Quantifying NFL's 2024 Quarterback Competitions

There are open questions for a handful of NFL teams for Week 1 starters. Looking at those decisions from a Bayesian, range-of-outcomes approach

A lot will be made of tiny, preseason samples for the NFL rookie and veteran quarterbacks engaged in battles to start Week 1 of the 2024 season. We can’t see the more substantial evidence each quarterback provides in practice and training camp, so we’re left to make way too much of a handful of preseason drives. Making predictions based on training-camp clips or preseason play is mostly an exercise in confirmation bias, especially after seeing so much go wrong for quarterbacks who looked decent in the preseason once the NFL games mattered.

What we have for the veterans competing to start across the NFL is much significant and predictive evidence from their play during previous regular seasons. For rookies, you can do worse than simply looking at historical ranges of outcomes tied to draft position, with a discount for typically weaker play as a rookie.

In this analysis, I’ll walk through the quantified ranges of outcomes for the veterans competing for starting jobs in Pittsburgh, Las Vegas, Minnesota, New England and Denver, leveraging my preferred method: Bayesian updating. I won’t belabor the usefulness of Bayesian updating here, but it has been useful for looking at quarterback decisions in the past, and forms the foundation of my weekly quarterback rankings published during the season.

Long story short: we have expectations for expected ranges of outcomes based on draft position coming into the NFL (our prior), and we can use Bayesian statistics to adjust (update) that prior with NFL evidence, thereby forming new ranges of outcomes going forward (posterior). The more evidence we get, the more we believe in the efficiency produced, and the more narrow our expected ranges of outcomes going forward.

PITTSBURGH STEELERS: RUSSELL WILSON VS JUSTIN FIELDS

As of a week ago, the (limited) betting markets for Week 1 NFL starters gave Russell Wilson a better than 80% chance of being the Steelers’ first choice. It appears that a couple wobbly drives from Justin Fields in their first preseason game was enough to essentially shut down the market. According to the ranges of outcomes produced from Bayesian updating, it makes sense that Wilson will get the nod in Week 1.

Wilson shows a more narrow range of outcomes going forward, tied to his overwhelming sample of evidence as an NFL quarterback. Based on what he’s done the last few seasons (which is weighed with a premium in the model versus older performances), Wilson probably won’t recapture the Hall-of-Fame level from earlier in his career. At the same time, Fields has a lower expected efficiency (middle of the range of outcomes), with a lot more downside. A team with a veteran coach and well-established history of defensive strength will likely trade Fields' theoretical upside for Wilson’s stability without blinking.

Fields’ best NFL campaign was two years ago, but even that was only slightly better than Wilson’s worst of 12 NFL seasons, coming during his first year in Denver. Another factor to consider was that Fields’ efficiency in his best season (2022) was largely driven by a small number of outlier rushing plays.

We could see Fields later this season if Wilson struggles, but the Steelers schedule is fairly friendly for whoever wins the Week 1 job, starting @ ATL, @ DEN, VS LAC, @ IND, VS DAL and @ LV.