I thought I'd take a look at some of the quantities surrounding even-strength goal-scoring. Again, I'm looking at the persistence of each of these rates in even and odd games, just for initial shots (no rebounds.)
Talent | Regression to Mean | R^2 |
Goals/GP | 28% | 51% |
Shots/GP | 1% | 98% |
Shot Location | 11% | 82% |
Shooting % | 74% | 9% |
So players quite clearly control the volume of shots they generate and the locations they take the shots from. But actually putting the puck in the net is a very transient talent, so the number of goals a player scores is, as you might imagine, not exactly deterministic.
If we combine shot volume with the previous analysis I did of shooting percentage and shot location, we get the following breakdown of true and transient talent:
Percent | Type | |
Shots/GP | 39 | Talent |
Shot Location | 14.5 | Talent |
Shot Location | 2 | Transient |
Shooting % | 11.5 | Talent |
Shooting % | 33 | Transient |
A player's goal-scoring rate – on a per-game basis – is about 65% driven by his true talents – whether he's a shooter, where he shoots from, and his ability to pick corners from a given spot on the ice over and above other players. The other 33% is shooters making their shots or whatever you want to call it. Again, this analysis is restricted to initial shots at even-strength – if we included rebounds, then you'd see that goal-scoring was even more of a talent, and if we further include power-play goals, you'd see that goal-scoring was even more talent-driven (or driven by a coach's decisions regarding PP time.)
I think (hope) this resolves some of the confusion around the notion of shooting being luck-driven. On a given shot, from a given spot, luck rules. But when we look at the sum total of a player's skills to figure out how many goals he's going to score, true skills dominate.