Here is the first part of a series that looks at historical trends regarding NFL games and betting. Fits perfect into cirrustheduck.com huh? Cat pics and football analytics. The trends are based on historical lines dating back to 1978.
In general, we will look largely at probabilities given a line more or less than a certain amount. This is opposed to looking at trends at a given line. Although that would be ideal (how often does one team cover when favored by five and a half points), but that method severely decreases sample sizes and requires extra fitting and smoothing to identify trends.
The first trend we’ll look at is the probability that a team covers the spread based on how many points they are favored by. Below, we have this probability for home teams determined along with 90% confidence intervals. .
One can see that there is a definite trend. In general, the more points that a home team is favored by, the more likely it will not cover the spread. The left hand side of the plot shows essentially how often the home team covers the spread in all situations (favored by more than -20 points). The plot shows that the home team covers about 48% of the time. Similarly, the visiting team covers about 48% of the time, and the rest of the games push.
When the home team is favored by more 10 points or more, they cover the spread only 45% of the time. Around this point, it may prove profitable to start betting the visiting team. Finally, when favored by 15 points, the home team covers 35-40% of the time. Betting the other side is almost certainly profitable in the long term.
You may also notice how the trend jumps around a bit and the confidence intervals increase towards the right side of the plot. This is because the sample size decreases as you hit the extreme cases (large spreads). Something you should notice with later posts as well. That isn’t to say you should ignore the data at the ends, but you should be aware that it is less likely to be accurate than data points with a tight confidence interval.
Another small thing that appears to be significant is a bump at 4 points favor. Interestingly, data shows that the home team is more likely to cover with a 4 or 5 point spread than a 2 or 3 point spread. It is still not profitable to bet long term at those odds, but it is an unexpected trend (that appears significant). The reason may be due to how the game is played in those situations. For example, a team that is down 2 or 3 points will be less aggressive than one that is down by 4 or more (needing a touchdown towards the end).
If we take a look at the same plot but for the visiting team, a few things are different.
Just like the home team, there is a relatively quick decline in the probability the visiting team covers when favored by more than 5 points. What is unusual, however, is an increase in covers when favored above 10 points. This could be due to smaller sample sizes and meaningless. Perhaps, though, visiting teams with huge spreads only happen when they are significantly better than the home team and lead to some blow outs.