Understanding 2021 Rookie Quarterbacks with Clustering and Bayesian Forecasting: How much confidence do we still have in Trevor Lawrence?

After the 2020 season was a very good year for rookie quarterbacks on the field, the 2021 season marked the return of the usual growing pains for first-year signal-callers. Zach Wilson, Trevor Lawrence and Justin Fields struggled to produce, while Trey Lance hardly saw playing time behind Jimmy Garoppolo. Only Mac Jones enjoyed a strong start to his NFL career, making the postseason as a result.

Among the 31 quarterbacks who were on the field for 500 of their team’s snaps this year, the rookies rank 12th (Mac Jones), 23rd (Justin Fields), 26th (Trevor Lawrence), 28th (Zach Wilson) and 29th (Davis Mills) in PFF grade. Apart from Jones, this doesn’t sound very encouraging, but it’s often not fair to compare a rookie's performance to their experienced peers, as the average first-year quarterback increases their production by almost a standard deviation from Year 1 to Year 2.

Hence, we will look at how the 2021 rookie quarterbacks stack up among first-year signal-callers from the past to get a better understanding of how to evaluate their seasons. We will follow last year’s template and look at the quarterbacks from three different angles:

1. Performance in stable and unstable metrics
2. Quarterback clustering
3. Estimating the distribution of future outcomes 

Click here for more PFF tools:

Rankings & Projections | WR/CB Matchup Chart | NFL & NCAA Betting Dashboards | NFL Player Props Tool | NFL & NCAA Power Rankings
Best Bets Tool

Looking at different metrics

PFF's database contains thousands of data points, but some might be less interesting to look at for rookie quarterbacks than others. Last year, we examined some of our stats for rookie quarterbacks and determined how each statistic in the rookie season correlates to the same statistic in future years.

We ended up looking at the following set of metrics, some of which are stable and some of which are more unstable for the future. A higher R-squared value means the metric is more stable going forward.

Here is how each how the 2021 rookie quarterbacks (minus Lance, who didn't play enough snaps) rank in the stable metrics among rookies since 2006:

And here is how they rank in the unstable metrics among rookies since 2006:

Quarterback Clustering

We will use a clustering algorithm to classify and visualize statistical profiles for the rookies from both a performance and stylistic standpoint. The goal of quarterback clustering is to reduce the complexity of all the information we have about quarterbacks, namely the metrics occurring in the tables above.