There has been an increasing push in the hockey community towards the integration of advanced statistics and that was highlighted over the summer when the Toronto Maple Leafs started scooping up talent known for their use of analytics, like new assistant general manager Kyle Dubas. Meanwhile, general managers and coaches from around the league have openly endorsed analytics as a tool for evaluating talent.
Don’t forget, for everything NHL, check out Rotoworld's Player News, and follow @Rotoworld_ HK and @RyanDadoun on Twitter. Also a big thank you to Gus Katsaros (@KatsHockey) for his help with this article. He is the Director of Analytics & Pro Scouting with McKeen’s Hockey and is also a contributor with Rotoworld.
But what does that mean for fantasy owners? Are advanced statistics something that will eventually become a mainstream option (option being the key word here as opposed to default category)? If they are, then are they something that you should consider using?
Well let's start at the beginning with the simple question: What counts as an advanced statistic? For some that's a question that's been long since answered, but if that's not the case for you then this will double as Advanced Statistics 101. The first lesson being that the "advanced" is part of the name rather than definition. As you'll see below, the math involved isn't as intimidating as you might assume.
When talking about advanced statistics in the NHL, you're going to hear two words a lot: Corsi and Fenwick. Both attempt to do the same thing: Act as a proxy to approximate puck possession.
Corsi measures a team's shots on goal, missed shots, and blocked shots against versus the opposition's shots on goal, missed shots, and blocked shots for. The basic idea being that the team that's more commonly dictating the play will get more opportunities to shoot the puck. Fenwick is the same idea, only it doesn't factor in blocked shots under the argument that blocking shots is a reflection of the skill of the opposing team.
So far we're just dealing with adding and subtracting a few numbers, but as you might suspect, we can use that as the base for more in-depth analysis.
Let's focus on the player statistics because that's ultimately were our interest lies. Just as Corsi or Fenwick can be used to measure a team, the same can be applied to players by looking at what happens when said player is on the ice. For example, Jonathan Toews raw Corsi For events lies at 655 and his Corsi Against events is 488, per the War on Ice. That gives him a Corsi plus/minus of plus-167. All of those numbers and subsequent statistics deal with five-on-five situations, which we'll address later, but for now it's enough that you know that's the case.
The league leader in that regard is Drew Doughty with a plus-239 while Tyler Myers is at the bottom of the NHL's standings at minus-362. So tying this back into fantasy hockey, you could see how Corsi plus/minus could eventually be a possible substitute for traditional plus/minus for those so inclined. You might also wonder if that's really a better indicator of a player's individual skill than traditional plus/minus. After all, the bottom nine players in terms of Corsi plus/minus are all Buffalo Sabres and that has more to do with the overall skill of the team rather than the individual.
You can use advanced statistics to attempt to neutralize that though, or at least make the team's overall skill less of a factor. How? Well, first you turn the player's Corsi For/Against into a percentage form, or CF%. So Myers' CF% is 35.2%, which means that when he's on the ice, his team takes 35.2% of the shots fired. With that in hand you can look at what the percentage is like for the team when he's not on the ice, in this case 39.3% (aka CF%off). By taking your CF% and subtracting it by your CF%off, you're left with the relative Corsi (or CF%Rel). In Myers case that's -4.1, which basically means that the Sabres have been doing a better job from a Corsi perspective when Myers is not on the ice compared to when he is playing.
Because Corsi%Rel is, as the name suggests, percentage based, small sample sizes can mess with it. So, as is the case with a goaltender's save percentage, you'll want to have a base minimum number of games played in order to qualify for consideration on the leaderboard. If you set the cutoff at a minimum of 20 games (and I'm just doing that arbitrarily for the sake of this article), then Manny Malhotra is in last place at -18.3 as he has a fairly bad CF% (33.6) on a pretty good Montreal squad. On the other end of the spectrum, two-time Selke Trophy winner Patrice Bergeron is leading the pack with a plus-9.7 CF%Rel.
Maybe there's potential there for CF%Rel to be used as a category at some point down the road as a lot of fantasy leagues don't deal with a skater category that produces an average rather than cumulative number. Incorporating a statistic like this could also add more strategy to the selection process as you're no longer just evaluating a player, but also trying to determine if he'll be more valuable than the average player on the team he plays for. That makes guys like Arizona's Martin Erat, who is excelling from a Corsi perspective on a relatively poor team, more desirable than they otherwise would be.
That's a couple examples about the types of statistics that might be used in fantasy leagues some days, but without even going deeper into analytics, there's so much in the way of variation. I briefly touched on the fact that the Corsi figures I was using were 5-on-5 even strength numbers. Obviously shot totals are going to be different in power-play/penalty killing situations and incorporating them can contaminate the results. However, there's nothing to say that you can't have a CF%Rel category for even strength and then another for the man advantage. You can even throw in situational variables, like what the player's CF%Rel is during power-plays when his team is leading by two goals.
Also, while I was using Corsi as the base for the sake of convenience, we could have done the same thing with Fenwick.
There are also analytics that don't involve Corsi or Fenwick. One interesting example is PDO, which is a player's on-ice shooting percentage combined with the on-ice save percentage. The idea is that those numbers should equal 1 or 100 depending on how you're scaling it. If the player's PDO is significantly higher than 1/100 then theoretically he's been lucky and is due to regress while a low PDO might indicate that he hasn't been rewarded for his efforts, but should be in the future. It doesn't make for an ideal fantasy category because in a perfect world, everyone's PDO should move towards the same point over time, but it is a great statistic for fantasy owners to be aware of.
One of the most common questions fantasy owners have is if a breakout player has staying power or if a slumping veteran will bounce back. While PDO isn't the full answer to that question, it can be used as a tool to gauge if the player in question has been playing over his head or conversely has been unlucky.
Looking at it from a different perspective, using a player's PDO can help you attempt to skew trades in your favor by going after potentially over/undervalued players. Just keep in mind that, as is the case with all analytics, it's a helpful tool but not a cure-all.
We're not at the point yet where analytics are easily accessible fantasy categories, although at the rate they're being adopted, maybe that day isn't far off. I doubt they'll become mainstream as most casual players are already comfortable with the current categories and might not welcome a change to a system that already works. Still, for those seeking variety, it would be interesting to experiment with when the opportunity arises.
In the meantime, if you're interested in advanced statistics, War on Ice is a great resource. It's worth noting that NHL deputy commissioner Bill Daly suggested in November that the NHL website might end up including analytics in the future. If that ever happens, then these statistics will become a lot easier for the larger hockey community to get behind.