NCAA FBS Competitive Balance - Part III

Previously I have looked at competitive balance in the NCAA Football Bowl Subdivision (FBS) for all games played in a season (regular season FBS games, regular season Football Championship Subdivision (FCS) games and post-season bowl games), and then for just regular season (FBS and FCS) games.  Now what I want to do is just look at regular season games against only by FBS teams.  To do that I have deleted all the standings data associated with football opponents that at the time were not FBS teams.

To provide some perspective, I have included in the table below all three Noll-Scully competitive balance measures for the 2002 through 2012 NCAA FBS seasons.  You will notice that competitive balance is about 6% worse using only the regular season FBS standings data (column 3) as compared to the full standings data (column 1).  Given that the Big 10 and SEC conference have been talking about increasing the number of regular season conference games, I would expect that competitive balance would actually decrease by a small amount, as some teams that would have played an FCS school (and most likely won that game), will now have to play against a better opponent and have a lower winning percentage and thus lead to more competitive imbalance (higher Noll-Scully number).

Season		Full		Regular		Regular FBS
2002		1.539		1.527		1.651
2003		1.612		1.599		1.702
2004		1.462		1.458		1.549
2005		1.435		1.399		1.518
2006		1.579		1.560		1.633
2007		1.458		1.451		1.544
2008		1.458		1.466		1.588
2009		1.519		1.512		1.620
2010		1.526		1.508		1.581
2011		1.515		1.486		1.592
2012		1.579		1.561		1.698
Average		1.516		1.503		1.607

NCAA FBS Competitive Balance - Part II

Yesterday I blogged about competitive balance for NCAA FBS from 2002 to 2012, and reported that the NCAA FBS is very similar to the NFL.  I also gave a reason why different leagues of the same sport have similar levels of competitive balance, but different sports have different levels of competitive balance.  Today, I want to clean up one issue that I mentioned yesterday - that the data I was using had both regular season and post-season (bowl games) results in the standings data.  For comparison purposes, I will include the Noll-Scully competitive balance metric with the bowl games as well, which is the column labeled "Full" and just the regular season standings data labeled "Regular".

Season		Full		Regular
2002		1.539		1.527
2003		1.612		1.599
2004		1.462		1.458
2005		1.435		1.399
2006		1.579		1.560
2007		1.458		1.451
2008		1.458		1.466
2009		1.519		1.512
2010		1.526		1.508
2011		1.515		1.486
2012		1.579		1.561
Average		1.516		1.503

As you can see from the table above, the NCAA FBS "league" is a little more competitive without including the bowl game results.  (Remember, the closer the Noll-Scully gets to one the more that the league would have an equal playing strength in a statistical sense).  So, with the average over the time period moving from 1.516 to 1.503 over the last eleven years, competitive balance excluding the bowl games improves by less than one percent.

That still leaves the question about how much would the "leagues" competitive balance change if I excluded non-league games.  Non-league games would be games played against FCS schools.  I will work on figuring out which teams are FCS schools for the eleven seasons, and then drop them from the standings data and re-calculate the Noll-Scully competitive balance metric.  I hope to get to this tomorrow, but no guarantees.

How Balanced is NCAA FBS competition?

A question that I have been thinking about lately is how does the NCAA FBS compare in terms of competitiveness to other sports leagues?  In other words, we hear a lot about one conference having the best teams overall, but that would mean that other teams in that conference would have to be some of the worst, since they play a majority of their games each season against excellent conference rivals.  So how balanced is NCAA FBS football competition?

In our book, The Wages of Wins, we take a look at competitive balance using a measure devised by Roger Noll and Gerald Scully, which we call the Noll-Scully competitive balance measure.  Basically, what it measures is the actual standard deviation of winning percent in a sports league relative to a standard deviation of winning percent if wins and losses were randomly distributed, using the binomial distribution.  The closer a league gets to one the more balanced the league from a statistical viewpoint.

In The Wages of Wins, we report that for the NFL, from 1922-2006 that Noll-Scully competitive balance metric had an average of 1.56 and for the AFL (1960 - 1969) the Noll-Scully metric had an average of 1.58.  We also show that other types of sports have similar measures of competitive balance among different leagues.  One sport that we did not report was NCAA football, and here is a first pass at clearing up that oversight.

So, I got the NCAA football schedules from 2002 to 2012 and calculated each season's Noll-Scully measure of competitive balance.  The first is with all the games that were played in each season, which includes both bowl (post-season) games and games against non-FBS teams.  To be fair, when we calculated the Noll-Scully for the NFL and the AFL we did not include post-season games and there were no games against non-league opponents.  In the coming days, I will re-calculate the Noll-Scully deleting out the post-season games and then deleting out games against non-FBS teams and report each in a separate blog post.  As for now, here is the full sample over the last eleven years.
 

Season		NS
2002		1.539
2003		1.612
2004		1.462
2005		1.435
2006		1.579
2007		1.458
2008		1.458
2009		1.519
2010		1.526
2011		1.515
2012		1.579
Average		1.516

You will notice that the Noll-Scully does not change very much from one year to another and that the average over the 2002-2012 seasons is fairly similar to the NFL and AFL over the time periods reported in The Wages of Wins.  This fits with our extension of Gould's hypothesis that the underlying population playing a sport has an impact on it's level of competitive balance. (Gould was trying to explain the disappearance of the 0.400 hitter in MLB).  Our extension is that the increase in the underlying population of individuals playing a particular type of sport and having particular skills impacts the level of competitive balance in the sport.  So we argue that given the tiny population of people seven foot or taller makes basketball less competitive than a sport like soccer which does not rely on drawing highly skilled players from such a small population.  Given that soccer is the world's most played sport, it should be the most competitive and overall soccer has a lower Noll-Scully competitive balance measure than say hockey, baseball, or basketball.

2012-2013 NBA Competitive Balance

Now that the NBA finals are upon us, I thought that I would look back at the NBA regular season and calculate competitive balance using the Noll-Scully measure of competitive balance.  As Dave Berri has written previously and as we wrote in The Wages of Wins, the NBA compared to the NFL, MLB and NHL is rather competitively imbalanced.  During David Stern's first 27 years as NBA commissioner, the NBA averaged a Noll-Scully competitive balance of 2.8 for the seasons 1984-85 to 2010-11.  So looking back at the 2012-2013 NBA regular season, I have calculated the Noll-Scully equal to 2.81 (using the sample standard deviation of winning percent) and 2.76 (using the population standard deviation of winning percent).  In either case the NBA is similar in its measure of competitive balance as it has during the last three decades.

2013 NHL Goalie Performance

This past weekend I heard a number of hockey announcers state that the most important player in the playoffs was the goaltender.  That reminded me that I had not finished the regular season NHL goalie analysis.  For our purposes, the NHL goalie measure analyzes the number of wins above average a NHL goalie produces in a given time period (in this case the lockout shortened season). So here are the top 10 NHL goalies using our Wins Above Average measure of an NHL goalie.  The updated coefficient on goals against is -0.341814 over the seasons from 1995/96 to 2013.  Using that measure of the impact of goals against on standings points, yields the following top 20 NHL goalies during the 2013 regular season.

	Player	Team		WAA		SV%
1	Sergei Bobrovsky	CBJ		3.720		0.932
2	Craig Anderson	OTT		3.364		0.941
3	Henrik Lundqvist	NYR		2.863		0.926
4	Tuukka Rask	BOS		2.860		0.929
5	Antti Niemi	SJS		2.518		0.924
6	Cory Schneider	VAN		2.152		0.927
7	Jimmy Howard	DET		2.137		0.923
8	James Reimer	TOR		2.054		0.924
9	Corey Crawford	CHI		1.850		0.926
10	Robin Lehner	OTT		1.745		0.936
11	Devan Dubnyk	EDM		1.563		0.920
12	Braden Holtby	WSH		1.550		0.920
13	Viktor Fasth	ANA		1.025		0.921
14	Ben Bishop	OTT, TBL		0.968		0.920
15	Ray Emery	CHI		0.792		0.922
16	Jason LaBarbera	PHX		0.791		0.923
17	Chad Johnson	PHX		0.777		0.954
18	Kari Lehtonen	DAL		0.732		0.916
19	Nikolai Khabibulin	EDM		0.712		0.923
20	Ryan Miller	BUF		0.668		0.915

NHL Attendance Analysis

Now that the NHL regular season is over, let's take a look at NHL team attendance and see what effect (if any) the NHL lockout had on fan attendance as compared to the past few years.

I grabbed the NHL attendance data from ESPN from the 2000/2001 season to the current 2012/2013 NHL regular season, sorted the data and then ran a t-test (much like I did after the recent NBA lockout).  Here is the average home attendance from 2005-2006 to 2012 (or actually 2013) NHL regular seasons.  The pink line is the most recent NHL regular season and notice that even thought there were 17 fewer home games the average attendance for the teams is very similar to the average home regular season attendance in the previous non-lockout regular seasons.  In other words, there is not much change in attendance from 2011-2012 to this season.

 

Using more formal statistical analysis, I calculated the t-test for the last two NHL regular seasons and found that there is no statistical difference between the home regular season attendance in the 2011-2012 and 2013 seasons.  For those curious, the t-test was  0.318 using a two sample equal (and unequal) variance measure.

2013 NHL Pay and Performance

Today let's take a look at how well NHL (relative) team payroll and team regular season performance relate to each other.  I am interested in how payroll and team performance related for both just this (lock-out shortened) season and over a longer period of time, so I will present both below.

For NHL team payroll data, I usually use USA Today's payroll database, but USA Today does not have that data for the 2013 NHL season, so I found NHL team compensation data at the National Hockey Leagues Players Association (NHLPA) website, which I assume is very accurate.  Note this is total compensation, not compensation against the payroll (or salary) cap.  The NHLPA notes that their measure of team compensation "is comprised of base salary plus signing bonus for the current season".  Prior data comes from USA Today's NHL salary database, so there may be some differences in the two.  For NHL team performance data, I used the 2013 regular season standings data reported on  ESPN's website.  Prior regular seasons standings data come ESPN as well.  Here is the step-by-step details of how I calculated the relative payroll - team performance relationship.

So, for just the 2013 NHL season, running the numbers I find that the relationship between relative payroll and regular season performance is positive and statistically significant and results in that relative payroll "explains" 29.7% of NHL regular season performance, which is rather high as compared to the NFL.  Again by "explain" I mean that the amount of variation in relative payroll that is related to the amount of variation in regular season performance.  Even so that still leaves 70% of regular season performance not explained by payroll.  I will leave it to you to decide if that is a lot or a little.

Over a longer time period (2000-2001 to 2013) seasons (without the 2004-2005 cancelled season) we see that the relationship between relative payroll and regular season performance in the NHL is positive and statistically significant.  Relative payroll "explains" about 24.8% of NHL regular season performance, which is very similar to prior estimates, resulting in that relative payroll does not explain over three quarters of team performance.  As we contend in The Wages of Wins, the argument that team payrolls determine team regular season performance does not seem to be as big as some claim.