The “Smartest” NHL Teams

Continuing our series on win efficiency (which is also part of our class on predictive sports analytics) we now turn to the NHL.  We offer the NHL rankings largely without comment.  At the top we have the Red Wings, Sharks and Canucks.  At the bottom it’s Edmonton, Florida and Columbus.  Okay, maybe one comment.  Based on all of our NHL analyses over the last year, I think we can say that the best place to be a hockey fan is Detroit.

Rank

Team

1

Detroit

2

San Jose

3

Vancouver

4

Ottawa

5

Boston

6

Nashville

7

New Jersey

8

Anaheim

9

Pittsburgh

10

Minnesota

11

Chicago

12

Dallas

13

Montreal

14

St. Louis

15

Buffalo

16

Arizona

17

Washington

18

Colorado

19

Philadelphia

20

NY Islanders

21

Tampa Bay

22

Calgary

23

Los Angeles

24

Carolina

25

Winnipeg

26

Toronto

27

NY Rangers

28

Edmonton

29

Florida

30

Columbus

 

The “Smartest” MLB Teams

We had a couple of reasons for the previous post regarding the relationship between spending and winning across leagues.   The post was intended as a discussion point for our class – “Predictive Sports Analytics”.  But, the post was also background for our next series of analyses that focus on identifying the “smartest” or most efficient spenders in each of the four major professional leagues.

Today, we take a look at Major League Baseball.  The analysis itself is fairly simple.  We use the last 12 years of data on spending and winning to estimate a linear regression model.  We also include fixed effects for each team in this equation.  As a side note, an examination of residual trends might be a little better as management teams tend to change over time.

The fixed effects or team level intercepts provide an indication of whether teams over or under perform relative to what they invest.  The analysis is kept simple since it is for class, but it could easily be extended to include non-linear effects, or perhaps the dependent variable could be post-season qualification models as a binary logit.

The results of our analysis are both as expected in many ways and surprising in others.  At the top of the list we have the early adopter of analytics, the Oakland A’s.  We will take this as evidence of the value of investing in analytical capabilities.  At number two on the list we have the Atlanta Braves.  In positions three through five we have the Cardinals, Yankees and Giants.  The Yankees came as a bit of a surprise given their enormous payrolls.

At the bottom of the list we have Baltimore (note that this current season data is not used).  To some extent, the simplicity of the model might be unfairly penalizing Baltimore.  We would explain why, but this seems like a good question for class!  The other notable but unsurprising member of the bottom five is the Chicago Cubs.  This pains Professor Lewis too much to write about.

1

Oakland

2

Atlanta

3

St Louis

4

NY Yankees

5

San Francisco

6

Boston

7

LA Angels

8

Philadelphia

9

Cleveland

10

Minnesota

11

Chicago White Sox

12

Texas

13

Cincinnati

14

LA Dodgers

15

Toronto

16

Arizona

17

San Diego

18

Miami

19

Houston

20

Tampa Bay

21

Milwaukee

22

Seattle

23

NY Mets

24

Washington

25

Colorado

26

Detroit

27

Chicago Cubs

28

Pittsburgh

29

Kansas City

30

Baltimore

 Mike Lewis & Manish Tripathi, Emory University, 2014.

 

A Quick Look at the Relationship between Payroll & Winning Across Leagues

Note: This post is related to work currently being done in a class called “Predictive Sports Analytics” at Emory University.

This week, we are taking a (simple) look at the relationship between payrolls and winning across the four major professional sports league.  We begin today with a quick look at the strength of the payroll-winning relationship across the leagues.

The payroll variable that we use in this analysis is relative payroll.  To construct this variable, we simply divide each team’s payroll by the league average that year.  For example, if a team has double the league average the relative payroll measure would be 2, and if they had a payroll of exactly the league average, the variable would equal 1.

Please note that we did say this is a simple look.  The analysis could be extended in a variety of ways (such as using a division average as the denominator).  But, we are using this analysis as a discussion point for our Predictive Sports Analytics class here at Emory, and we want to keep the analysis clean.

So, let’s compare some basic descriptive statistics across leagues.  The tables below list measures of payroll and winning rate dispersion across leagues (we are doing this analysis using data from 2001 to 2013).  The first table includes data on relative payrolls.  The standard deviation (std. dev.) is a measure of spread in which higher values indicate greater variation and lower values indicate a tighter range.

The league with the greatest dispersion (variance) in payrolls is MLB.  Baseball is followed by the NHL and the NBA.  The NFL has the smallest amount of variability.  These results are not surprising given the differences in collective bargaining agreements, and revenue sharing policies across leagues.

Payroll Dispersion

League

Std. Dev.

Minimum

Maximum

MLB

.408

.19

2.85

NBA

.195

.42

1.82

NHL

.238

.13

1.80

NFL

.126

.59

1.46

The next table shows the variability in winning rates across leagues.  This is potentially important information for league commissioners and media companies.  It is unclear what the right level of competitive balance is for generating maximum fan interest.  But, at least at some level, fans probably want to see more even match-ups, and small market residents want their team to have a chance to be a winner.

Interestingly, we find the greatest dispersion in winning rates in the NFL – the league known for parity.  This is likely due to the fact that we are studying one year winning rates, and the NFL has a relatively short season.  The league with the most balance is MLB – the league with the least revenue sharing, and the most concerns about large market dominance.

Winning Percentage Dispersion

League

Std. Dev.

Minimum

Maximum

MLB

.073

.265

.716

NBA

.151

.106

.817

NHL

.099

.292

.837

NFL

.193

0

1

This brings us to the real reason for today’s post: A look at the “strength” of the relationship between spending and winning across leagues.  This is done by computing the correlation between relative payroll and winning, and also by estimating a linear regression of winning as a function of relative payroll.  The table below contains the results.

League

Correlation

Equation

MLB

.44

Win% = .42 + .08 * Rel_Pay

NBA

.32

Win% = .25 + .25 * Rel_Pay

NHL

.45

Win% = .35 + .20 * Rel_Pay

NFL

.17

Win% = .25 + .25 * Rel_Pay

We see that the relationship between winning and payroll is the strongest in MLB and the NHL.  While the “hard” salary cap leagues have the weakest relationship.

One more note: this is just the start.  As the weak progresses, we are going to examine the most efficient and least efficient teams in each league.  In other words, we are going to see what front offices get the most bang for their buck and who struggles.  Stay tuned. 

Mike Lewis & Manish Tripathi, Emory University 2014.

BBC Mundo: Por qué en EE.UU. no hay publicidad en las camisetas de los deportistas

BBC Mundo: Por qué en EE.UU. no hay publicidad en las camisetas de los deportistas

“Los fanáticos lo aceptan pero hay algo así como un chiste cultural. Ves a los pilotos todos llenos de logos. No creo que los deportes más tradicionales quieran compararse con la hipercomercialización de la Nascar”, comentó Michael Lewis, profesor de Mercadeo de la Universidad Emory.

Too Gay to Play? A Quantitative Analysis of Michael Sam’s Pre-Season Performance

The following is a guest post by Dr. Thomas Smith of the Goizueta Business School, Emory University.

Michael Sam, the first openly gay player drafted into the NFL, was cut by the St. Louis Rams this past weekend. Several analysts, including Mike Freeman of Bleacher Report, identified that cutting a player with performance stats similar to those of Sam is “almost unprecedented… and basically unheard of .” (Freeman’s article)  We know Sam had 11 combined tackles and 3 sacks, but are these the only way to measure Sam’s performance?  Let’s look at the stats of the defensive ends that made the cut this past weekend:

Table 1: Statistics of DEs Making the 53-Man Roster (2014 Pre-Season)

Table1 Source: www.NFL.com; www.espn.go.com (players on IR are not included in this table)

Table 2: Statistics of Michael Sam

Table2

Source: www.NFL.comwww.espn.go.com 

According to statistics, the DEs that made the cut had an average of 8 combined tackles and a little less than 1 sack. Sam’s stats better these marks by quite a bit. If these are the only statistics that matter, then the Rams (and most every other team) are missing out. However, if other performance metrics are considered – say those that are captured in the annual NFL Combine, then Sam was a little behind the ball. Sam is 0.14 seconds slower in the 40 yard dash, nearly 0.50 seconds slower in the 3-Cone drill, has a lower vertical leap, smaller broad jump and has a little less upper body strength than his counterparts.

But who is to say which statistic is the ‘silver bullet’, so to speak? If I look at the 40-yard dash and 3-cone performance then Jackson Jeffcoat (4.63 and 6.97) is the standout performer – yet didn’t make the Seahawks. Why? The Seahawks do not need him on defense – they are amazing without him.  Is it possible that Sam didn’t make the Rams because they didn’t need him – they were covered at DE? Absolutely.

Is it unprecedented that a player with Sam’s statistics on tackles and sacks gets left off a team (including the practice squad)? Looking back over the past four seasons (2010 – 2013) it appears that Mr. Freeman is on to something.  Looking at only the defensive ends with combined tackles within 1 and/or sacks within 1 of Michael Sam’s 2014 pre-season performance you see the following:

Table 3: DEs with Pre-Season Statistics Similar to Michael Sam (2010-2013)

Table 3

Source: www.NFL.com

So, Freeman knows his material. In the last four years, players with tackles and sacks similar to Michael Sam end up on a team (or practice squad in the case of Matt Broha). But, does this mean that the NFL is homophobic? No doubt people in the NFL are. However, this doesn’t prove that he’s not on the Rams because the Rams are homophobic. Several of the players listed above were picked up as free agents – they didn’t all play for the team that drafted them. And, at the moment, rumors are swirling that the Cowboys are interested in Sam for their practice squad (Note: he has since been signed by the Cowboys).

What does this mean? There is more than one statistic out there – we have seen plenty of players with better (and worse) statistics (all statistics) relative to Sam both make the NFL and get cut.

Thomas More Smith, Emory University, 2014.