Put your analytical skills to the test...learn how to time this market!
-->

Projecting 2014-15 NHL Standings Using Fenwick & Corsi Possession Statistics


“Riches do not consist in the possession of treasures, but in the use made of them.” ~ Napoleon Bonaparte

We are at that point. You know, that point a decade from now when looking back you wonder when advanced hockey analytics first became widely accepted. There have been pockets of analysts dubbed basement dwelling nerds rather than the leaders, innovators and forerunners they are. But now we have a pocket in Toronto – the world’s largest hockey market – and in the front office no less.

Advanced statistics is the stenograph for what the eye sees. You can comment on how the team was outplayed, how the forecheck was ineffective or how the face off was always in our zone. But can you quantify this? Can an audit trail be developed that holds players, lines, pairings and/or teams accountable.

And, of course, you cannot overlook systems. I’m sure it is simply the inspiring pregame speeches that have any Mike Babcock or Claude Julien or Joel Quenneville team perennially atop the standings. It couldn’t be their understanding of analytics. It couldn’t be their belief that puck possession, at its most fundamental, wins hockey games. It couldn’t be. Or could it?

The Maple Leafs have their crack statistical staff at the ready. Will the information at their disposal translate into success on ice? Will they learn to adapt? Perhaps someday.

I contend that a puck possession game must become ingrained in the players before acceptance. And acceptance must always occur before success. As the 2014-15 NHL season gets underway, I have set out in the following analysis to prove that an aggregate of strong puck possession players will produce more points in the standings for their team. For those players who have not bought in – or who have only played under coaches who have yet to drink from the cup (proverbially, but also most likely, literally) – I have hypotheses for them too.

PUCK POSSESSION STATS

I figure I should ensure we are all on the same page before I rattle off terms like Fenwick, Corsi and 5-on-5 Down 1. I will go through a high level explanation of certain puck possession statistics below, but I recommend referring to Advanced Hockey Stats – An Introduction for a more fulsome dissection.

My interpretation of key puck possession statistics is below. The theory behind these statistics is that teams must be in possession of the puck in order to shoot on the net. Therefore, the more shots taken, and the fewer shots taken against you, are a strong indicator of how a player or team controls the game.

Fenwick For (FF) = On Ice Shots For + One Ice Missed Shots For
Fenwick Against (FA) = On Ice Shots Against + One Ice Missed Shots Against
FF % = FF ÷ (FF + FA)

Note that the Fenwick statistics are all “on ice”. Therefore, the player doesn’t need to be shooting to improve their Fenwick; they just need to be on the ice. In other words, the player is rewarded for simply contributing to the shot. Fenwick statistics can be attributed to an individual player or an entire team.

Next is Corsi, which is the same as Fenwick with blocked shots added to the equation.

Corsi For (CF) = On Ice Shots For + One Ice Missed Shots For + Blocked Shots For
Corsi Against (CA) = On Ice Shots Against + One Ice Missed Shots Against + Blocked Shots Against
CF % = CF ÷ (CF + CA)

Both Fenwick and Corsi are used as situational statistics. If you think about it, how relevant is assessing puck possession of a player while he’s on the powerplay or shorthanded. It wouldn’t be valuable to dock Jonathan Toews or Patrice Bergeron for killing penalties when the opposition has a significantly higher likelihood of possessing the puck. Consequently, puck possession stats are often measured separately in scenarios such as the following:

    • 5-on-5
    • 5-on-5 Up 1 or Down 1: With a one goal lead or conversely down one goal
    • 5-on-5 Close: When the game is within one goal in the first two periods or tied in the final period *
    • 5-on-5 Home or Road

Definition from stats.hockeyanalysis.com

There are also Fenwick and Corsi “relative” statistics. I have ignored these statistics for my analysis because this relative measure adjusts the base figures for the puck possession ability of the player’s teammates while he is on the ice. Consequently, the relative stat only applies when comparing one player to another on the same team. Once a player has been dealt, the relative metric is tossed out the window.

BUT WHICH PUCK POSSESSION STATISTIC IS MOST DIRECTLY LINKED TO SUCCESS?

I am out to prove that a team’s standing at the end of the season can be projected based on the individual puck possession statistics of their skaters. To begin, I studied team Fenwick and Corsi figures from various 5-on-5 scenarios over the past five NHL regular seasons and compared the figures to final point totals. Note that I extrapolated results of the lockout-shortened 2012-13 regular season over the standard 82 games.

I used linear regression to assess which statistic is most directly linked (i.e., most highly correlated) to success. Correlation is measured using an R2 factor that ranges from -1 to 1. Correlation is strongest as R2 approaches 1 and weakest as it approaches 0. A negative R2 is, in simplistic terms, uncorrelated.

The following graph shows the correlation between regular season point totals and the team’s 5-on-5 Fenwick For %. For reference, a Fenwick For score of 50% means that a team is shooting on or at the net as many times as their opposition is. Above 50% indicates greater puck possession; below 50% indicates less. The formula at the bottom of the graph is the origin of the regression line that best fits the historical results. This formula could, in effect, be used as a predictor for future results.

Bob -nhl -1


The 5-on-5 Fenwick results have a positive R2 equal to 0.2855 which means that Fenwick stats are correlated with regular season point totals. I wasn’t expecting perfect correlation (i.e., R2 = 1); therefore, the question now becomes how to assess Fenwick versus other measures – like Corsi, for example.

Bob -nhl -2


The R2 resulting from the comparison of Corsi to team point totals is less than with Fenwick. This indicates that Corsi is less correlated, or in other words, adding blocked shots to the equation does not improve the metric’s predictive ability.

The regression above determined that Fenwick is the better metric. But is pure 5-on-5 even strength the best on ice scenario to project 2014-15 standings? The table below shows R2 results for several 5-on-5 Fenwick scenarios. 

R2 Factors from the Correlation of Fenwick to Points
Over Past Five NHL Regular Seasons

On Ice Scenario

R2 Factors

5-on-5 – All

0.2855

5-on-5 – Up 1

0.2954

5-on-5 – Tied

0.3230

5-on-5 – Down 1

0.2956

5-on-5 – Close

0.3332

5-on-5 – Home

0.2777

5-on-5 – Road

0.2486


The “close” scenario had the highest R2 after running each scenario through the linear regression test thereby proving to be the best of the bunch for predicting the final standings.

Bob -nhl -3


As a reminder, 5-on-5 Close refers to situations where the play is 5-on-5 even strength when the game is within one goal in the first two periods or tied in the final period. It is under this scenario that one can best assess the true makeup of a team. With a lead, a team’s nature is likely to relax. Down one, it’s to press. With respect to home and road splits, matching lines has too much of an impact on the variables at play to be useful as a predictor over a balanced schedule split equally home and away.

Consequently, the linear regression formula to project every NHL team’s 2014-15 regular season point total, i.e., the final standings, is:

Projected Team Points = (249.1 x 5-on-5 Close Fenwick) – 32.6

But how can one arrive at the 5-on-5 Close Fenwick number for each team when so many variables have changed since last year? Coaches have been fired and hired. Free agent frenzy took place. And rookies have emerged from camp with roster spots.  Earlier, I contended that puck possession must become ingrained in the players. The next step is to analyze the 2014-15 rosters and the historical Fenwick data for each player in order to establish an estimate for each team’s 2014-15 5-on-5 Close Fenwick.

ESTIMATING 2014-15 TEAM FENWICK STATISTICS

The first challenge was arriving at each team’s 5-on-5 Close Fenwick projection when this season’s squad consists of returning players, rookies and those from several other teams. Consequently, I needed to convert each player’s individual Fenwick scores into a team score.

To begin, I used each team’s 2014-15 depth charts and assigned each player to forward lines (F1, F2, F3, F4) and defensive pairings (D1, D2, D3). I determined the average 5-on-5 Close time on ice (TOI) per game over the past three seasons for each line and pairing. The next step was to determine the percentage of 5-on-5 Close TOI allocated to each line and pairing and, eventually, to arrive at a % TOI Allocation for each player.

NHL Average 5-on-5 Close Time on Ice – Forwards from 2011-12 through 2013-14

Forward Line

Average TOI
5-on-5 Close

% TOI Allocated

% TOI Allocated To Line

% TOI Allocated To Player

F1

8:37

28.14%

16.89%

5.63%

F2

8:04

26.33%

15.80%

5.27%

F3

7:23

24.09%

14.45%

4.82%

F4

6:34

21.44%

12.86%

4.28%

Total

30:40

100.00%

60.00%

20.00%


The % TOI Allocated to Line is determined by multiplying the % TOI Allocated by 3/5 to represent the ratio of forwards on the ice in a 5-on-5 situation. Likewise, the % TOI Allocation to Player is determined by multiplying the % TOI Allocated to Line by 1/3 to represent the ratio of forwards on a line.

The table below presents the same calculations for defensive pairings. 

NHL Average 5-on-5 Close Time on Ice – Defensemen from 2011-12 through 2013-14

Defensive Pairing

Average TOI
5-on-5 Close

% TOI Allocated

% TOI Allocated To Line

% TOI Allocated To Player

D1

10:45

36.08%

14.43%

7.22%

D2

9:49

32.94%

13.18%

6.59%

D3

9:14

30.98%

12.39%

6.20%

Total

29:50

100.00%

40.00%

20.00%


Once the percentage TOI allocations were calculated, I took each player on a 2014-15 NHL depth chart and determined their average 5-on-5 Close Fenwick score over the past three regular seasons. If they played fewer than the past three seasons, I determined the average over their number of year’s played.

2014-15 rookies like Sam Reinhart and Jonathan Drouin created a unique problem. My solution was to determine the average 5-on-5 Close Fenwick for rookies over the past two seasons. It just so happened that the average rookie 5-on-5 Close Fenwick score was 50.0%.

To demonstrate how I built the team 5-on-5 Close Fenwick score, I will use the Buffalo Sabres 3rd line of Marcus Foligno, Drew Stafford and Sam Reinhart as an example. Each player’s three year average 5-on-5 Close Fenwick is as follows:

  • Marcus Foligno – 47.4%
  • Drew Stafford – 42.2%
  • Sam Reinhart – Not available since he is a rookie; therefore, I used 50.0%

Then, I multiplied each of these averages by the F3 allocation of 4.82% from the table above. This resulted in each player and the line contributing the following to Buffalo’s projected 5-on-5 Close Fenwick score:

  • Marcus Foligno – 47.4% x 4.82% = 2.3%
  • Drew Stafford – 42.2% x 4.82% = 2.0%
  • Sam Reinhart – 50.0% x 4.82% = 2.4%
  • Buffalo 3rd Line Total = 6.7%

I continued on with these calculations for the entire Sabres team and each forward line and defensive pairing contributed the 5-on-5 Close Fenwick scores summarized below. These scores were then summed to determine the team’s projection.

Buffalo Sabres
2014-15 Projected 5-on-5 Close Fenwick

Line/ Pairing

Player (3-year Average 5-on-5 Close Fenwick)

% TOI Allocated To Player

Contribution To Team 5-on-5 Close Fenwick

F1

Z. Girgensons (45.0)

M. Moulson (50.2)

B.Gionta (48.2)

5.63%

8.1%

F2

T. Ennis (44.0)

C. Hodgson (43.4)

C. Stewart (50.4)

5.27%

7.3%

F3

M. Foligno (47.4)

D. Stafford (42.2)

S. Reinhart (50.0)

4.82%

6.7%

F4

N. Deslauiers (41.2)

B. Flynn (43.0)

C. McCormick (46.0)

4.28%

5.6%

D1

T. Myers (44.6)

J. Gorges (50.6)

7.22%

6.9%

D2

A. Meszaros (47.9)

M. Pysyk (41.3)

6.59%

5.9%

D3

R. Ristolainen (37.9)

A. Benoit (49.7)

6.20%

5.4%

Total

45.8%


I proceeded to project the 2014-15 5-on-5 Close Fenwick scores for each team. Once determined, each projection was applied to the regression formula below to estimate each team’s total points for the season. 

Projected Team Points = (249.1 x 5-on-5 Close Fenwick) – 32.6

The table below summarizes each team’s point total and 5-on-5 Close Fenwick from the 2013-14 NHL season and my projections for this season. 

Projected 2014-15 5-on-5 Close Fenwick and Team Points
Compared to Actual 2013-14 Results

Team

2013-14
5-on-5 Close Fenwick

2013-14 Points

Projected 2014-15 5-on-5 Close Fenwick

Projected 2014-15 Points

Anaheim Ducks

50.0%

116

49.7%

91

Arizona Coyotes

49.8%

89

49.6%

91

Boston Bruins

54.1%

117

53.4%

100

Buffalo Sabres

40.9%

52

45.8%

82

Calgary Flames

47.3%

77

48.5%

88

Carolina Hurricanes

48.1%

83

47.4%

85

Chicago Blackhawks

55.2%

107

55.0%

104

Colorado Avalanche

46.8%

112

48.5%

88

Columbus Blue Jackets

50.7%

93

49.0%

89

Dallas Stars

51.8%

91

50.7%

94

Detroit Red Wings

51.2%

93

53.6%

101

Edmonton Oilers

43.6%

67

48.1%

87

Florida Panthers

49.4%

66

50.3%

93

Los Angeles Kings

56.7%

100

56.0%

107

Minnesota Wild

48.7%

98

48.4%

88

Montreal Canadiens

48.4%

100

49.4%

91

Nashville Predators

50.7%

88

50.1%

92

New Jersey Devils

53.9%

88

51.8%

96

New York Islanders

49.5%

79

51.2%

95

New York Rangers

53.6%

96

51.0%

94

Ottawa Senators

50.8%

88

50.7%

94

Philadelphia Flyers

48.2%

94

48.7%

89

Pittsburgh Penguins

50.2%

109

50.9%

94

San Jose Sharks

54.9%

111

52.4%

98

St. Louis Blues

53.2%

111

51.7%

96

Tampa Bay Lightning

51.7%

101

50.7%

94

Toronto Maple Leafs

41.6%

84

47.1%

85

Vancouver Canucks

52.2%

83

52.2%

97

Washington Capitals

47.5%

90

48.5%

88

Winnipeg Jets

49.7%

84

50.1%

92


In reviewing these projections, it’s important to note that the results do regress towards the mean. The projected point totals have much less variance than what will eventually transpire. Other than the linear regression model itself, one key contributing factor is the attrition of underperforming players (i.e., low 2013-14 Fenwick scores) into retirement, the minors or Europe and their replacement with promising rookies or those recently acquired through trades or free agency. Each season provides this type of rebalancing that isn’t fully measurable until all 82 games are contested.

PROJECTED 2014-15 NHL STANDINGS

I have reorganized the results above into two conference graphs to better analyze the projected success of each franchise. The position each team is projected to rank, and thereby qualify for playoff hockey, is the true assessment of each team’s success.

In the Eastern Conference, it’s not surprising to see Babcock and Julien’s respective puck possession teams from Detroit and Boston lead the way. The far end sees a Sabres team coming off as poor a puck possession season that there has been since Corsi and Fenwick have become statistics and a Maple Leafs team who had an astounding and unfavorable disparity between shots for and against in 2013-14.

Bob -nhl -4


The Maple Leafs and Sabres may reside at the bottom, but both teams seem to have improved their expected 5-on-5 Close Fenwick by 5-6%. Some of Toronto’s worst offenders were Tim Gleason, Dave Bolland and Nikolai Kulemin who are now scattered throughout other Eastern Conference organizations. Instead, Dave Nonis brought in role players like Roman Polak, Mike Santorelli and Daniel Winnik with 5-on-5 Close Fenwick scores averaging above 50% over the past few seasons.

Other Eastern Conference teams improved as well. The New York Islanders and Florida Panthers are assembling many more pieces from teams with heavy puck possession systems. Key acquisitions included the Islanders early season deal for Johnny Boychuk and the Panthers offseason free agent signing of veteran blueliner Willie Mitchell.

When examining who would and wouldn’t make the playoffs, you can’t help but notice the Devils, Islanders and Senators are in and last year’s Eastern Conference finalists, the Montreal Canadiens, are out. Other Eastern teams with playoff aspirations, but low 5-on-5 Close Fenwick scores, are the Columbus Blue Jackets, Philadelphia Flyers and Washington Capitals.

The Western Conference projections below are littered with the cream of the crop among the top six. The West is loaded with playoff caliber teams; which mean there are still strong franchises in Colorado and Minnesota finding themselves near the bottom when it comes to puck possession.

Bob -nhl -5


The Los Angeles Kings and Chicago Blackhawks have combined to win four of the last five Stanley Cups, so if Fenwick is truly a statistic correlated with success, it should be no surprise to see both these teams atop my projections. Likewise, the Edmonton Oilers bring up the rear and they have failed to qualify for the playoffs over the past eight seasons despite some of the best young talent from recent drafts.

Edmonton is among a handful of Western Conference teams that have attempted to address puck possession. The Oilers have acquired the likes of Mark Fayne and Benoit Pouliot. But I think the most notable movement is in the other direction. St. Louis has brought in players from teams that have historically poor puck possession. One of the new faces from the offseason includes Paul Stastny. But I’m using a three year average, so the Blues projections are still affected by Jordan Leopold and recent trade deadline deals for Jay Bouwmeester and Steve Ott.

The only surprising inclusions among the projected playoff teams are Nashville and Winnipeg. For the Predators, Seth Jones had a strong 5-on-5 Close Fenwick of 53.4% in his rookie year which is 5.4% higher than Shea Weber’s three year average. In addition, Nashville’s James Neal and Derek Roy have solid puck possession statistics in comparison to the rest of their new team.

The Jets, on the other hand, are loaded with an existing lineup that have underachieved despite high 5-on-5 Close Fenwick scores. Significant contributors such as Evander Kane, Andrew Ladd, Tobias Enstrom, Blake Wheeler and Michael Frolik have all averaged over 50% for the past three seasons which is not bad for a team that has failed to make the postseason since moving north from Atlanta.

At the end of the day, these projected standings are only estimates. However, over the past several NHL seasons there has been clear historical evidence linking Cup winners to puck possession. NHL organizations are paying attention to Fenwick and Corsi and making roster moves to address such needs. The data is worth its weight in gold and those in possession, and willing to implement, will reap the rewards.

Bob Sullivan writes periodically for SportingCharts.com and can be followed on Twitter at @mrbobsullivan.

NOTES

  • The puck possession data has been obtained from stats.hockeyanalysis.com.
  • The 2014-15 NHL depth charts are from STATS LLC as at October 11, 2014

 



Latest Articles


TEAM UP