2024 NFL Futures Bets: Season-Long QB Interception Props

Aug 09, 2024

Nathan Ray Seebeck-USA TODAY Sports

Quarterbacks have thrown one or more interceptions in exactly 50% of games over the last three seasons when we filter for at least 10 pass attempts. And when you adjust this number upward for the historical probability of multi-interception games and downward for QB injury rates, you get a baseline expectation of 9.5 interceptions per season. And the most common season-long interception line at DraftKings Sportsbook is…9.5. Just another reminder that oddsmakers are good at their jobs.

Despite this, I’ve had success betting weekly QB interception prop bets. My interception model correctly predicted whether or not a QB would throw an interception 62% of the time last season. I took that model and made some adjustments to it to look at these full-season props. I’ll dig into what variables are in the model, what benchmarks matter, and then highlight a few lines that I think are weaker than the rest.

Pass Rate Over Expectation

This is going to sound obvious, but it’s an important point to make. In a lot of ways, when you’re projecting QB interceptions, you’re projecting a team’s pass rate. More pass attempts mean more chances for an interception. With that in mind, an important variable in the model is pass rate over expectation (PROE). PROE starts by calculating how often a typical team would pass or run in a given situation (i.e. down, distance, score differential, etc.). And then it compares what a team actually did in that situation. Teams that passed more than expected have a PROE above zero, and the opposite is true for teams that passed less than expected.

The graph below is a partial dependence plot (PDP) of QB interceptions. PDPs are a great way to shine some light on the inner workings of machine learning models. They show how each variable affects the model’s predictions while averaging out the effects of the other variables. It’s easiest to focus on the orange line in the graph. As that orange line increases above zero on the y-axis, it means QBs are more likely to throw interceptions. The main takeaway here is interception probabilities start to increase at a PROE of 2.5% and really take off over 4.0%.

So, what teams had high PROEs last year? The graph below shows PROE on the x-axis and weekly fluctuations in PROE on the y-axis. The three teams that stand out are the Chiefs, Bengals, and Commanders. Both Sam Howell and Patrick Mahomes exceeded our 9.5-pick benchmark, while Burrow was pacing to hit that in an injury-shortened season. But we do need to layer on some additional context for these passes. Let’s start with average depth of target (aDoT).

Context on Pass Rate Matters

Passes farther down the field are more likely to be picked off. It’s more difficult to be accurate on these deeper passes, and there is more time for a defender to react to the ball. The graph below is a PDP of QB average aDoT. And it’s a pretty clear trend. As a QB’s average aDoT exceeds eight yards, we see a significant increase in interceptions. So, where did players like Mahomes and Burrow rank in terms of aDoT last year?

The graph below shows completion percentage over expectation on the x-axis and average air yards per attempt on the y-axis. You’ll find both Mahomes and Burrow near the very bottom of the graph, well below our eight-yard benchmark. So, despite their league-high PROEs, both Mahomes and Burrow limit their interception odds by keeping their typical pass depth down. Not to dig too deeply into these specific players, but there are signals that this could change for both players in 2024. The Chiefs drafted the fastest player in combine history, while we’re hoping for improved health with Burrow.

There are three additional variables that provide important pass rate context. The first is a team’s pace. Team’s that snap the ball with a lot of time remaining on the clock generate more pass opportunities. And more pass opportunities mean more interceptions. And pace is most frequently determined by the QB with a few glaring exceptions like offensive coordinator Kellen Moore whose teams always play quickly.

The second contextual variable is accuracy rate. As a QB’s accuracy rate increases, we see fewer interceptions, which shouldn’t surprise anyone. What I found is accuracy rates above 72.5% really start to drop interception probabilities. And the third contextual variable is a QB’s scramble and throwaway rate. Again, we need pass attempts for interceptions to happen. Anything that prevents a dropback from turning into a pass attempt lowers a QB’s interception odds. And so, as a QB’s scramble and throwaway rate exceeds 6% of his dropbacks, we start to see interception rates fall.

Other Variables that Impact the Model

There are four other variables that I wanted to mention before turning to my favorite lines for the season. A really important one is the pressure rate an offense allows. Pressured passes have a higher probability of being picked and a pressure rate above 32% is a key benchmark for interception predictions.

The next variable worth mentioning is Vegas game lines. There is a sharp drop in interception expectations when a team is favored by six or more points. And that makes a lot of sense. The fact that the team is significantly favored might say something about their opposing defense and makes it more likely that they’ll run the ball later in the game to wind down the clock.

And the final two variables are connected. They are a QB’s turnover-worthy play (TWP) rate and interception rate. I include both variables because they provide important context. More specifically, if a QB’s TWP rate is significantly higher than his interception rate, I would expect more interceptions in the future. He might be getting a bit lucky with defenders dropping his would-be interceptions.