Home Field Advantage Part II: What’s the Weather?

The successful home underdog strategy in NFL lines helped popularize the use of data analysis in gambling and predictions. In Part I, we looked at basic tenets of home field advantage with a focus on wins, losses, and point differential. As one might expect, Seattle came out on top, performing almost 11 points better at home than on the road, with Baltimore and Arizona trailing close behind. The time, we will look deeper into home-field advantage, with a focus on the weather.

Home Underdogs

We’ll start with the famous home underdog phenomenon, popularized by  author Steven Levitt. He noted home underdogs have historically covered against the spread (“ATS”) with enough frequency that a bettor could beat the vig (a sportsbook’s minimum commission on each wager) simply by blindly betting every home underdog. Levitt’s data covered the 1980-2001 time period, so let’s update and expand that a bit to cover 1978-2013.

Below is a breakdown of activity from that period showing the number of games, average points scored, points allowed, winning percentage, record ATS, and how often the game went Over or Under the over/under total. Pushes have been removed. As a reminder, a bettor needs to win approximately 52.4% of the time to beat a standard -110 vig.

All Games	#	Scored	Allowed	W %	ATS	Over	Under
Home Favorites	5611	23.87	17.82	67.76%	48.55%	50.51%	49.49%
Road Favorites	2723	23.02	19.08	62.32%	47.37%	52.53%	47.47%
Home Underdogs	2723	19.08	23.02	37.50%	52.63%	52.53%	47.47%
Road Underdogs	5611	17.82	23.87	32.03%	51.45%	50.51%	49.49%

Home underdogs ATS remain the best performing group in this subset, although not by too much. Oddly, though it has received much less publicity, the Over has also historically done well enough to beat the vig in home underdog/road favorite situations. However, these margins are razor-thin in both cases – approximately 0.2% ahead of break-even, which amounts to one bet every five seasons. You’re not doing much with this, but it’s still rare to see such an apparently easily exploitable strategy.

What about more recently?

Levitt’s paper was published in 2004, and has received substantial popular republication since. Has the effect shrunk in size as more people have become aware? Here are the last ten years of home underdogs ATS:

Year	ATS
2004	45.57%
2005	35.53%
2006	58.97%
2007	48.91%
2008	43.24%
2009	47.19%
2010	50.62%
2011	53.49%
2012	51.11%
2013	52.38%
Total	48.85%

Levitt wrote a blog followup in 2006 proclaiming that the system continued to work, yet that appears not to be representative. Overall, during the last ten years, home underdogs have covered less than half the time, and have beaten the vig only twice. The single strongest year of the last ten was 2006. If Levitt had tried that same update in 2005, he’d have gotten demolished (35% ATS!). Either way, whether it’s in response to public knowledge of the home underdog effect or not, the evidence doesn’t support just blindly betting all home underdogs anymore.

Weather Effects

A common theme in discussing home field advantage is the edge that cold weather teams, like the Patriots, hold at home. Luckily, provides us with data on the temperature for every game, so we can examine this effect. With that data, we can see that this theme is largely correct – home teams do significantly improve in the cold.

Here’s what the data looks like for games played below 50 ℉, for all years:

Temp < 50 ℉	#	Scored	Allowed	W %	ATS	Over	Under
Home Favorites	1519	24.05	17.08	69.91%	50.24%	51.82%	48.18%
Road Favorites	712	22.02	18.80	61.10%	45.56%	52.12%	47.88%
Home Underdogs	712	18.80	22.02	38.48%	54.44%	52.12%	47.88%
Road Underdogs	1519	17.08	24.05	29.76%	49.76%	51.82%	48.18%

Home favorites are now winning outright nearly 70% of the time (up from 68% previously). Meanwhile our home underdogs have become significantly more profitable, well above the breakeven rate of 52.4%. Additionally, the colder it gets the more the Over comes in on the Total. (NOTE: This may not mean what it looks like. Overall scoring is lower in colder weather: 41.82 points per game overall vs. 41.06 when the temperature is below 50 ℉. Rather, books are shifting the Total to adjust for the colder temperatures, but teams are scoring a bit more than books expect.)

Here’s what it looks like below 40 ℉:

Temp < 40 ℉	#	Scored	Allowed	W %	ATS	Over	Under
Home Favorites	774	24.41	16.74	72.35%	52.30%	51.91%	48.09%
Road Favorites	341	21.66	18.75	59.53%	43.77%	52.80%	47.20%
Home Underdogs	341	18.75	21.66	39.59%	56.23%	52.80%	47.20%
Road Underdogs	774	16.74	24.41	27.13%	47.70%	51.91%	48.09%

Home teams generally are faring better and better. Below 40 ℉, betting on all home teams becomes profitable, even the favorites. The home underdog betting advantage in particular looks strong in cold weather.

Lets keep going, to below 35 ℉:

Temp < 35 ℉	#	Scored	Allowed	W %	ATS	Over	Under
Home Favorites	472	24.76	16.38	76.06%	55.27%	53.45%	46.55%
Road Favorites	199	21.71	19.14	56.78%	40.84%	55.28%	44.72%
Home Underdogs	199	19.14	21.71	42.21%	59.16%	55.28%	44.72%
Road Underdogs	472	16.38	24.76	23.52%	44.73%	53.45%	46.55%

More of the same. At these temperatures, betting all home teams is strongly profitable. Additionally, betting the Over starts beating the vig as well. (NOTE: Scoring per game remains roughly flat at around 41 points per game below 35 ℉ and below 50 ℉. The difference in success betting the Over comes purely from lower scoring expectations at these temperatures.)

Here’s what it looks like below 30 ℉:

Temp < 30 ℉	#	Scored	Allowed	W %	ATS	Over	Under
Home Favorites	252	24.58	16.44	75.00%	54.25%	54.07%	45.93%
Road Favorites	112	21.64	19.49	56.25%	40.74%	57.14%	42.86%
Home Underdogs	112	19.49	21.64	41.96%	59.26%	57.14%	42.86%
Road Underdogs	252	16.44	24.58	24.21%	45.75%	54.07%	45.93%

And here’s the data for below 25 ℉:

Temp < 25 ℉	#	Scored	Allowed	W %	ATS	Over	Under
Home Favorites	120	25.96	16.25	78.33%	55.93%	57.14%	42.86%
Road Favorites	59	22.32	20.93	52.54%	39.66%	64.41%	35.59%
Home Underdogs	59	20.93	22.32	47.46%	60.34%	64.41%	35.59%
Road Underdogs	120	16.25	25.96	20.83%	44.07%	57.14%	42.86%

The same trends continue. The colder it gets, the stronger home underdogs perform, and the more the Over comes in. Oddly, and this may be a sample size issue, below 25 ℉ scoring actually increases, spiking to 42.55 points per game (higher than the average across all temperatures).

These trends continue at the temperatures drop further, although the sample sizes become vanishingly small. Just for fun though, here are results for below 20 ℉, 15 ℉, and 10 ℉:

Temp < 20 ℉	#	Scored	Allowed	W %	ATS	Over	Under
Home Favorites	63	25.22	17.13	74.60%	51.61%	59.68%	40.32%
Road Favorites	33	21.45	21.61	51.52%	40.63%	72.73%	27.27%
Home Underdogs	33	21.61	21.45	48.48%	59.38%	72.73%	27.27%
Road Underdogs	63	17.13	25.22	23.81%	48.39%	59.68%	40.32%

 

Temp < 15 ℉	#	Scored	Allowed	W %	ATS	Over	Under
Home Favorites	38	24.87	16.79	76.32%	51.35%	52.63%	47.37%
Road Favorites	20	20.95	22.8	45.00%	31.58%	80.00%	20.00%
Home Underdogs	20	22.8	20.95	55.00%	68.42%	80.00%	20.00%
Road Underdogs	38	16.79	24.87	23.68%	48.65%	52.63%	47.37%

 

Temp < 10 ℉	#	Scored	Allowed	W %	ATS	Over	Under
Home Favorites	20	26.60	19.05	75.00%	45.00%	55.00%	45.00%
Road Favorites	13	17.00	25.46	15.38%	7.69%	69.23%	30.77%
Home Underdogs	13	25.46	17.00	84.62%	92.31%	69.23%	30.77%
Road Underdogs	20	19.05	26.60	25.00%	55.00%	55.00%	45.00%

Yup. Below 10℉, only one road favorite has covered since 1978, and only two have won games outright. Home favorites meanwhile have ceased keeping pace with the home underdogs.

However, remember what we saw in the table showing annual results since 2004, the year Levitt published his paper: outside of a couple seasons, the home underdog effect vanished (48.85% ATS). Those figures didn’t account for temperature, so let’s take a look at 2004-2013, below 40 ℉ degrees:

Temp < 40 ℉	#	Scored	Allowed	W %	ATS	Over	Under
Home Favorites	222	25.76	17.74	70.72%	50.68%	51.82%	48.18%
Road Favorites	93	24.23	19.42	63.44%	47.06%	55.91%	44.09%
Home Underdogs	93	19.42	24.23	35.48%	52.94%	55.91%	44.09%
Road Underdogs	222	17.74	25.76	28.83%	49.32%	51.82%	48.18%

The Over grades out well, and while the home underdog ATS result is enough to beat the vig, it’s a smaller advantage than it was before the Levitt paper.

Here’s 2004-2013, below 35 ℉:

Temp < 35 ℉	#	Scored	Allowed	W %	ATS	Over	Under
Home Favorites	142	26.49	17.81	72.54%	55.00%	55.32%	44.68%
Road Favorites	46	25.11	20.83	58.70%	42.86%	67.39%	32.61%
Home Underdogs	46	20.83	25.11	41.30%	57.14%	67.39%	32.61%
Road Underdogs	142	17.81	26.49	26.76%	45.00%	55.32%	44.68%

Getting better. 57% ATS is strong, and enough to make a real profit (But the Over again is the real cold weather winner). As a word of caution however, here’s what it looks like below 30 ℉:

Temp < 30 ℉	#	Scored	Allowed	W %	ATS	Over	Under
Home Favorites	76	26.42	18.07	67.11%	52.00%	58.67%	41.33%
Road Favorites	27	26.19	20.44	59.26%	57.69%	70.37%	29.63%
Home Underdogs	27	20.44	26.19	40.74%	42.31%	70.37%	29.63%
Road Underdogs	76	18.07	26.42	31.58%	48.00%	58.67%	41.33%

All of a sudden, the home underdogs are getting killed. This is a good reminder of the danger of splicing data too much, and of doing too much data mining generally. Sample sizes become vanishingly small, which increases the risk of flukes that look important.

Generally, the home underdog effect does seem to have shrunk, in all temperatures, since Levitt brought it to popular attention. On the other hand, to my knowledge, no popular blogger has yet made any substantive comment on the profitability of betting the Over in cold weather games. That effect appears robust across time periods so far, and increases as temperatures drop, which is something to watch closely as the NFL season closes in on colder weather.

Konstantin Medvedovsky writes about football science, in both college football and the NFL.