes, -4.30 G/60, opponent PP = +4.77
Phillips 319 minutes, -3.39 G/60, opponent PP = +4.76
Vermette 195 minutes, -3.39 G/60, opponent PP = +4.76
Kelly 215 minutes, -3.70 G/60, opponent PP = +4.76
Schubert 78 minutes, -0.77 G/60, opponent PP = +4.39
Heatley 64 minutes, +1.87 G/60, opponent PP = +4.27
There isn't much spread in opponent PP (0.21 standard deviation league-wide), so you can't compare Ottawa's top unit to Schubert and Heatley, who only get PK time in very specific circumstances - against inferior teams or against a team's 2nd or even 3rd PP unit. In fact, Heatley's opponent PP is the second-lowest in the league for anyone with 60 minutes of PK time.
So the statistics don't suggest that Heatley and Schubert should be on the ice. In fact, we don't know what would happen if Heatley played against the other team's first PP line - but we do know that he's not doing it right now, and that we shouldn't be misled by his pristine PK stats.