More on Luck With Interceptions

Last week, I asserted while looking at turnovers in 2015 that recovering fumbles and actually being able to haul in passes that a defender gets a hand on were luck-based activities. Yesterday, I expanded the view on fumble recoveries and found that they really did seem to be random. Today, I'm doing the same in-depth look at interceptions.

Intuitively, interceptions would seem to be based less on luck than fumble recoveries. Catching a ball without it hitting the ground first requires more skill than grabbing an object on the ground. A football travels in ways that are difficult to predict when it's bouncing around, but a player can hit it around or have it roll around and still be able to recover the fumble. Snatching the ball from the air is a far less forgiving activity.

Using data from CFBStats.com, here is the distribution of interception-to-passes defended ratios from 2008-15. Passes defended as a stat is simply pass breakups plus interceptions. The rates are rounded to the nearest whole percentage point and only include games between FBS teams.

The distribution still looks bell-shaped for the most part. There is some weirdness near the top that I assume has to do with randomness, the rounding involved here, or both.

That said, the right side is steeper than the left side. Some of that is probably the zero percent lower bound, as it's impossible to intercept fewer than zero passes. I didn't think that covered it entirely, though, so I investigated the matter further.

The median was 22% and the average was 21.9%, so those line up nicely. The numbers of teams on either side of the average are just about the same. Of the 985 teams over the eight seasons covered, 499 registered a rate below the average and 486 were above it. That symmetry is fuel for the randomness hypothesis here.

However, the number of teams within one standard deviation below the average was 353, while only 235 were within one standard deviation above the average. Meanwhile, 220 teams were between one and two standard deviations above the average, whereas just 126 were between one and two standard deviations below the average. Those are huge differences, and they suggest that something is happening here that is not mere luck. The standard deviation was 6.6%, if you're curious.

I think the answer to the puzzle can be found by going even more to the extremes.

An issue I discussed in last week's post on underdog strategy is that the smaller the number of occurrences of something, the more likely it is that you'll end up with a result that is far from average. Therefore, we should expect to see a lot of low numbers of total passes defended among the highest and lowest interception percentages.

The average number of passes defended for all teams was 51.8 per season. The standard deviation was 14.2, so any team with 37 or fewer total passes defended in a season was at least one standard deviation below average.

Of the teams with the 50 lowest interception percentages, 16 of them had a passes defended count at least one standard deviation below the mean. This was a sample of 5.1% of all teams, but it contained 9.6% of the teams that were at least one standard deviation below average in total passes defended (16 of the 167 such teams). In short, the teams with the lowest interception ratios had a disproportionately high number of teams with low total passes defended.

On the other end of the spectrum, the teams with the 50 highest interception percentages featured 11 teams with a low total passes defended count. That is still disproportionately high—11 of 167 is 6.6%, versus a sample of 5.1% of total teams—but it's not as disproportionately high as with the lowest interception percentages.

Further, only one of the teams with the 50 lowest interception percentages also had a total passes defended count of more than one standard deviation above the average. This situation is as expected, because the more passes defended, the more a team should trend towards the mean.

Among the 50 highest interception percentages, five of them were more than one standard deviation above average in total passes defended. Those worst of those teams was 2014 Louisville, which went 9-4. Two others went undefeated—2009 Boise State and 2013 Florida State—while two more lost just one game—2009 Texas and 2014 TCU.

This is where the fundamental difference between fumbles and interceptions comes in.

Fumbles are more likely on some plays than others. A play with an option pitch will generate more fumbles than an I-formation handoff, for instance. Fumbles are also more likely on pass plays than run plays, and that is in large part due to the fact that fumble rates are about 15 times higher on sacks than on runs.

Strategically, teams don't often make the decision to switch to run plays with a greater chance of fumbles. Far more often is a situation where a team is leading in the second half and switches to the most basic and conservative runs to minimize the chance of turnovers.

Teams will switch to a strategy of passing more when they're behind, which will increase the likelihood of fumbles. However, teams that are behind don't just pass more; they pass more aggressively to try to gain large chunks of yards at a time. When these teams ramp up the passing, they're not adding a bunch of safe screens and three-yard slants. They're chucking it ten or more yards up the field as much as they can, and those passes are more likely to be intercepted than shorter tosses.

Further, quarterbacks of teams that are behind tend to try to be heroes. So not only are they throwing higher risk passes, but they're probably going to be forcing throws into tighter windows than they would on the same play in the first quarter. Plus, Hail Mary passes have sky-high interception rates. A team that is throwing the ball because it's behind late is ramping up its interception risks on several fronts.

Of all the teams that both had an interception rate and a total passes defended count at least one standard deviation above average, there wasn't a bad team in there. I told you five of them already. The others were:

• 2008 Florida, 13-1
• 2008 Iowa, 9-4
• 2010 Hawaii, 10-4
• 2010 Tulsa, 10-3
• 2012 Kent State, 11-3
• 2013 Houston, 8-5
• 2014 Boise State, 12-2
• 2014 Ohio State, 14-1
• 2015 Arkansas State, 9-4
• 2015 West Virginia, 8-5

Not all of them were world beaters, but 11 of the 15 won at least ten games, seven went to BCS/New Year's 6 bowls, and three won national championships.

The takeaway (ahem) from yesterday and today's posts is that it's probably impossible to increase a team's fumble recovery rate by focusing on it or trying harder. However, it does seem to be possible to jack up interception rates by being a good team and forcing opponents into situations where interceptions are more likely. There also is the the fact that a team can be coached to jump more routes—something that probably will increase the interception rate, which is seen here, while potentially increasing the rate of big plays allowed, which isn't seen here.

Oh, but remember how I said there was that one team in the bottom 50 in interception percentage that had an unusually high number of total passes defended? It was the 2008 Texas Longhorns, whose defense caught just 8.5% of the passes it got a hand on. The team had the 18th lowest rate among 985 teams in the eight-year span I looked at. A year later, the Longhorns picked off 37.9% of the balls they got a hand on, the 15th highest rate out of 985 teams.

The '09 Texas secondary benefitted from the maturation of young talent and being in the second year under defensive coordinator Will Muschamp, but this anecdote is your reminder that a good amount of interception percentage still is just luck.