clock menu more-arrow no yes mobile

Filed under:

Puck Possession Fraud: Your New Jersey Devils

With the Devils set to miss the playoffs for the third year running, CJ and I looked at shot differential metrics from another angle to try and see if it would shed any light on the possession fraud that has been the New Jersey Devils.

"Need some help here, Sunny"
"Need some help here, Sunny"
Ed Mulholland-USA TODAY Sports

Corsi (all shooting attempts). Fenwick (all unblocked shooting attempts). Shot differential. You’ve heard these terms on this site and elsewhere countless times. Score-adjusted Corsi, Fenwick Close, Time and Venue Adjusted Corsi among others are incremental improvements on the raw shooting attempt totals that the NHL records.  They tell us that teams that generally out-attempt their opposition will do better in the long run. However, that’s not always the case is it, my fellow New Jersey Devils fans?

Here at In Lou We Trust, we've witnessed a possession giant miss the playoffs the last two seasons, fire its coach and replace him with two co-coaches and the GM, and now the Devils have flat-lined and only Cory Schneider has kept this team from joining the Buffalo Tanks in the standings.  In 2012, the year the Devils made a run to the Stanley Cup Finals, they were in the top half of the league in 5v5 CF%. The following, lockout-shortened season saw the Devils rise to second-best in the league, just behind the Los Angeles Kings. Last season, the Devils fell one spot to 3rd overall, behind those same Los Angeles Kings and the Chicago Blackhawks.  Now, they're just above the bottom five in CF%. How the times have changed.

The Devils improved on their puck possession each of the prior three seasons, yet have fallen further and further down the standings. Why? Sure, some will point to abysmal goaltending by Martin Brodeur last season and an unbelievable run of failure in shootouts, but there’s another way to look at how the team has performed - which could give us a clue as to what's going wrong this season.

"Microstats," as they are commonly referred to, are the events tracked by people like Corey SznajderEmmanuel PerryJen Lute Costellamyself, and several others: Zone Exits, Zone Entries, Zone Entry/Exit Targeting, and Passing. In last year’s summer series on what I found out by examining the passing data from the Devils 2013 – 2014 season, I learned that how efficient each team was mattered more than the amount of shot attempts they put up. This was true for passing efficiency (SAGE, Shot Attempt Generation Efficiency), and overall team efficiency (SF/CF, Shots For/Corsi For, or Corsi Efficiency; SF/FF, Shots For/Fenwick For, or Fenwick Efficiency).

When I say Corsi and Fenwick Efficiency, what I'm simply referring to is the number of Corsi or Fenwick attempts a team requires to generate their total shots. So, if a team has twenty shots, thirty Fenwick events, and forty Corsi events, their Fenwick Efficiency would be (20/30) 66% and their Corsi Efficiency would be (20/40) 50%.

The idea here is that if a team is more likely to convert Corsi/Fenwick events into actual shots on goal, they simply won't need to attempt as many shots. If a team doesn't need to attempt as many shots, their possession numbers may look worse, but they ultimately achieved their goal of getting a shot on net. Why do I need a few missed and blocked shots if my team can create opportunities on net more effectively than their opponents?

The Devils, during the prior two seasons, have been one of the best teams in Corsi differential, or CF%. However, rather than thinking of Corsi as a proxy for possession by counting each shot attempt, perhaps we should think of it as measuring successes (shots) and failures (attempts that miss the net, are blocked, or deflected away). When I started thinking this way, I began to think Corsi is making teams like the Devils out to be more than they are.

Before we get to the big picture, let's take a look at each Devils skater. Using data from War on Ice, we can pull the on-ice Corsi, Fenwick, and Shot totals, both for and against the Devils, for each player in 5v5 situations. If a player was on the ice for fewer than 100 Corsi events, I did not include them. I then divided the shot totals into the Corsi and Fenwick numbers and came away with a Corsi Efficiency (or percent) For and a Corsi Efficiency Against. This tells us when each player is on the ice, the likelihood of the team generating a shot from an attempt, and the likelihood of the opposition generating a shot from an attempt as well. We'll start with the defense, as we always do.

Def_C%

These are arranged left to right in ascending Corsi% Against: players that restrict opposition efficiency the best are on the left and the worst are on the right. For the defense, only Adam Larsson has a positive difference in Corsi% For (C%F) and Corsi% Against (C%A). Put another way, when Larsson is on the ice, the Devils are 2.4% more likely for their shooting attempts to register on goal than their opponents. Apart from Larsson, Andy Greene was the only other blueliner close to breaking even in terms of C%F and C%A. Let's remove blocks and take a look at the Fenwick chart.

Def_F%

When we remove blocks, we still see Larsson and Greene as the best defensive duo, while maintaining respectable offensive efficiency numbers. Mark Fraser, Jon Merrill, and Bryce Salvador bring up the next three in terms of defensive Fenwick efficiency. The Devils are actually most efficient going forward when Merrill is on the ice when we remove blocks.  Larsson and Greene are the only Devils defensemen to have a net gain in terms of Fenwick Efficiency. On to the forwards!

Fwd_C%

When opposing teams attempt any sort of shot, it has the least amount of change of becoming a shot when Steve Bernier is on the ice. That's nuts. You also don't think of the next group of players as interfering with shot attempts: Tuomo Ruutu, Stephen Gionta, Scott Gomez. Only Gionta, Gomez, and Jordin Tootoo have a net Corsi% gain. Gionta and Tootoo are shoot-first players and take several long range shots that are harmless, so that makes sense; Gomez has the highest SAGE (Shot Attempt Generation Efficiency) on the team, so this makes sense.

It's alarming how efficient teams are when Patrik Elias is on the ice, or that Jaromir Jagr doesn't provide a net gain here. To be honest, this entire team is a mess right now, so I'm not surprised so many players are in the red here. Let's see how things look with Fenwick.

Fwd_F%

Removing blocks paints a better picture, most likely, based on how we think of these players. If Gomez is on the ice and the shot attempt is not blocked, it's likely getting through on goal due to his exceptional setup ability. Travis Zajac also does really well in that department.  Bernier stays in the top five; Jagr and Elias jump forward.  And Michael Ryder wonders why he's benched.

Okay, so let's look at the team picture. Below is a table that shows the net difference between Corsi and Fenwick efficiency for each team in the league since the start of the lockout-shortened season until today.

Team_C%F%_D

Now, before you all run to the comments to say this doesn't mean anything because of where some teams are in this table, remember why we're here: the Devils were one of the best puck-possession teams in the league the last two seasons, but other factors contributed to their downfall. This article is about exploring what some of those other factors might be and to answer why the Devils had such great possession numbers even when we never really felt like they were that dominant of a team.

What this table shows is that the Devils were 4.2% less efficient than their opponents on all shooting attempts, which was the second worst mark in the league ahead of only Ottawa, and 3.7% less efficient than their opponents when we remove blocks from shooting attempts, dead last in the league. This means that when teams played the Devils, they had an easier time getting their attempts on net compared to any other team. Conversely, the Devils had a more difficult time getting their own attempts on net: it was more difficult for them to do this than any other team. Talent? System failure? Random chance?

So what, right? Why does any of this matter? Two points: from a descriptive standpoint, this a more intuitive look into offensive and defensive evaluation. We know if a player is on the ice for more or less shots than their opponents, but how effective is the team at converting those attempts into shots, or limiting the opposition attempts from becoming shots? There must be something these players are doing when they are on the ice that enhances the team's offense and defense. The opposite is true as well: some of these players don't do enough to enhance either side of possession. Is it movement? Passing? Shooting? Zone Exits and Entries?

As I track the rest of the season, this is something I want to focus on: off the puck movement that increases the chance of a shot attempt becoming a shot for the Devils, and decreases the chance of this same event for the opposition. So, look for that in March as I gather data/video on some of these players and try to offer up an explanation for why the team is more efficient on offense and defense with certain players on the ice.

The second point is that teams that are more efficient seem to win more. Entering this season, my merry band of trackers and I have expanded upon our passing metrics while still looking at Corsi and Fenwick efficiency of teams alongside their Corsi and Fenwick differentials. Looking at only games that have ended in Regulation (since we’re only tracking 5v5 data), here is what the data shows after having tracked 227 games.

Pass_Efficiency_Chart

Now, these and all subsequent Corsi, Fenwick, and Shot totals you will see are the raw totals, unadjusted for score effects. It is the same with the passing data. We see SAGE outpacing a team’s Corsi and Fenwick Efficiency. We also see Corsi Efficiency (similar to last season with the Devils) outpacing Fenwick and Shot differentials by a decent margin. Naturally, I began to wonder if team efficiency is something I should pay more attention to.

So, in comes your fellow ILWT writer, CJ Turtoro. CJ has the mathematical background I wanted to check to see if my theory was correct—that how efficient teams were mattered more than their corsi/fenwick/shot differentials. CJ pulled data from each game since 2006 that finished in regulation or overtime and simply took the raw Corsi, Fenwick, Shot differentials, as well as the raw Corsi and Fenwick efficiencies. CJ will fill in the technical details below.

CJ: The goal was to see if Corsi and Fenwick efficiency were more strongly correlated to winning than raw Corsi and Fenwick totals. Using the same raw data used by AC Thomas and War on Ice to generate their stats, I compiled full sets of Corsi, Fenwick, Shot, Corsi Efficiency (C%) and Fenwick Efficiency (F%) for all games decided in Regulation or single OT.

Game outcome is a Boolean variable. This means that there are only two ultimate possibilities: a win or a loss. Therefore, we chose to analyze the winning percentages of the game winner of each category. As an example, let’s say the Devils play the Capitals, win every category except F%, and they also win the game. Since the winner of the first 4 categories (Devils) won the game, those four categories are awarded a win. Since the winner of F% (Capitals) lost the game, they are not awarded a win. For those who are sticklers for statistical correctness, I subsequently used a proportion-comparing z-test to verify that differences were statistically significant. The results are below:

C% Results

If we analyze game outcomes with regards to all situations – a total game summary – Corsi efficiency is far and away the best statistic with Fenwick efficiency in second. Corsi efficiency winners won almost 60% of the time, whereas all the raw statistics were under 50% and Fenwick efficiency was 53.8%. However, efficiency is affected by situation just like Corsi and Fenwick totals. As an example, Corsi efficiency is about 63% when playing a man down as opposed to 54% at evens. So we had to look at two commonly analyzed situations. The first qualifier is to filter out anything not happening 5 on 5. The second filter is to only analyze when the game is "close." This is defined as being within one goal in the first two periods and/or being tied thereafter.

The results for 5v5 were similarly encouraging. Curiously, all winning percentages went down but the Corsi and Fenwick efficiencies had maintained a similar gap over the raw stats. However, when the stats were filtered out to only analyze the "close" portions of the game, the raw stats all went up to around 56% and the efficiency stats fell to 53.3% for Corsi and 50.7% for Fenwick.

Considering the curious high winning percentages in the first two investigations, I decided to analyze the correlation directly between efficiency and goals for percentage (GF%). Regardless of what contortion of the relationship I ran, there was somewhere between 0% and 1% correlation between our efficiency statistics and goal differential for all three situations. Now why is there a statistically significant increase in winning percentage, but zero correlation with scoring? Your guess is as good as mine. I’m open to ideas.

The file with the results can be found here .

Ryan: Our findings raised more questions than answers, as CJ and I don't present this as an "Aha!" posting, but simply pursued something that piqued our interest. How efficient teams are when they pass the puck not only has a strong correlation to teams winning games, but I've noticed a pattern between this passing efficiency and how many goals a team will score and their shooting percentage. It was very surprising to see that the relationship did not carry over to overall efficiency? Could it be that non-passing events are more noise than value?

So, why is the relationship not as clear when we look at a team's overall efficiency? Why do Corsi and Fenwick efficiency outpace shot differential metrics in all situations, 5v5 situations, but take a back seat during close situations? Like CJ said, I don't have the answers, but I have a few thoughts.

If a team is more likely to score a goal or get a shot on net from successful passing, how efficient teams are in that phase of the game will have a stronger impact on things like shooting percentage, goals, and winning, than just their overall efficiency. Identifying phases of the game is where we'll continue to isolate what matters and turn it into a data point. Below is a chart looking at how efficient the Devils were last season and through the first 54 games of this season when generating offense through passes.

SAGE

And now here are the overall Corsi and Fenwick Efficiency metrics compared against their opposition for the same time period.

C%

F%

Despite their strong possession numbers last season, their inefficiencies at passing the puck and overall play likely counted more towards their poor finish in the standings. Moreover, while the Devils have been great at suppressing shots the past few seasons - though certainly not this current one - their opponents are more likely to generate a shot from any type of shot attempt than the Devils are. This is the takeaway here: despite their two seasons of strong possession numbers, the Devils had to work harder to get a shot on goal than their opposition on most nights.

Put another way: Is every other team more skilled at passing or shooting than the Devils? This season, I might say "yes," but it's likely a systemic problem the Devils have had since the start of the 2013 - 2014 when I began tracking this data. It's undoubtedly a bit of both--the lack of talent and poor systems play--that have reduced this team to its current state.

So, how can the Devils improve for future seasons? By becoming more efficient both offensively and defensively. Possession alone is not a guarantee of success. Possession and efficiency together could yield more success; and it might be a better way to evaluate teams. The Devils need to find phases of the game such as zone exits, zone entries, passing that players excel at and acquire them.  If the team has an analytics department, maybe it’s time they got off their tails and did something. If CJ and I can pull this together in a few weeks while working full time jobs, an NHL team with dedicated resources should be able to provide the detailed data to have management put together a more competitive lineup. This current Devils team just doesn’t cut it. At the time of this writing, I'd just witnessed the Devils get put through the ringer by the Edmonton Oilers of all teams.  This team, this front office, this entire organization needs to hit the reset button.

Conclusions: Again, this is not suggesting the death of shot differential metrics or anything of that sort. Running the numbers for score-adjusted totals would be the next step in this analysis. Also, once my group's passing data is up on War on Ice, we'll be able to look at pass-generated possession like we do for Corsi: how well do teams pass the puck when a certain player is on the ice? How well do they defend against passes? Ideally, we could look at on-ice efficiency - both passing and overall - with this data and the group at War on Ice.

What CJ and I hope this piece does is to illustrate that by peeling back the surface data of shooting attempts, there's still an overwhelming amount of data and analysis still yet to be collected and worked on that will provide for deeper, more meaningful discussions. It's great the NHL will be posting Corsi, Zone Starts, PDO, etc. on their site this month. Judging by the chatter coming out of the Ottawa Conference last weekend, the message is more focused on "Where do we go now?"  That answer will come as we continue to determine why exactly our favorite team is or is not successful.