The Curious Case of Home Field Advantage – Part I

Everyone knows it’s good to play at home. It seems obvious, and the results in every major sport, including the NFL, confirm its importance. Home teams score more points, allow fewer, and win more often. It’s a little less clear why, although there’s no shortage of speculation: sleeping in your own bed; travel fatigue; crowd noise; biased officiatingstaying up all night chasing tail; and any other number of extracurricular activities.

All NFL teams benefit from playing at home, but some benefit more than others. As home wins continue to pile up in the NFL, I thought it might be interesting to examine some basic facts about home-field advantage (“HFA”) in the NFL in an attempt to seek some answers.

Using data from Pro-Football-Reference.com, we’ll look at 8,472 NFL regular season games from 1978-2013, and see what NFL HFA looks like. The basics:

 

N = 8472 Scored Difference Winning %
Home 22.26 2.74 57.81%
Road 19.52 -2.74 41.99%

(NOTE: Winning percentages don’t add to 100% because of ties.)

 

Nothing too surprising there. Home teams score 2.74 points more than road teams, and win substantially more often. However, those are just league-wide averages. Here are the individual team-level data, sorted by the gap between home/road point differential.

Team Home Pt Diff W% Road Pt Diff W% H/R Pt Split H/R W% Split
ARI -1.07 0.495 -8.89 0.274 7.82 0.22
KAN 4.16 0.604 -3.64 0.379 7.80 0.23
DEN 6.28 0.716 -1.12 0.473 7.40 0.24
BUF 2.79 0.579 -4.61 0.378 7.40 0.20
DET 0.79 0.493 -6.55 0.275 7.33 0.22
SEA 3.71 0.607 -3.19 0.399 6.90 0.21
BAL 3.28 0.590 -3.58 0.386 6.86 0.20
HOU 1.70 0.537 -5.11 0.369 6.82 0.17
PIT 6.64 0.700 0.02 0.495 6.62 0.20
STL 0.22 0.480 -6.28 0.354 6.49 0.13
CHI 3.88 0.623 -2.55 0.433 6.43 0.19
MIN 3.84 0.642 -2.53 0.396 6.37 0.25
GNB 5.32 0.648 -0.96 0.444 6.28 0.20
CIN 0.94 0.521 -4.84 0.335 5.78 0.19
JAX 2.03 0.559 -3.64 0.388 5.66 0.17
DAL 4.79 0.637 -0.83 0.472 5.62 0.17
SFO 6.43 0.648 0.93 0.518 5.50 0.13
ATL 0.51 0.544 -4.71 0.367 5.21 0.18
TAM -0.34 0.496 -5.50 0.317 5.16 0.18
IND 2.09 0.588 -2.85 0.469 4.94 0.12
MIA 4.11 0.636 -0.69 0.481 4.80 0.16
TEN 2.43 0.596 -2.24 0.493 4.67 0.10
OAK 0.00 0.478 -4.49 0.375 4.49 0.10
NYJ 1.47 0.514 -2.95 0.426 4.42 0.09
NWE 5.18 0.658 0.79 0.500 4.40 0.16
SDG 3.20 0.572 -1.20 0.432 4.39 0.14
NYG 2.34 0.575 -1.76 0.477 4.10 0.10
CLE -0.72 0.471 -4.63 0.360 3.91 0.11
RAI 3.23 0.640 -0.66 0.540 3.89 0.10
WAS 2.23 0.569 -1.33 0.446 3.56 0.12
CAR 1.18 0.520 -2.13 0.428 3.32 0.09
RAM 1.20 0.530 -1.89 0.455 3.09 0.08
NOR 1.38 0.507 -1.70 0.472 3.08 0.04
PHI 3.39 0.582 0.34 0.512 3.05 0.07

 

The chart above presents each team’s point differential at home and on the road, as well as their winning percentages. There’s good reason to think that point differential is more important than wins and losses in evaluating success in the NFL, but both are worth looking at.

A few things stand out:

1. Only three active teams have been at least .500 on the road since 1978 – the Patriots, the Eagles, and the 49ers. It is hard to win on the road – even the Steelers, Packers, and Broncos failed to reach .500 during that time.

2. There are some large gaps between teams both with respect to splits for point differential and for winning percentage – for instance, the Eagles have enjoyed less than half the home-field advantage of the Cardinals.

3. There’s only a small relationship between team quality and the size of your home-field edge. The correlation is only -0.15, indicating good teams have slightly smaller home-field advantages than do poor teams.

The one big caveat here is that this data isn’t stadium-specific; as such, the analysis reflects what cities have had the biggest home-field advantage rather than which stadiums. For example, if we think Seattle’s “12th man” is partly due to the acoustics of CenturyLink Field, then we may be missing that by including data from the Kingdome.

[infobox title='”Home” Field: A Note on the Data’]Since we’re examining the home-field advantages of cities, rather than specific stadiums, the game results has been grouped as such — even for franchises with tenures in multiple cities. “RAI” and “RAM” refer to the Los Angeles incarnations of the Raiders and Rams franchises respectively. Teams are otherwise grouped by city, such that Houston includes both the Oilers and Texans, St. Louis comprises the Cardinals and Rams, Baltimore covers the Colts and Ravens, and Cleveland accounts for both versions of the Browns. A potential factor is the Packers’ use of County Stadium in Milwaukee for several home games annually from 1953 through 1994, after which Lambeau Field resumed its role as their exclusive host stadium. In our data set, dating to 1978, there are 53 such games which, though held in a city two hours south of Green Bay, were clearly played before home crowds. The Packers went 33-19-1 in those games. In a similar arrangement, the Buffalo Bills played six regular-season games in Toronto’s Rogers Centre from 2008 through 2013, winning five. In the aftermath of Hurricane Katrina, which rendered the Superdome unusable, New Orleans split their eight 2005 “home” games between the Meadowlands (against the Giants in their stadium), Louisiana State University (Tiger Stadium in Baton Rouge), and the Alamodome (in San Antonio). The Saints won just once at these venues, beating the Bills in Texas. In addition, the home/road data has not been adjusted to account for neutral site games, including eight played in London between 2007 and 2013, and one in Mexico City in 2005.[/infobox]

Accordingly, here’s data for each team since they moved into their current stadium (Levi’s Stadium excluded, since we’re not using 2014 data yet):

 

Tm Since Home Pt Diff W% Road Pt Diff W% H/R Pt Split H/R W% Split
SEA 2002 7.72 0.716 -3.10 0.398 10.82 0.32
BAL 1998 8.46 0.742 -1.17 0.442 9.63 0.30
ARI 2006 2.88 0.643 -6.04 0.339 8.91 0.30
NYG 2010 3.83 0.583 -4.25 0.458 8.08 0.13
KAN 1972 4.16 0.604 -3.64 0.379 7.80 0.23
NYJ 2010 0.08 0.625 -7.46 0.292 7.54 0.33
STL 1995 1.15 0.493 -6.36 0.354 7.51 0.14
BUF 1973 2.79 0.579 -4.61 0.378 7.40 0.20
MIN 1982 4.36 0.641 -2.30 0.405 6.66 0.24
CHI 1926 3.88 0.623 -2.55 0.433 6.43 0.19
GNB 1957 5.32 0.648 -0.96 0.444 6.28 0.20
DET 2002 -1.95 0.398 -8.18 0.227 6.23 0.17
HOU 2002 0.61 0.500 -5.20 0.352 5.82 0.15
JAX 1995 2.44 0.576 -3.25 0.396 5.69 0.18
SFO 1960 6.43 0.648 0.93 0.518 5.50 0.13
DAL 2009 2.09 0.500 -3.25 0.438 5.34 0.06
PIT 2001 7.16 0.708 1.93 0.552 5.23 0.16
IND 2008 2.38 0.700 -2.53 0.500 4.90 0.20
ATL 1992 1.15 0.589 -3.58 0.399 4.73 0.19
CLE 1999 -2.90 0.393 -7.63 0.277 4.72 0.12
TEN 1999 1.96 0.571 -2.62 0.491 4.57 0.08
OAK 1966 0.00 0.478 -4.49 0.375 4.49 0.10
TAM 1998 1.88 0.550 -2.53 0.408 4.40 0.14
SDG 1967 3.20 0.572 -1.20 0.432 4.39 0.14
DEN 2001 4.16 0.646 -0.18 0.521 4.33 0.13
MIA 1987 2.33 0.587 -1.61 0.452 3.94 0.13
CIN 2000 0.80 0.548 -2.63 0.394 3.42 0.15
PHI 2003 4.10 0.550 0.89 0.563 3.21 -0.01
NOR 1975 1.38 0.507 -1.70 0.472 3.08 0.04
NWE 2002 11.42 0.864 8.38 0.705 3.05 0.16
CAR 1996 0.31 0.485 -2.20 0.434 2.51 0.05
WAS 1997 -1.09 0.477 -3.35 0.383 2.26 0.09

 

Seattle now rises to the top, more than a point better than the Ravens. Since 2002, when the Seahawks moved into CenturyLink, only the Ravens and the Patriots have had better results at home. (The Eagles have somehow managed to win less often at home than on the road, but given their home/road point differential split is still 3.21 points one assumes that’s just statistical noise).

It’s worth remembering what exactly we’re measuring here: the gap between how well a team plays at home and on the road. There’s nothing inherently good about a large home-field advantage (if the effect is anything other than statistical noise in the first place). The relationship between team quality and home-field advantage is effectively zero. Teams like the Patriots (spanning most of the Belichick-era) have played so well on the road that it would be almost impossible for them to display a large home-field advantage. The opposite is true of Seattle – their “12th man” effect is as much a result of poor play on the road as strong play at home.

I tried playing around with some pieces of data to see if anything correlated strongly with home-field advantage, testing against 89 different variables for each team. Among factors considered: game-time temperatures, attendance figures, offensive and defensive points per game, rushing and passing yards and attempts, touchdowns, sacks, and score by quarter. Some takeaways:

1. Sack yards lost in road games had the strongest effect (.49 correlation with home-field advantage), meaning teams like the Seahawks and Ravens lost more yards to sacks in road games than teams like Washington and the Panthers. This is indicative of the fact that a large home-field advantage is in some measurejust poor play on the road.

2. Attendance was not a strong predictor. Home attendance, even controlling for stadium size, was effectively meaningless. There was a -0.35 correlation between road attendance and home-field advantage (meaning teams like the Seahawks and Ravens had low road attendance). An obvious causal mechanism does not pop out however.

3. Fourth-quarter home point differential had a .43 correlation with home-field advantage overall. This is basically what we’d expect. Good home teams “rose to the occasion” in the fourth quarter at home much more often than they did on the road.

4. Other variables, such as weather, team play style (e.g. running vs. passing), and other elements of team quality (e.g. rushing yards per attempt and completion %) did not have strong relationships.

It’s worth noting that there is some real possibility for spurious correlations here. Even these extremely modest results (correlations below ~.6 are barely worth writing about) could be just flukes. This is the risk of testing 89 different variables. Even if there was no effect, we’d expect some results of equivalent size simply because of chance.

Finally, one more pivot point: Some (namely economist Toby Moscowitz and co-author Jon Wertheim) have suggested that home-field advantage is mostly or entirely a result of biased officiating. While there is good reason to be skeptical of this result, I thought it would be interesting to take a look at one of their key findings: that NFL home-field advantage dramatically decreased in size after instant replay was instituted in 1999. (The theory being biased officiating is less salient once officiating can be reviewed and errors corrected.).

As we covered above, overall home-field advantage during the 1978-2013 time frame was 2.74 points/game, resulting in a 57.8% home winning percentage. From 1978-1998 (before the current replay system), those numbers were 2.9 points per game and a 58.3% home winning percentage. Since 1999, the figures have indeed dropped ‒ to 2.6 points per game and a 57.1% home win rate. However, while there has been a definite change, it’s hard to read too much into it and attribute the entirety of the results to officiating. The last 14 years have also seen changes in NFL team air travel, curfews, scheduling, and other less obvious factors.

Overall, it’s hard to say how much we’ve really learned here, other than that home field is a persistent NFL effect that has a huge impact on winning and losing. The causes are elusive – the majority of the data fails to reliably predict anything about how strong a team’s home-field advantage is, and sample sizes cloud the few decent correlations we do have.

In Part II, we’ll look at what NFL point spreads and over/unders can teach us about home field, as well as look as some general gambling trends focused on home field, including the notorious home underdog phenomenon.

Leave a Reply

Your email address will not be published.