While I might not be in the majority here, I think interceptions are the most gut-wrenching non-injury-related plays to witness in football.
One careless flick of the wrist and the most beautifully orchestrated drive is stunted and becomes a chance for the opposing defense to flip the field and help their offense to a score of their own. It’s a waste of effort and a killer for the emotions of the team and its fans, and the results can be catastrophic.
But interception rates are not just big-time hazards for real-life football. They can also be the difference between a good fantasy quarterback and a waiver-wire option.
Our own JJ Zachariason previously looked at how touchdown rates can help us find fantasy value, which inspired my own analysis of how interceptions affect quarterback potential. So let’s dig in.
Risky Business
Interceptions matter for fantasy football.
Everyone knows they’re bad for your NFL team, and throwing a pick might not lose as many fantasy points (-2) as a touchdown adds (4) in standard leagues, but there are quite a few effects that the fantasy box score doesn’t measure. Think about this: toss the ball to the opposition and you not only get -2 points for the play, and you also lose the chance for your quarterback to complete more passes, earn more yards, and score a touchdown on that drive. You lose all those opportunities.
Say we look at player seasons from 2000 to present: quarterback fantasy points correlated with intercepted passes, which amounts to a coefficient of determination (R squared) of -0.46, or a moderately negative relationship (meaning the more interceptions thrown, the fewer points scored). While this is not an ironclad relationship, with the way passing yards, touchdowns, and completions work, there is a significant impact on scoring potential when a quarterback gives the ball away.
Unsurprisingly, the average fantasy points scored by a passer tossing multiple interceptions in a game last year was 10.4, more than 25 percent less than the average quarterback performance (14.2 points) last year. Of the 104 games where a quarterback threw multiple picks, just 18 of them put up 17.4 points or more -- the average score for a replacement-level fantasy starting quarterback.
Now I don’t know about you, but I don’t like staring down a 17 percent chance that my fantasy quarterback is worth starting in any given week. That means that we need to limit the chance that our fantasy quarterbacks throw interceptions.
Interceptions can also be somewhat unpredictable, though. Did any of you expect Case Keenum, after a 2016 season that saw him post a 3.4 percent interception rate (11 picks on 322 attempts), to put up an interception rate of just 1.5 percent a year ago? That's nearly a drop of two full percentage points. Let's be real -- you did not see that coming.
Time is a Flat Circle
The thing about Keenum, though -- who now holds a career 2.1 percent interception rate -- is that he is very unlikely to repeat his incredible 2017 season. In fact, he is no more likely to repeat this performance than Tom Brady was to repeat his 0.8 percent rate from 2016, or Matt Barkley was to encore at a rate of 6.5 percent from that same season.
Of those passers to tally 200 or more attempts since the 2000 season, there have been just 9 instances where a quarterback maintained an interception rate of 1.5 percent or lower. Of those nine, only five actually lowered the rate of interceptions per passing attempts thrown from one year to the next.
That said, any accurate passer will stay relatively accurate and any wild passer will stay relatively wild. We’re not expecting a quarterback who has had pinpoint accuracy his whole life to all of a sudden just spray the ball all willy-nilly, nor are we expecting any players but those on the furthest ends of the spectrum to regress to the historical average that hard.
Furthermore, there's not really any sort of career momentum or "mappable" development for NFL quarterbacks in terms of their ability to avoid turning the ball over. Instead, we have to look at regression and how passers’ rates should be expected to shift based on other players as a whole.
Some players may be better or worse than the norm at preventing interceptions, but regression gets everyone.
Expected Interception Rate
In the same vein as JJ’s study, I laid out the interception rate and the next-year difference in interception rate for quarterback seasons from 2000 to 2016 (we exclude 2017 data because there is no “next year” data) yet) for passers with at least 200 passing attempts.
The chart below shows us a clear trend for how far interception rates regress from one year to the next. The x-axis (horizontal) shows us any given player's interception rate in a season, and the y-axis (vertical) shows us by how many percentage points it changed the next year.
The dotted line going across the middle shows us that passers have tended to even out right around a 2.8 percent interception rate (as shown by a 0% shift). That means that if they have an interception rate in that ballpark this year, we shouldn’t expect them to deviate too far from that in the coming year. Take that dot right in the upper middle of the graph, for example. That data point represents Blaine Gabbert's 2012 season, when he posted a 2.2 percent interception rate that then ballooned by 5.9 percent the next season.
What this helps us do then is plug in a player’s interception rate for one year along the x-axis and find out what we should expect it to be the next year on the y-axis. By using this tool, we could have expected Gabbert’s interception rate to rise by a little under 1.0 percent in 2013, showing that his case was clearly an outlier.
However, in the earlier part of the new millennium, passing was not nearly as efficient as it is today. We’ve cut the sample size for our model back to the 2012 to 2016 period, which is still a significant group of players, but specifically ones whose efficiency rates match the current environment we see today a bit better.
Looking to 2018
Again, with a slight adjustment to the trend-line due to more recent efficiency standards, we’re looking at a stabilization point of about 2.7 percent. The table below shows the results when we plug in every potential 2018 starter’s 2017 interception rate.
| Player | 2017 INT Rate | Expected INT Rate |
|---|---|---|
| Tyrod Taylor | 1.0% | 2.3% |
| Alex Smith | 1.0% | 2.4% |
| Tom Brady | 1.4% | 2.4% |
| Case Keenum | 1.5% | 2.4% |
| Jared Goff | 1.5% | 2.4% |
| Drew Brees | 1.5% | 2.4% |
| Carson Wentz | 1.6% | 2.5% |
| Philip Rivers | 1.7% | 2.5% |
| Matthew Stafford | 1.8% | 2.5% |
| Russell Wilson | 2.0% | 2.5% |
| Mitch Trubisky | 2.1% | 2.6% |
| Josh McCown | 2.3% | 2.6% |
| Matt Ryan | 2.3% | 2.6% |
| Eli Manning | 2.3% | 2.6% |
| Joe Flacco | 2.4% | 2.6% |
| Kirk Cousins | 2.4% | 2.6% |
| Andy Dalton | 2.4% | 2.6% |
| Blake Bortles | 2.5% | 2.6% |
| Jameis Winston | 2.5% | 2.6% |
| Ben Roethlisberger | 2.5% | 2.6% |
| Aaron Rodgers | 2.5% | 2.6% |
| Derek Carr | 2.5% | 2.6% |
| Dak Prescott | 2.7% | 2.7% |
| Cam Newton | 3.3% | 2.8% |
| Marcus Mariota | 3.3% | 2.8% |
| Deshaun Watson | 3.9% | 2.9% |
| Trevor Siemian | 4.0% | 2.9% |
This is interesting to see, though it is not entirely useful for fantasy football. What is more useful is looking at quarterbacks from last season who had high interception rates despite good numbers everywhere else, or on the inverse, low ones that masked ugliness elsewhere. For this part, we’re taking their expected interception rates for 2018 and multiplying them by their 2017 pass attempts to see how much they could -- at least in theory -- return to the average.
The table below shows us all the qualified quarterbacks (200-plus pass attempts) we should expect to have lower interception totals in 2018, based on the data.
| Player | 2017 INT Rate | Expected INT Rate | Interception Difference |
|---|---|---|---|
| DeShone Kizer | 4.6% | 3.0% | -7.6 |
| Trevor Siemian | 4.0% | 2.9% | -3.8 |
| Brett Hundley | 3.8% | 2.9% | -2.9 |
| Marcus Mariota | 3.3% | 2.8% | -2.4 |
| Cam Newton | 3.3% | 2.8% | -2.4 |
| Jay Cutler | 3.3% | 2.8% | -2.1 |
| Deshaun Watson | 3.9% | 2.9% | -2.1 |
The most important players to note here are Marcus Mariota, Cam Newton, and Deshaun Watson, who should all drop a little over two interceptions based on their attempts last year. Watson -- who started just six games in 2017 -- was actually on pace for over 21 interceptions last year before his injury, but his expected rate for the season puts him closer to 15. Only Mariota is going particularly late in fantasy drafts right now, but this bit of knowledge should make fantasy owners a bit more certain in selecting him.
The final table shows us quarterbacks whose expected interception rates suggest they should have turned the ball over more and, therefore, could do so this coming season.
| Player | 2017 INT Rate | Expected INT Rate | Interception Difference |
|---|---|---|---|
| Alex Smith | 1.0% | 2.4% | +6.9 |
| Tom Brady | 1.4% | 2.4% | +6.1 |
| Tyrod Taylor | 1.0% | 2.3% | +5.9 |
| Drew Brees | 1.5% | 2.4% | +5.1 |
| Case Keenum | 1.5% | 2.4% | +4.7 |
| Jared Goff | 1.5% | 2.4% | +4.6 |
| Jacoby Brissett | 1.5% | 2.4% | +4.5 |
| Philip Rivers | 1.7% | 2.5% | +4.3 |
| Matthew Stafford | 1.8% | 2.5% | +4.1 |
The law of averages expects Alex Smith’s ridiculous 2017 season to return to a reasonable place, and not just because he’s in a new offense. Simple math shows us that he should have thrown about seven more interceptions than he did last season, while Brady neared that mark as well. That said, their production elsewhere was so good that their ranking from last year wouldn’t have been affected in a big way.
For lower-tiered guys like Keenum and Jared Goff, however, adding almost five interceptions would have taken them from startable territory to bye-week backup for standard fantasy leagues.
Simply by understanding how the average passing play functions, we sharpen our edges over our opponents even more. Will one interception make or break your fantasy season? Absolutely not. But if you’re counting on a player like Alex Smith to sustain a ridiculously low turnover rate and provide safety (at least in that form) at the position, you could find yourself blindsided come the middle of the season.