Watching a sports TV show yesterday I saw an interesting study on cold weather’s impact on NFL totals. The big problem I saw with the study was that it looked only at a period from 2002 thru 2006. This was far too narrow a slice of time to really be meaningful in my opinion. I found the study online (done by Brian Burke) and examined the methodology used. As the data did not include the last 5 years, nor the bulk of what is considered the “modern football era” I felt it would be very worthwhile to expand the study for a complete analysis.
Essentially the study was done by isolating cold weather teams and looking at just their home games over 3 periods during the football calendar:
- Regular Season, Prior to Week 14 (Essentially Sep thru the 1st week of Dec)
- Regular Season from Week 14 onward (mid Dec thru early Jan)
- Playoffs (early Jan thru Conference Championship games)
The cold weather teams I isolated for use in the study were all teams with avg December temps no higher than 44, with avg December winds above 10 mph, thus producing avg wind chills in December of 38 degrees or lower. This gave me 8 different cities (9 teams) to use in the study, whose combined December windchill averaged to be exactly freezing (32 degrees). The teams selected (from coldest avg Dec wind chill) GB, Buf, Cle, Chi, NE, KC, Pit, NYG, NYJ.
Because we know scoring has increased in recent years in general, I felt it would be helpful to separate the data into two periods. The first was just 2002-current (2011) so we could look at just the data since divisional realignment and inclusive of the high scoring game of today. The second period was a complete look back from 1989 thru present, to confirm if this is a long term trend which can be confidently relied upon.
In addition to looking at these 9 cold weather teams at home in general, I also separated out two teams in particular (Green Bay and New England) as both are hosting playoff games in what is expected to be frigid weather this weekend.
The findings were very interesting and are detailed below.
Period 1: 2002 – 2011
As you can see (click on chart for larger view) during this most recent period of time, the average total during the cold late season months decreased from 41.5 to 39.7 (1.8 points). But the cold weather actually lowered the scoring by just 1.2 points on average, therefore, it’s affect was over-compensated for by the linesmakers.
Moving into the playoffs, the average total was very close to that of the average during the warmer months of the regular season (in fact it was slightly higher). But the total points scored was significantly higher (45.8) than it was even in the cold late season weeks (41.3). As a result, 59% of the games went over the total, by an average of 4.2 ppg.
So as to the question: does scoring decrease in cold weather, the question is not 100% answered by this study. Scoring in the late season is decreased as compared to the warmer months of the regular season. However, potentially because better offenses make the playoffs, scoring actually is significantly higher when the weather would be at its coldest point during the NFL calendar (early-late January).
What is answered (and perhaps it’s more important from a gambler’s perspective) is that games in cold weather cities are significantly more likely to see higher scoring and more overs than games in the late regular season and even the warm months of the regular season, which certainly is a surprise.
As it relates to Green Bay and New England, the two cities who host playoff games this weekend: Green Bay is completely in line with the points scored in all cold cities: lowest during the late season weeks but highest in the playoffs. And when it comes to playing the total, the playoff games have gone over the posted total by an avg of 5 ppg, a larger margin than all cold weather cities combined.
However, when it comes to New England, this city is an anomaly. Home playoff games in New England have seen far fewer points scored in playoff games than any other type of game, and these games tend to go under, both from an overall percentage (67% unders) and looking at average margin, as these games go under by an avg of 4 ppg. That is a 9 ppg swing when comparing the totals of GB and NE: GB goes over by 5 ppg, NE goes under by 4 ppg.
Period 2: 1989 – 2011
As you can see from the above, the results are very consistent with the 2002-2011 period. Clearly, over this larger period of time, scoring was down and as such, overall ppg numbers and totals set are lower. However, the key findings are consistent:
1. In cold weather cities, fewer points per game are scored late in the regular season, when it is colder than the warmer months of the regular season.
2. However, games tend to go over at a higher rate because linesmakers lower their average total by almost 2 ppg, yet scoring decreases by just over 1 ppg.
3. Playoff games are not affected by cold weather – more ppg are scored in the playoffs than even in warm weather, games go over at a larger percentage and by a larger amount of ppg than in the warm months of the regular season.
4. New England is not in line with the rest of the cold weather cities – their playoff games see significantly lower scoring and significantly more games going under by a significantly larger ppg (7.5).
Improving this Study
Clearly a way to improve this study would be to obtain precise weather data for games and determine how the precise weather affects scoring and games going over/under the total. This study was done using the same methodology as the study I saw reported on TV, which did not use precise weather data. It was based on averages: average temps, avg winds, avg snow and avg wind chills. Clearly there will be some playoff games included in the study which saw 50 mph winds and wind chills in the negative range, while others saw 50+ temps in sun and wind-free conditions. These results are averaged together and not studied independently. If I had significantly more time, and access to detailed weather data, I would gladly perform a more comprehensive analysis.
From data I have seen online from Cold, Hard, Football Facts, they did compare two sets of “cold weather” to non-cold weather. In a nutshell, what they found was:
In 818 games featuring “Freeze or Snow” (temperature <=32 degrees OR snow, ice, or sleet present) since 1941, scoring decreased by 1.3 ppg and passing yards decreased by 5.8 ypg.
In 105 games featuring “Extreme Freeze or Snow” (temperature <=15 degrees OR heavy snow OR wind chill <= (-10) degrees) since 1941, scoring decreased by 2.8 ppg and passing yards decreased by 18.7 ypg.
However, there was no analysis done as to how the scoring related to the total points posted on the games, which is what matters most to gamblers. While scoring may decrease by 1.3 ppg in “Freeze or Snow”, if the linesmaker lowers the total by 5 ppg, there is absolutely no value in betting the games under.
Eventually I am positive the missing link will be filled in, by myself or by someone else.