There was a positive relationship between rush percentage and career EPA per play, meaning that quarterbacks who run more during their rookie seasons tend to have greater success in their careers.

as the number of rush attempts divided by the sum of pass attempts, rush attempts, and sacks. I took the natural log of draft pick because players picked later in the draft have less of a gap between them than players picked earlier For example, the 1st overall pick and 20th overall pick have a sizable gap in talent, but the 200th and 220th draft pick have a far smaller gap in talent. I included this variable in order to capture the overall talent of the player, since players that are picked earlier are more likely to outplay their rookie statistics.

The first was the natural log of draft pick. I included two more variable in the multiple linear regression.

From the scatterplot below, we can see that more aggressive quarterbacks, those with a greater aDOT, tend to have better careers. The average depth of target (aDOT) of a quarterback tells us how far, on average, he throws the ball downfield. Intended air yards per attempt is the same as average depth of target. In addition to completion percentage, I added intended air yards per pass attempt as another independent variable to the regression. For these reasons, I chose to use completion percentage as an indicator for passing ability, even though the statistic obviously has its shortcomings. Another interesting result was that rookie and career completion percentage (r = 0.702) had a higher correlation than rookie and career EPA per play (r = 0.679). The correlation between rookie completion percentage and career EPA per play was 0.596, indicating a moderate positive relationship. Rookie completion percentage had a greater correlation to career EPA per play than rookie yards per attempt, touchdown rate, or interception rate. The first variable I included in the multiple linear regression was rookie completion percentage. Therefore, I carefully chose my variables. Since there were only 43 players in the sample, I had to limit the number of independent variables I would use in any multiple linear regression to a small number. The goal of this project was to determine a quarterback’s career EPA per play based on his first-year statistics. The distribution of EPA per play is approximately Normal with a mean of 0 and a standard deviation of 0.108. As seen from the rankings, EPA per play seems to rank quarterbacks relatively well. As said on ESPN, “EPA knows that not all yards are created equal.” If a team gains 5 yards on a 3rd and 10, they will have a negative EPA, but if they gain 5 yards on a 3rd and 1, the EPA will be positive.įor reference, the top quarterbacks since 2006 ranked by EPA per play and a histogram of quarterback EPA per play are shown below. Taking the difference between the Expected Points after the play and the Expected Points before the play gives the Expected Points Added on a play. It gives a value for the expected points scored by the offense on a drive based on the down, distance to first down, and distance to endzone during a play. Therefore, I opted to use expected points added (EPA) per play to rank the quarterbacks.Įxpected Points Added accounts for the down and distance of a play. While passer rating is helpful and simple, it has many flaws, like giving all plays the same weight. In order to quantify quarterback performance, I had to select a statistic I would use to rank quarterbacks. In total, the sample consisted of 43 players. The filters set on the sample of quarterbacks ensured that each player had an adequate sample of plays in both their first-year seasons and their careers. I chose to use players drafted after 2005 because air yards, a metric that I will use in the final model, was first available in the 2006 season. This sample was then filtered to only include players with at least 500 NFL pass attempts and 3 NFL seasons. The initial sample of players I used included all quarterbacks that were drafted (or went undrafted) in 2006 or later that had at least 100 passes in their first season of play.

Joe burrow stats