In Week 1 of 2015, a 49ers team that went onto finish 5-11 and fire their head coach (RIP Tomsula) mopped the floor with a Vikings team that would finish 11-5 and make the playoffs. The Jeff Fischer-led Rams somehow beat the mighty Seahawks multiple times. Things don’t always go according to plan in the NFL. Situational factors such as weather, bye weeks, divisional opponents, travel distance, and time of the game are often cited as an explanation for why those things don’t go according to plan. In the article below, I explore whether any of these factors have been statistically associated with wins, points, and offensive efficiency. These factors by themselves should not be the sole driver of decision making for picking fantasy football lineups, making bets, or predicting wins, but they do serve as a piece to the puzzle and should not be ignored. They can certainly highlight situations that should be avoided (such as starting QBs playing in 20+ MPH winds), as well as situations that should be targeted (such as the under on Thursday Night Football). Many of the factors covered below are included in our NFL models and serve as contributing factors to our win, spread, total predictions, and ELO ratings, which we will be publishing every week of the upcoming NFL season.
Most of the numbers below use NFL games from 2002-2016 and are shown from the perspective of the home team (e.g., if I say teams playing indoors win 58% of the time, that means the home team won 58% and the away team won 42%). A few reference points for home team performance from 2002-2016: home teams have won 57.6% of games and covered the spread in 49.1% of games, with an average margin of victory of 2.7 points. The median Vegas Total was 43 points, and the over has hit at a 49.9% rate, which is pretty remarkable. Home teams have scored an average of 23.3 points while holding their opponent to 20.6 points (giving total points scored of 43.9). Passing games have averaged 225 yards on 6.7 yards/attempt (YPA), 61.4% completion percentage, and a QB rating of 87.2. Running games have averaged 118 yards on 4.2 yards/rush (YPR).
Throughout the article, I also use correlations. Correlation is a measure that explains the relationship between two variables, and ranges from -1 to +1, where +1 is an exact positive linear relationship between two variables (i.e., if one variable increases by 5, the other variable increases by 5) and -1 an exact negative linear relationship between two variables (i.e., if one variable increases by 5, the other variable decreases by 5). A correlation of 0 would represent no linear relationship between the two variables.
A common weather-related narrative is that mother nature has a strong(er) effect on the passing game. The data does not provide much support here, as seen in the correlation matrix below, which shows that temperature and wind speed do not correlate strongly with completion percentage, QB rating, or the number of passes attempted:
|Stat||Comp %||QB Rating||# of Pass Att|
In terms of wins and losses, weather has essentially no correlation once again. The correlation between margin of victory (MOV) and wind speed is -0.01 and the correlation between MOV and temperature is 0.05, implying that there is basically no linear relationship between these variables. However, there are some interesting splits that arise if we look at temperature/wind speed in specific ranges, rather than from a purely linear standpoint. The table below shows splits of games played above/below 32 degrees, which highlights jumps in home team winning percentage and MOV in colder weather:
|Temp||Home Win%||Home MOV||Home ATS%||Over%||Total Pts||YPA||YPC|
Presumably, these increases are due to home teams being more accustomed to playing in the cold. It is also worth nothing that many cold-weather teams have won a lot over the last decade (Patriots, Steelers, and Packers) and thus will contribute higher-than-average home win percentages to this sample. Another note from this split – more games have gone over the Vegas total in cold games (54.0%) than warmer games (49.7%). Vegas may be dropping totals too much based on weather, which has caused cold games to go over more often. On the flip side, in the 25 games played below 32° in the past two seasons, the over is an abysmal 9-16 (36%). Perhaps Vegas has started over-correcting for this in recent years, or perhaps 25 is just too small a sample of games. If we extend to our entire data set (1994-2016), we have 342 games played below 32°, and the over is 184-154-4 (54.4%) in those games.
As you might suspect, home teams do not see a change in Win % at differing levels of wind speed, as playing in high-wind environments is not something any teams are extremely familiar with. However, we do see notable decreases in total points scored and passing volume/efficiency once wind speed gets high enough:
|Wind Speed||Total Pts||YPA||Pass Yds||Pass Att||Comp Pct||QBR||YPR||Rush Att|
The differences between, say, the 1-5 MPH bucket and the 6-10 MPH bucket are not very drastic. However, when comparing the 20+ MPH splits to the low-end wind speeds, there are substantial drop-offs across the board (e.g. 37.0 total points scored / 45.9 total points scored). For fantasy and DFS purposes, I would strongly consider fading QBs/WRs/TEs in games with 20+ MPH winds (NFLweather.com is a good resource for checking game weather). For betting or forecasting wins, teams that are extremely reliant on passing success could struggle to put up their usual points in high-wind outings.
Similar to wind speed, playing indoors vs. outdoors does not provide much of an advantage for the home team in terms of winning games. BUT, we do see notable increases in scoring and passing game success for indoor games, as highlighted in the splits below. For fantasy football and DFS, this can help identify players who are more likely to be involved in high-scoring games and players on passing offenses that should be firing on all cylinders. While this shouldn’t be the sole driver of fantasy football decision making, at the very least, indoor games could be used as a tiebreaker between similar players. For betting, Vegas is well aware of the fact that more points are scored in these games, as illustrated by the difference in the average total from the table above (42 for outdoor vs. 45 for indoor) and the fact that overs for indoor games only hit slightly more frequently (51.9%) than overs in outdoor games (49.3%).
|Stadium||Home Win%||Home MOV||Home ATS%||Over%||Avg Total|
|Stadium||Total Pts||YPA||Pass Yds||Pass Att||Comp %||QBR||YPR||Rush Att|
Bye Weeks and Days of Rest
When comparing to a “normal” week (i.e., 7 days between games), there are solid bumps in Win %, MOV, and ATS % for teams coming off of a bye week. Though, the increased winning marks are not accompanied by substantial increases in offensive metrics:
|Week Type||Home Win%||Home MOV||Home ATS%||Total Pts||Over %||YPA||QBR||YPR|
Albeit less-drastic changes, teams also frequently play on either 6 or 8 days of rest when they are coming off/going into a Monday night game. These splits are not as severe relative to bye weeks, but there are notable dips in points scored, Win % and ATS % when playing on 6 days of rest, and the over has hit at a much higher rate (57.2%) when the home team is playing on 6 days of rest:
|Days of Rest||Home Win%||Home MOV||Home ATS %||PPG||PAPG||Total Pts||Over %|
|6 Days Rest||54.4%||-2.6||45.0%||24.7||22.1||46.8||57.2%|
|7 Days Rest||57.7%||-2.5||49.1%||22.9||20.4||43.4||49.1%|
|8 Days Rest||56.1%||-2.1||49.4%||23.6||21.5||45.0||50.9%|
Maybe it is just me, but divisional games have always seemed more competitive and lower scoring than expected. Here are some basic splits between division games and non-division games, which agree with my gut feeling to some degree, but do not highlight any drastic differences:
|Game Type||Home Win%||Home MOV||Home AvgSpread||Home ATS%||Total Pts||Avg Vegas
Digging a bit deeper, the table below further splits games on (1) whether the home team is an underdog or a favorite, and (2) whether it is a division game. Notice the drastic differences between the underdog vs. favorite categories:
|Game Type||Home Win%||Home MOV||Home ATS%||Total Pts||Yds/Gm||YPA||YPR|
|Dog / Non-Div||38.8%||+3.7||51.0%||44.2||323||6.3||4.14|
|Dog / Div||33.6%||+5.2||48.8%||43.6||315||6.2||4.09|
|Fav / Non-Div||68.4%||-6.2||49.7%||44.4||356||7.0||4.27|
|Fav / Div||66.9%||-5.9||46.5%||42.8||350||6.8||4.28|
Teams can get put in a pretty tough spot when it comes to travel, especially West Coast teams like the Raiders, Rams, Chargers, and Seahawks, who often fly 2,000+ miles across the country for a game (or 5,000 miles according to Jon Gruden). To start, travel distance only has a correlation of 0.03 with MOV, indicating a weak/non-existent relationship between the two. The table below provides further support to this theory, showing how teams have fared at different travel distances. Most apparent, Win % and ATS % do not seem to have a strong trend to them until the high-mileage trips, where there is a drop off in the competitiveness of the road team:
|Travel Dist||Home Win%||Home MOV||Home ATS%||PPG||PAPG||Over %|
|< 300 Miles||54.8%||-1.9||48.2%||22.4||20.5||48.7%|
|300 – 499 Miles||58.7%||-2.9||49.6%||23.3||20.4||46.4%|
|500 – 799 Miles||55.6%||-2.3||48.1%||23.2||20.9||49.0%|
|800 – 999 Miles||59.0%||-3.0||49.8%||23.5||20.5||50.8%|
|1,000 – 1,499 Miles||55.5%||-1.8||47.1%||22.8||21.0||50.3%|
|1,500 – 1,999 Miles||58.3%||-2.8||49.9%||23.5||20.7||51.4%|
|> 2,000 Miles||61.2%||-3.8||52.0%||24.0||20.2||52.9%|
My first thought was that divisional games (which are typically shorter travel) were clouding these numbers, but even after removing all division games, the numbers don’t shift much apart from the < 300 Miles category, which becomes less competitive after removing divisional games (i.e., higher Win % for the home team).
Another narrative to address here is teams traveling far on a short week – thankfully I was able to dig up some useful data to support this one. Extending to our entire data set (1994-2016) to get a sample of 104 games, when home teams face a non-division opponent that is playing on a short week (i.e., less than 7 days since their previous game) and traveling 1,500+ miles, the home team has won 67.3% of games and covered the spread 63.0% of the time, winning by an average score of 24.5 to 19.5.
Time of Game
Nearly all NFL games are played at set days/times: Sunday early, Sunday late, Sunday night, Monday night, or Thursday night. The table below highlights the differences in outcomes among these different times. Interestingly, home teams have the worst Win % and MOV on Monday Night Football. This falls in line with the idea that road teams “come to play” on Monday Night and often put up a good fight. Sunday Night and Thursday Night Football see the best marks for home teams in terms of Win %, ATS %, and MOV. Lastly, Thursday Night games see the fewest points scored (42.3) coupled with the lowest Over %, with the over only hitting 45.2% of the time. I would be hesitant to bet on road teams and overs on those usually-thrilling Thursday Night games.
|Game Time||Home Win%||Home ATS%||Home MOV||Total Pts||Over %|
2 thoughts on “The Impact of Situational Factors on NFL Games”
Comments are closed.