Getting Pucks on Net: Luck or Skill?

As with most metrics in hockey, there are so many variables that we need to take into account when evaluating an individual player. If you can recall from our player usage saga, some of these factors include, but are not limited to: zone starts, quality of teammates, quality of competition. 

Shooting percentage is a variable that certainly falls victim to some of these contextual clues. The main one, for shooting percentage, being quality of goaltender faced. This blog will attempt to eliminate the goaltender variable by looking at a players individual unblocked shot attempt on net percentage, or what I am now dubbing, iUSAF%. This is simply individual shots on net divided by individual fenwick, a barometer of a players' skill of getting pucks on net. Goals don't matter here, just the ability to hit the net when you get it passed the line of defense (unblocked shot attempts). 

In an attempt to judge whether or not this is a skill or luck, two sets of player month-to-month data will be compared (November to December and February to March - minimum 8 games played in each month) and one set of season to season data (13/14 to 14/15 - minimum 40 games played in each season). These data sets include only forwards.

Expected outcome is more luck in the month-to-month comparison, and skill in the season-to-season comparison. We will judge luck versus skill by calculating the R^2 of our data sets. Lower R^2 will have us leaning towards luck, while a higher R^2 will have us leaning towards skill.

November to December:

NDPS is the vector naming convention (November to December Player Shooting)

NDPS is the vector naming convention (November to December Player Shooting)

November to December shows no fit, with an R^2 of 0.0008602. Complete luck.

February to March:

February to March shows no fit with an R^2 of 0.002668. A little better than November to December, but still complete luck.

13/14 to 14/15 (season-to-season):

Season to season also shows no fit with an R^2 of 0.01108. Complete luck. 

Hypothesis = right in the sense that the season-to-season data definitely has more 'skill' than the month-to-month data sets; but also wrong because the R^2 for season-to-season data dictates that skill does not 'outlast' luck in the end.

There seems to be no predictive power in a player's shooting efficiency. Eliminating the skill of goalies, since we are looking at shots that are on net (including goals, of course) and the effectiveness of opposing team's shot-blocking by only accounting for unblocked shot attempts, we still do not see a skill-based explanation for a players efficiency.

Players do not seem to have any control over their shot efficiency.