SportsNerd: How well does statistical performance predict winning percentage?

Updated: November 29, 2012

P-ing on the NFL

Welcome to yet another installment of SportsNerd. Now that all of the PCA results have been presented, it’s time to examine the overall regression results to see how well all of the principal components together explain regular season winning percentage over the past ten years. The results will be presented in two ways:

1) The overall regression results for the ten year time span.

2) Total weighted principal component (TPCA) scores vs. winning percentage on a year-by-year basis.

The TPCA scores are each team’s values for every individual PCA multiplied by their standardized slopes as determined in the initial regression analysis. These cumulative scores rank teams on the basis of offense, defense, and special teams combined, taking into account variation in all factors over the past decade.

Overall Regression Results

The initial regression analysis included all twelve PCs, but KPPC2, the PC for kickoff variables, was not a significant factor in explaining winning percentage. Because of its non-effect, KPPC2 was dropped and a second regression analysis was run with the remaining eleven PCs. Remember that although KPPC2 contained all of the kickoff variables, number of kickoffs was also contained in KPPC1, that otherwise explained punting variability. Thus, number of kickoffs remains in the regression while variables pertaining to kickoff yards, etc., do not.

The eleven remaining PCs together significantly (P <0.01) explained 75% of the variation in regular season winning percentage (r2 = 0.75). The overall regression had an intercept of 0.500, which is what we should expect–the most average of teams is expected to win half its games. Individually, each of the remaining eleven PCs also significantly explained variation in winning percentage (for each P <0.01) and their standardized slopes are in Table 1 below.

Table 1

OPC1: Passing O 0.640 < 0.01
OPC2: Rushing O 0.293 < 0.01
OPC3: O Penalties -0.268 < 0.01
DPC1: Smashmouth D -0.307 < 0.01
DPC2: Passing D -0.242 < 0.01
DPC3: Interceptions 0.246 < 0.01
DPC4: Aggressive D 0.262 < 0.01
PKPC1: Punting 0.182 < 0.01
FGPC1: Field Goals 0.127 < 0.01
RETPC1: Punt Returns 0.115 < 0.01
RETPC2: KO Returns 0.089 < 0.01

From these slopes it is clear that passing offense has been, over the past ten years, the relatively most important factor in winning during the regular season. The effects of the remaining offensive and defensive variable are about half as strong, whereas the effects of special teams are less important.

Using these standardized slopes, total weighted principal components (TPCA) scores were calculated for all 320 teams  against which to plot winning percentage. This provides a visual representation of how these numbers vary and where teams stand relative to one another. That’s right, you are about to be overwhelmed by scatter plots!

Figure 1

What this plot tells us is that for every 1 point increase in TPCA score (within a range of scores from -2.3 to 2.3), one can expect a 0.200 increase in winning percentage, or roughly 3 additional wins per season. Meanwhile, the average team with a TPCA of 0 can expect to go 8-8. So, an above average team with a TPCA score of 1 should win 11 games, whereas a sub-standard team with a TPCA score of -1 should win just 5 games. The best team on this plot with the highest TPCA score is the 2007 New England Patriots. The worst: the winless 2008 Detriot Lions. Table2 contains the best twenty teams and Table 3 contains the worst ten teams for TPCA over the past ten years.

Table 2

TPCA Score
Win Pct
1 New England Patriots 2007 2.232 1.000 SB
2 Green Bay Packers 2011 1.816 0.938 P
3 Philadelphia Eagles 2002 1.806 0.750 CC
4 New Orleans Saints 2011 1.806 0.813 P
5 Kansas City Chiefs 2003 1.765 0.813 P
6 San Diego Chargers 2006 1.739 0.875 P
7 New Orleans Saints 2009 1.616 0.813 SBW
8 Baltimore Ravens 2006 1.580 0.813 P
9 Minnesota Vikings 2009 1.446 0.750 CC
10 Seattle Seahawks 2005 1.445 0.813 SB
11 Green Bay Packers 2003 1.441 0.625 P
12 Denver Broncos 2005 1.421 0.813 CC
13 Baltimore Ravens 2008 1.367 0.688 CC
14 Seattle Seahawks 2007 1.360 0.625 P
15 New Orleans Saints 2006 1.349 0.625 CC
16 St. Louis Rams 2003 1.344 0.750 P
17 Pittsburgh Steelers 2004 1.316 0.938 CC
18 Pittsburgh Steelers 2010 1.308 0.750 SB
19 San Diego Chargers 2010 1.260 0.563 NP
20 Chicago Bears 2006 1.222 0.813 SB

Table 3

TPCA Score
Win Pct
320 Detroit Lions 2008 -2.316 0.000
319 Oakland Raiders 2009 -2.146 0.313
318 St. Louis Rams 2009 -1.985 0.063
317 San Francisco 49ers 2005 -1.945 0.250
316 Indianapolis Colts 2011 -1.802 0.125
315 Cleveland Browns 2008 -1.801 0.250
314 Houston Texans 2005 -1.725 0.125
313 Tampa Bay Buccaneers 2011 -1.718 0.250
312 Tampa Bay Buccaneers 2009 -1.648 0.188
311 St. Louis Rams 2008 -1.595 0.125

You’ll notice in Table 2 that with the exception of the 2010 Chargers, all of the top twenty teams at least made the playoffs, and many of them made it at least as far as their respective conference championships. Only three made it to the Super Bowl, however, and of those only one actually managed to win it all. So, while TPCA does a great job of explaining variation in regular season winning percentage and can pretty reliably predict playoff berths, it does not predict Super Bowl winners very well.

Year-by-Year Regular Season Winning Percentage vs. TPCA

To provide an idea of how things work out within any given year, the following ten figure plot winning percentage against TPCA scores within each of the past ten years. Note that in all of these plots, the slopes and intercepts  are very close to the ten-year overall slope of 0.19 and intercept of 0.500, meaning that the relationship between TPCA and winning percentage is remarkably stable. Additionally, although values range from 68-80%, the amount of variation explained within each year is similar to the overall value of 75%.

For 2011, all of the playoff teams are identified as are the Super Bowl winner and loser. For the remaining years, only the Super Bowl winner is marked with a red circle.

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

What becomes apparent from these plots is that rarely does the best team statistically during the regular season end up winning the Super Bowl. Only once in the past ten years has this happened, when the New Orleans Saints won in 2009. On the other hand, a team can be categorically middle of the road during the regular season and manage to win.

This is the case with the New York Giants in both their recent Super Bowl wins. In fact, as Table 4 demonstrates, the Giants of 2007 and 2011 are the worst two regular season teams to win Super Bowls in the last ten years. As a Packers fan, I’m not sure if that should come as relief or should sink me into an even deeper depression with regard to the Giants.

Table 4

Super Bowl
Season OPC Rank
Season DPC Rank
Season TPCA Rank
XLVI New York Giants 12 21 17
XLV Green Bay Packers 10 3 5
XLIV New Orleans Saints 1 1 1
XLIII Pittsburgh Steelers 23 5 7
XLII New York Giants 11 16 11
XLI Indianapolis Colts 3 24 10
XL Pittsburgh Steelers 7 15 6
XXXIX New England Patriots 7 14 2
XXXVIII New England Patriots 20 2 9
XXXVII Tampa Bay Bucanneers 26 2 3

What we see here is that regular season prowess and winning percentage are critical to making the playoffs, but are no guarantee of playoff performance. This calls for a playoff-specific analysis similar to the one performed on regular season statistics, which may be presented at a later date.

In the meantime, the next installment of SportsNerd will continue looking at regression analysis, examining which of the PCs are significant predictors of winning percentage within years, and how those patterns change from year to year.

Note on the title: “SportsNerd” is a regular column that will look at trends in the NFL using fancy-pants statistical methods. The coy, if vulgar, “P-ing on the NFL” part refers to the p-value, which is nearly ubiquitous in statistical analysis. The “p” refers, roughly, to the probability of obtaining a result as strong as the one observed if the null hypothesis is true.

If that just made your ears bleed, good: statistics, like football, is supposed to hurt a little bit.

About the author

SportsNerd columnist Co-winner of the prestigious Water Cooler Sports Award for Best Statistical Study, Andy Froehle is proud to call the Green Bay Packers his team (literally: he has a worthless piece of paper that says so!). Andy uses statistical analysis extensively in his work and is interested in applying the same kinds of techniques to analyzing football data, with the chief goal of making your math-hole bleed a little bit. Feel free to contact him with any questions, comments, rants, raves, suggestions for future analyses at He also accepts donations in cash or in kind.


Um... 2246 pts

 EvilDonkey  jwoude23 yeah, I have no feelings for her for a long time. Still find her attractive though. But in a primal way. I keep that shit under wraps . She does like to flash me her bush now and the which is weird.



J. Walter Weatherman
J. Walter Weatherman 1 Like

You got something ELSE broken now Nate?  You need to pay that bookie back, stat!

Um... moderator

  See, I used to hurt myself getting really angry. this was a few years back. The anger management worked really well but now my body, which got used to the abuse, finds ways to break itself without my permission. 

Maized and Confused
Maized and Confused


Now it's just emotional scars from being a Bears fan?

Um... moderator

    The Bears have been pretty good to me as a fan. at least in recent years. The Cubs now...

ReggiesNoJackpondavre moderator 4 Like

so Motley Crue is giving away a bunch of signed stuff to the first person who posts a screenshot of their facebook page when it hits 2 million...



do they not know that ive been trained in the art of quick draw screenshots on this website?? xD

adambballn 4 Like

   A Tommy Lee penis pump?


    if it'll get it as big as his who wouldn't want it? just bleach the shit out of it first.

Preparation_A 1 Like


I wouldn't touch that thing with Bea Arthur's dick.


     I wish I could steer a boat with mine...


So a fight broke out between two chicks at Camacho's wake today. (the boxer who was recently killed) not a big surprise

Gabes22 moderator 1 Like

  Bitches, they do be crazy.


  i am assuming they were his sex slaves and they fought over who he loved the most. I am getting flashbacks from a cheers episode

Gabes22 moderator 4 Like

Because that's what being friends means, right hun? We don't fuck.  So I just get to be your emotional tampon but we don't fuck. That sound about right.

Nate, you are a Kinison fan, heed his advice!!!

Um... moderator

  I love Kinison. I am certainly not my exes emotional tampon. I have told her before she has to use her current men for that shit.

EvilDonkey moderator 4 Like

  truer words have never been spoken




Um... moderator

My ex wife, who is recently married, just called me up on my broken phone and asked if I wanted to go sit in a hot tub with her.

Bro Montana
Bro Montana

  Do it! And never call her back.

Preparation_A 2 Like


They have kids together. The no call back is unlikely to work.

Preparation_A 1 Like



Well, they may have to talk to each other about their children occasionally.



They have kids together because they humped without protection.

Big Cheese
Big Cheese 1 Like


You with your bad advice. Stop it.

Bro Montana
Bro Montana 1 Like

    I give great advice. 

Big Cheese
Big Cheese


Pardon granted. You're welcome.

jwoude23 Bear Down
jwoude23 Bear Down moderator

  "Sit in a hot tub?"  So that's what they call it these days!


Also, take this as a compliment.  It means you were way better in bed than her new hubby, and she already misses you.  

Um... moderator

  this is her third man since we broke up 7 years ago.

Gabes22 moderator 2 Like

        I was thinking a couple of these.

adambballn 2 Like

        She needs this...


Um... moderator

      She's married. It does not get much more convenient or effortless than that. In fact, I think she is knocked up right now.

Gabes22 moderator

      Translation: She has nobody and wants to get laid with as minimal effort as possible.

EvilDonkey moderator 5 Like

    Next time just point at it and say: "our kids came out of that."

Um... moderator

    yeah, I have no feelings for her for a long time. Still find her attractive though. But in a primal way. I keep that shit under wraps . She does like to flash me her bush now and the which is weird.

EvilDonkey moderator

    She will only use you to feel pretty. Go into it with selfish reasons. Don't get feelings.

Um... moderator 2 Like

  It would be funny if she forgot me. We were married 10 years and have 2 kids.

jwoude23 Bear Down
jwoude23 Bear Down moderator 1 Like

  yet she still remembers you.  You're the best she's had!!!

Big Cheese
Big Cheese


I assume you're sharing this for the comedic value and not actually considering it.