by: Michael Clifford | February 7th, 2013
To be sure, advanced statistics in the NHL are still in their infancy.
This is not like baseball where we have a Bill James that pioneered everything. In fact, it’s been just a little over 30 years since hockey coaches have been using videotape to study opponents. The most commonly used advanced stat, ‘Corsi’ and all its variations, are a product of this millennium. Now you can understand why I say that this is still a relatively new concept.
Advanced hockey stats have really only begun being prevalent in the hockey writing and fantasy hockey world since the last lockout. This is because writers (and now fantasy hockey owners) wanted to understand why there would be such a variance between a player’s production from year to year. Now, there is a wide array of websites where you can research advanced statistics. I will provide a list of some of my favorite ones at the end of this piece.
Before I continue, I want to say that I don’t really lean on advanced stats just yet when formulating my approach to a fantasy hockey season. There are some stats I do factor in (PDO, being one of them and I will discuss that later), but some I don’t really rely on too much. There are a variety of reasons:
- Unlike baseball, a lot of advanced hockey stats are team-dependent. Baseball uses individual events e.g. 2-0 Pitch from a lefty, RISP with none out, how does a hitter hitter’s homerun rate vary from park to park and so on. In hockey, except for shootouts, there really aren’t very many individual events. How do you measure how often a player scores on a two on one? The quality of the goaltender, the positioning of the defenceman, the quality of the ice (early or late in a period), whether the puck is rolling or not etc. are all factors that play into how that two on one turns out.
- You can’t really just ask “how efficient is Player X on the power-play?” The answer is dependent on the players around him, the system of their PP setup, the aggressiveness/passiveness of the penalty kill and so on. Obviously, Chris Kunitz (F-PIT) is going to be more proficient on the PP, because of the talent around him; Derick Brassard (F-CBJ) less so. This is despite the fact that there is a negligible amount of talent difference between them.
- The value of a player is almost completely dependent on the players that surround him. In talking to Tom Fitzgerald (one of the guys who runs this site), I explained to him that point in this manner: Think of Pascal Dupuis (F-PIT). For the bulk of his career, he’s been no better than a second or third line players. All of a sudden, since joining Pittsburgh, he’s a first line player because he gets to play with Sidney Crosby. On just about any other team, he is a third liner and of no fantasy relevance (for the most part). Instead, he is a must-have bench player.
- The applications in fantasy (and I’ll get into this in a bit more detail) is still a bit precarious. There are some indicators of who will regress and who won’t, but the usefulness is still a bit vague. Again, I’ll explain why later.
Now that all that stuff is out of the way, time to get to the meat and potatoes.
The Most Common Advanced Stats
PDO is a relatively simple concept developed by Brian King for Vic Ferrari, an Edmonton Oilers blogger. It is simply the sum of the on-ice team save percentage – the save percentage of the goalies when a given player is on the ice – and the on-ice team shooting percentage – the shooting percentage of the team when a given player is on the ice. So if a player’s On-Ice SV% is .900 and the On-Ice SH% is 10%, this gives you 1.000, or in PDO terms just 1000. This 1000 rating is the norm; if a player dips below it, expect them to progress to the norm, if a player is above it, expect them to regress.
This example of PDO was shown on a blog at The Score. Last year, Tyler Seguin (F-BOS) was a remarkable +27 through the first 34 games of the season. Extrapolated for a full 82 game schedule, Seguin was on pace to be a +65 for the year. Considering only one player had gotten to +50 (Jeff Schultz, D-WSH, 2009-2010) since the last lockout, this was a highly unsustainable pace. According to the article, Seguin had a PDO of 1087 at the end of November. While he still had a good December (eight points, +8), he dropped off dramatically afterwards. In his next 47 games, Seguin went +7. Predictably, Seguin’s PDO dropped from 1087 at the end of November to finish at 1022 at the end of the season.
In this sense, PDO is a good way to use small sample sizes to see who is going to regress or return to the mean (1000) over the course of a season. It is very useful for identifying slow-starters and hot-starters.
For my money, PDO is the best advanced statistic to use when talking fantasy hockey. In short, it’s a great indicator on who you should buy-low on and who you should sell-high on. Let’s take Dion Phaneuf (D-TOR) as an example. At the time I’m writing this, he is sporting a 938 PDO. That is exceptionally terrible. Over his last four years, he’s sported a PDO between 979 and 1004. This would be a fantastic time for a fantasy hockey owner to buy low on Dion Phaneuf. Unless he’s in store for a season where he would be one of the worst players in hockey – only five players had a lower PDO in 2011-2012 with minimum 40 games played – then he has to improve from here on out. To what degree? That’s not possible to determine. But I highly doubt he ends up somewhere around -25 for the season. I would expect somewhere around zero, plus he should start to contribute offensively as well as a result.
The key to fantasy sports is also knowing when to sell high. When a player is producing at a completely unsustainable level, that is the time to try to find someone who will be willing to take him for a player greater than him. For those in fantasy baseball leagues, think of how many times you’ve seen a player go on a hot streak and an owner sell him for something better. If you managed to trade Josh Hamilton at the middle of the bsaeball season last year, you probably got someone back like Albert Pujols or Miguel Cabrera and got a great return on your trade. Fantasy hockey is no different.
Let’s take a look at Daniel Winnik (F-ANA). Winnik is currently sporting an 1153 PDO, good for second-best in the entire NHL for players who have played at least five games this year. This is despite the fact that over the last three years, Winnik has had a PDO of 972, 981 and 1012. This tells us that Winnik is due for a MAJOR regression, sooner rather than later. As a fantasy owner, if I had Winnik on my team, I would be looking to sell, sell, sell. More than likely, he was a waiver wire addition for most people. If you can turn him into a #2 or #3 defenceman, or a top-6 forward, then you have reaped incredible value. Not only did you get the most you possibly could from a WW addition, but you also turned it into a valuable fantasy player. This is how you win leagues.
There are several different applications of Corsi (named for Jim Corsi, current goaltending coach for the Buffalo Sabres), however the simplest one that is most widely used is called “On-Ice Corsi.” This is a metric that is used by some in place of the traditional plus/minus statistic. On-Ice Corsi tells us how good a team is at creating offence and possessing the puck when a certain player is on the ice. This is expressed as a shot-differential, but includes all the different varieties of shots.
On-Ice Corsi: (goals for + saves against + missed shots for+ blocked shots against) – (goals against + saves for + missed shots against + blocked shots for)
In short, it is the difference between the sum of all of MY team’s shots and the sum of all the OTHER team’s shots. I will make up an example to clarify:
Let’s say I’m Kris Letang, defenceman for Pittsburgh. In a single game, I was on the ice for: one goal for, nine saves against, four missed shots for and three blocked shots against. This gives us the first half of the equation; 1 + 9 + 4 + 3 = 17. I’m also on the ice for two goals against, five saves for, three missed shots against and two blocked shots for. This gives us the second half of the equation; 2 + 5 + 3 + 2 = 12. So the On-Ice Corsi equation gives us 17-12 = 5. My On-Ice Corsi for that game is +5. Any positive number is good.
Although in that scenario Letang would traditionally be a (-1) for the game, he had a (+5) On-Ice Corsi rating. This would indicate that the team generated more offence and controlled the play more often when he was on the ice and that the (-1), over the long run, would improve as long as his On-Ice Corsi rating stabilized.
It doesn’t take a genius to figure out the inherent issues with On-Ice Corsi. If you want a good read, look at this, written by an advanced stat guy that I look to and respect, Cam Charron. For those that don’t click through, I’ll explain.
On-Ice Corsi is completely player dependent. I’ll give you three guesses at who the On-Ice Corsi leader was from the 2011-2012 season among players who played at least 50 games… I’ll save time and just tell you: Alec Martinez (D-LAK). This is no knock against Martinez, he is a fine young defenceman who plays his role well. But I don’t think anyone would confuse him as one of the top defenceman, either on offence or defence, in the NHL.
Essentially, you see a lot of the same teams at the bottom and top of the On-Ice Corsi ratings. Last year, seven of the top ten leaders in On-Ice Corsi with at least 50 games played were from either the Los Angeles Kings or Boston Bruins. This is why, I believe in this sense, On-Ice Corsi does a better job at explaining a team’s overall ability to generate offence and possess the puck, rather than a specific player. The On-Ice Corsi of a player, as I’ve said, is dependent more on the team than the player themselves.
The application for fantasy hockey of On-Ice Corsi is fairly limited. While there are players in the top ten from last year I would want on my team, like Patrice Bergeron, Brad Marchand and Pavel Datsyuk, there are also players I would not even think of having on my team unless it was an insanely deep league (I’m thinking 400+ players rostered) like the aforementioned Martinez or his teammate Brad Richardson. For points-only leagues, On-Ice Corsi is almost useless. Of the top ten players in terms of On-Ice Corsi last year, only Anze Kopitar (F-LAK, 19.35 On-Ice Corsi, 76 points) finished in the top-20 in league scoring.
However, if you are in a standard rotisserie league, On-Ice Corsi can be an indicator of at least which teams you should target to draft from, if not individual players. Of the top-30 players last year in On-Ice Corsi, 19 of them came from just four teams. Not surprisingly, these teams were Los Angeles (6), Boston (5), Vancouver (4), Detroit (4). This bears out in the final traditional plus/minus standing at the end of the year; 14 of the top 24 players in terms of plus minus came from Boston, Vancouver or Detroit. There were only two players in the top-40 in +/- that came from L.A.; Willie Mitchell (T-25, +20) and Dustin Brown (T-34, +18). That’s what happens when you’re the second-lowest scoring team in the NHL.
As I just showed, On-Ice Corsi is difficult to apply to fantasy hockey. It is useless in points leagues, but if you’re looking for a boost in the +/- column with potential offensive upside, look for the teams of the players leading in On-Ice Corsi and not the individual players themselves.
Offensive and Defensive Zone Starts/Finishes
The final metric I want to talk about is O/D starts and finishes. How O-Zone start% apply to fantasy is clear, so is D-Zone start% (but more so and I’ll get into why). D-zone finishes aren’t as important as O-zone finishes and again, I’ll explain why.
The utility of offensive and defensive zone starts seems pretty obvious. Players that start in the offensive zone have a greater opportunity to produce offensively than the ones that start more often in the defensive zone.
The utility of offensive and defensive zone finishes is a little less obvious. When you talk about O/D starts, the reasons smack you in the face; you want your top offensive players to start in the O-zone with your top defensive players starting in the D-zone. But when you talk about O/D finishes, the reasons are less obvious. We know ‘why’ when talking O/D starts, but what is less clear is ‘why’ when talking O/D finishes.
But first let’s clarify everything for those that are new to the fantasy hockey circle. Offensive zone start% is the percentage of the time a given player starts their shift in the offensive zone e.g. on the face-off. The natural flow of hockey means that players who start less than half the time O-zone will finish in the defensive zone more often than they start in the offensive zone. This is because it is rare for a team to start and finish in the offensive zone, eventually the other team usually gets the puck out of the zone.
Defensive zone start% is exactly what it sounds like, the rate at which a player starts in the defensive zone. While this isn’t the best situation for a player to produce for your fantasy team, it still has its benefits that I will get into later.
So if you’re looking for players that get a more-than-usual opportunity to score, guys with high O-Zone start% is a good place to start. You see guys on the list of players so far this season who start on the O-zone greater than 70% of the time with names like Nash, Tarasenko, Malkin, Gaborik and Brad Richards. These are players being put in an optimal position to create offence. Also, you can assume their linemates won’t be too far behind, so keep that in mind as well. Don’t be fooled though; you will see a lot of fourth liners and #5-6 defencemen on that list as well. These players are not frequently put in defensive situations because quite often they are liabilities in this situation. Paul Bissonnette (F-PHX) starts in the O-zone over 75% of the time but no one would ever mistake him for a point producer.
A more interesting conversation is the relationship between defensive zone starts and offensive zone finishes. As I already said, the natural flow of hockey means a defensive zone start quite often leads to an offensive zone finish. But how or why this happens is what is critical.
I like to look for players that finish in the offensive zone more often than they start in the defensive zone for two reasons: Takeaways and Transition.
A player’s ability to take the puck away from another is a valuable skill. At the time of writing, of the top-11 players in the NHL in terms of takeaways, you find the following names among them: Erik Karlsson, Matt Duchene, Eric Staal, Pavel Datsyuk, Evgeni Malkin, Brad Marchand, Patrick Marleau, P.A. Parenteau and Thomas Vanek. Most of these players are among the leaders at their position for scoring so far this season.
Let’s look at 2010-2011 for a minute. Among rookie defencemen, Kevin Shattenkirk was the only player with at least 20 takeaways and less giveaways (41 takeaways, 23 giveaways). Is it shocking to anyone that he then, in turn, led NHL rookie defencemen with 43 points? That season, he started in the D-Zone 40% of the time.
In 2011-2012, Shattenkirk’s teammate Alex Pietrangelo was even more exceptional in offensive production. He had a +10 takeaway/giveaway ratio and finished with 51 points. He also started in the D-zone 47.2% of the time (finished in the O-Zone 53.3% of the time). His ability to take the puck away and transition the play factored in to his high point production and good plus minus.
Keep in mind, that this is not the be-all end-all. All I’m saying is that looking at a player’s ability to take the puck away with efficiency, if he has a high-ish D-Zone start% (40% or higher), can lead to scoring opportunities like 3 on 2s, or at the very least get the puck out of the zone and avoid a hit to the wrong side of the +/- column.
This section is a little bit more confusing than the other two, so I’ll simplify everything I just said in reference to O/D start/finish%
- The easiest thing to do is to look for players playing first or second line minutes that start frequently – > 55% of the time – in the offensive zone. These are the top players that are being put in optimal positions to produce.
- The second thing you should do is look at takeaway efficiency and defensive zone start%. This will give you an idea of the players who are able to transition the game quickly and can avoid, at the very least, having a bad plus/minus rating.
There are people much smarter than I currently developing more and more metrics for hockey. For now, most of these applications factor in to how a team is coached and constructed rather than fantasy production. However, there are indicators you can follow using things like PDO, On-Ice Corsi and O/D Start/Finish%.
Advanced-stats sites to visit:
BehindTheNet: This is where I get all my advanced statistics, including the ones in this article. Like I said, these stats are still in their infancy, so it only goes back to 2007-2008.
Hockey Analysis: Another great advanced-stat resource where blog posts are the norm. Get your thinking cap on.
Finally, there are many resources that track scoring chances for a particular team. Scoring chances are the best indicators of the quality of opportunities a team gets, so go ahead and look them up.