This has been quite the week for hockey metrics.
On Friday, a very well-known hockey blogger, author and radio personality Sean McIndoe a.k.a. DownGoesBrown posted an article on the basic ins-and-outs of hockey metrics over at Grantland. I strongly recommend reading it even if you have a grasp for hockey metrics as McIndoe has a unique way of making metrics funny and that’s not easy.
But in a much more nerd-friendly post, Michael Parkatti – an Edmonton Oilers blogger – wrote about a very controversial topic in the “advanced stats” community. His topic: the sustainability of shooting percentages. More specifically, he wrote about a player being able to maintain a high On-Ice shooting percentage. The difference between standard SH% rates and On-Ice SH%* is the former is specific to the individual’s shots on goal only, while the latter accounts for all shots on goal taken by a team while a specific player is on the ice. For example, Sidney Crosby’s SH% last year was 12.1% but his On-Ice SH% was 13.9%. This means he shot 12.1% himself but the team shot 13.9% while he was on the ice.
*note: On-Ice SH% factors in only even-strength, 5 on 5 play here.
A player’s individual shooting percentage can vary wildly over his career. I wrote about how the difference between Alex Ovechkin’s highest and lowest SH% seasons affected him. If you don’t want to read it, you only need to take this away; despite taking just 80 fewer shots in 2010-2011 compared to 2007-2008, he scored 33 fewer goals. If individual SH% was fairly constant, this would imply a shooting percentage of around 41%, and I can assure you this is not the case.
You can estimate a range of expected outcomes with regards to SH%. In Ovechkin’s case, it’s safe to say he’ll probably shoot between 10%-15%, but as I just demonstrated, this is a fairly useless estimation and there’s an outside chance it will still be flat wrong.
For the most part, if you see a player shoot at a very high rate over any extended period, it’s a safe bet that it will come crashing down sooner rather than later. James Mirtle wrote about this with regards to Nazem Kadri this year and, sure enough, he was right; after Kadri shot 20% through his first 36 games this past season, he shot 5.5% in his next 19 games (including playoffs). As with most things in life, the answer lies somewhere in the middle of the two extremes. These examples are available everywhere; Daniel Winnik scored five goals in his first five games this year shooting 38.5%, he proceeded to score one goal in his next 43 games shooting 1.2%.
Regression to the mean can be a cruel mistress.
On a macro level, it’s difficult for a team to drive scoring chances. In other words, a team cannot (that we know of) create a higher percentage of scoring chances relative to other teams (although Pittsburgh is trying), which is why puck possession is so important. If scoring chance/shot attempt ratio is fairly constant across the NHL, the only way to balance that in your favor is with more shot attempts i.e. puck possession. This is the exact same argument that I made in my Alex Ovechkin article; if he isn’t shooting at a high efficiency rate – like he did in his outlier 2010-2011 season – the only way to mitigate that is by shooting more (which he did this year compared to last year) and draw more penalties as a team (shooting percentages sky-rocket on power-plays).
What I did then was look at whether or not there have been players in the NHL recently that have been able to sustain both high shooting percentages and goal-scoring rates. I wanted to look at goal totals – which typically isn’t recommended as focusing on results instead of process will blur how you got there – because I want to see who has been able impact their team’s results in a significant manner. Alex Tanguay and Curtis Glencross have a combined career shooting% of 17.9% over their combined 19 full seasons in the NHL, yet neither of them has ever cracked 30 goals.
So I took a look at two sets of two seasons, 2009-2010/2010-2011 and 2011-2012/2013. The parameters for the first set are the player must have scored 50 goals over those two seasons and shot at least 14%. The parameters for the second set are 40 goals and shot percentage at least 14% (the discrepancy in goals is because of this lockout-shortened season). Here is the list for the first set, a list of 14 names. Here is a list for the second set, also 14 names. There is one name you can exclude from the second list in Jordan Eberle because he was not in the NHL in 2009-2010. Also, it’s dicey with Brad Marchand because he only played 20 games that season with the Bruins.
Nevertheless, you’ll notice something very quickly; there isn’t a lot of overlap. In fact, in those two sets, only Steven Stamkos and Thomas Vanek appear on both. It shouldn’t come as too much of a surprise, both are considered elite goal scorers. Even still, just 14.2% of the names on the first list appear on the second list, which gives you an idea of the probability of a player sustaining a high shooting percentage. While we don’t know for sure, it’s also probable that Crosby would have appeared on both lists had he been healthy in 2011-2012 when he missed 60 games. This would make it a 21.4% chance of a name to appear on the first list and then the second. However, that number for anyone not named Stamkos or Crosby is 8.3%. To put it simply: sustaining high shooting percentages with high production is very, very difficult.
This is an important concept in fantasy hockey. While earlier I showed that the old adage “he makes the players around him better” is actually true, the myth of the elite and efficient goal scorer is truthful in very few scenarios. Corey Perry scored 50 goals in 2010-2011, which is excellent. He also shot 17.2% that year and that mark ended up being far and away his five-year single season high and he has not cracked 40 goals in any other season. Jeff Carter and Jakub Voracek were both in the top-10 in NHL goal scoring last year; Carter shot 19.5% last year despite a career average of 11.5% and a previous high of 13.5%. Voracek shot 17.1% despite his career average being 10.4% which was also his previous career high.
There are mitigating factors, of course. Voracek went to an offensive team in Philadelphia and played alongside Claude Giroux, Carter is a natural goal scorer playing for a good possession team which inherently means more scoring chances. Sample size is the issue here as we see even a two season sample probably will not accurately portray a player’s true shooting prowess. However, it’s exactly these mitigating factors that make betting on a player sustaining a high shooting efficiency a fool’s errand.
The best way to avoid over-paying for players at the draft table is to avoid regression candidates and if you see players coming off career seasons like Carter and Voracek, let someone else over-pay for goal scoring.
