Monday, April 23, 2007

Temperature Differences and Home Field Advantage

Back in October 2006, the pro-football-reference blog posted a two part guest post I prepared on the effects of travel distance and climate here, http://www.pro-football-reference.com/blog/wordpress/?p=156 and http://www.pro-football-reference.com/blog/wordpress/?p=157.

This is a refinement of that information, plus the info with all games, not just divisional.

I tweaked the methodology on climate slightly. Previously, I used the average temperature for the months of September through December for each outdoor city, and the difference between the averages for two cities. The only slight problem with this is that not all cities with similar average temperatures have similar climates. To try to capture this better, in this post, I used the average monthly difference between the cities in question, then re-sorted the data.

An example of this is Seattle and Chicago. These cities are within 1 degree in average temperature over the whole football season. However, Seattle is the coolest city in September, while Chicago is much warmer, but by the end of the season, the numbers have reversed. My new methodology was to look at the absolute difference in degrees for each month, and average them. The other refinement was a slight tweak to the dome average temperature, I used 72 degrees (the data supported this, as like the outdoor vs outdoor games, the home field advantage between dome and outdoor was weakest against teams between 65 and 76 degrees).

I have included dome teams, but for now, I am excluding Denver. Denver is a case study on its own. Here is the data for all other divisional games, using the new climate methodology.

within 5 degrees difference, monthly average: 289-259-1 (0.527)
5.1 - 10.0 degrees difference, monthly average: 303-267-1 (0.532)
10.1-15.0 degrees difference, monthly average: 136-98-0 (0.581)
15.1-20.0 degrees difference, monthly average: 199-126-0 (0.612)
20.1-25.0 degrees difference, monthly average: 144-85-0 (0.629)
25.1 + degrees difference, monthly average: 72-47-0 (0.605)

Again, the effect of climate differences in divisional games is real and fairly strong. However, it is at the similar temperatures, not the extreme ones, where the difference is noticeable. Once divisional opponents are outside a 10 degree difference, the home field advantage increases dramatically, and beyond 15 degrees, it is fairly constant.

However, the effect is not the same with non-divisional games. Here is the overall data for all games played between 1986-2005 (except involving Denver) using the same criteria.

within 5 degrees difference, monthly average: 575-453-4 (0.559)
5.1 - 10.0 degrees difference, monthly average: 653-498-1 (0.567)
10.1-15.0 degrees difference, monthly average: 450-318-0 (0.586)
15.1-20.0 degrees difference, monthly average: 529-364-1 (0.592)
20.1-25.0 degrees difference, monthly average: 227-158-0 (0.590)
25.1 + degrees difference, monthly average: 119-77-0 (0.607)

So, there is a split between divisional games and non-divisional games. The overall effect still appears when we combine both. Why is this? Other than a random split? Two guesses:

First, the NFL may schedule divisional contests at a higher frequency both early and more importantly, later in the season (they do have to get 2 in vs each divisional opponent), and these are the same times when weather differences will be more extreme.

Second, maybe there is also a familiarity component to reducing home field advantage, so that visitors from similiar climates, who are also familiar with the specific venue, are better off than visitors from similiar climates, but unfamiliar with the specific city/stadium.

I did look at all cases since the merger where a team stayed in the same city it had been in for the previous 5 years, but moved into a new (but same climate) venue, to see what effect it had on home field against divisional opponents from a similiar climate. This would include: Buffalo (1973), NY Giants (1976), NY Jets (1984), Washington (1997), Cincinnati (2000), Pittsburgh (2001), New England (2002), Detroit (2002) and Philadelphia (2003). The sample size is not large enough to say anything with statistical certainty, but it was interesting that the home field advantage against similiar divisional opponents spiked in year 2 following the move, and disappeared thereafter.

These teams went 12-9-1 at home against their "climate" rivals in year 1 after the move, and an incredible 18-4 at home in year 2. They won about the same number at home and on road in the years leading up to the move, and reverted to this pattern in years 3-6 following the move. So, there may be something to the familiarity thing. I may also try to look at conference rivals who played at the same venue in more than 2 consecutive seasons, division rivals after relocation, and new divisional rivals (such as with realignment in 2002).

Finally, here are the point spread records since 2002, of the home team in divisional games, sorted by temperature difference.

Within 5 degrees on average: 59-75-2 (.441)
5.1 to 10 degrees on average: 56-76-4 (.429)
10.1 to 15 degrees on average: 48-48-4 (.500)
15.1 to 20 degrees on average: 25-15-0 (.625)
20.1 degrees or more on average: 30-30-0 (.500)