Identifying 2023 TD Regression Candidates: QBs (Fantasy Football)

Welcome back to the final article in our touchdown regression series, highlighting the quarterback position. This analysis was particularly intriguing given that the past season saw two of the most inefficient quarterback campaigns since 2010. One of those quarterbacks was Kenny Pickett, who recorded a historically low 1.8% touchdown rate. Should we expect another unproductive campaign in Pickett’s second season? Will his efficiency regress to the mean? To answer these questions, we can use the metric “Expected Touchdowns” to evaluate a quarterback’s efficiency. And by identifying regression candidates on both ends of the spectrum, we can determine which players to target and fade in your upcoming fantasy drafts.

If you are interested in a similar analysis for wide receivers, running backs, and tight ends, you can find those on the site as well!

Play-by-play data used for this metric was provided by nflfastR.

Calculating Expected Touchdowns

My Expected Touchdowns (xTD) model uses historical play-by-play data to determine the likelihood of a player scoring on any given play. This is based on a variety of variables, such as the down, distance to the goal, and the type of opportunity (rush attempt, pass attempt, or target). For targets and pass attempts, I also factor in the depth and direction of the throw. Once all of these variables are taken into account, my model will calculate an xTD value for each opportunity on a scale from 0 to 1.

To better understand xTD for quarterbacks, we can take a look at the following play by Josh Allen against the Tennesse Titans in Week 2:

• Type of Opportunity: Pass Attempt
• Pass Direction: Left Side of the Field
• Yards to Go: 10 Yards
• Yardline: Titans’ 46-Yard Line
• Down: 2nd Down
• Air Yards: 46 Yards

This particular play was a deep pass to Stefon Diggs in the end zone. Based on the data points, my model calculates an xTD value of 0.1379. This means that historically, 14% of plays in this specific situation have resulted in a touchdown. Since Allen and Diggs scored on this play, his Touchdowns Over Expected (or TDOE) is calculated as follows:

• Touchdowns Scored: 1.00
• Expected Touchdowns: 0.1379
• Touchdowns Over Expected: +0.8621

As you can see above, TDOE is the difference between the number of touchdowns scored and a player’s xTD value. Therefore, if we run this analysis for every one of Josh Allen’s opportunities, we can determine how many touchdowns he should have scored against the Titans. Based on his 38 pass attempts and one rush attempt, Allen’s xTD value was 2.9 in Week 2. Because he found the end zone four times, he finished that game with a TDOE of +1.1.

Naturally, Expected Touchdowns is a measure of opportunity and usage. Quarterbacks like Josh Allen and Patrick Mahomes are expected to lead the league in touchdowns because they operate in a pass-heavy scheme. Touchdowns Over Expected, on the other hand, is a measure of efficiency, which is both volatile and prone to regression. The assumption is that TDOE will eventually regress closer to zero over time. A perfect example would be Allen’s performance the following week against the Dolphins. Despite leading all quarterbacks in pass attempts in Week 3, Allen scored only twice against Miami (-2.9 TDOE), a significant decline from his efficient Week 2 performance against the Titans.

Interpreting xTD and TDOE

After calculating the Expected Touchdown value for every opportunity, we arrive at the chart above. The x-axis represents a player’s Expected Touchdowns (xTD), while the y-axis shows the actual number of touchdowns scored. The gap between a player and the trend line is their Touchdowns Over (or Under) Expected. The expectation is that if a player receives similar usage next season, they are likely to regress to the line and score much closer to their expected touchdown value. As I mentioned earlier, Expected Touchdowns is a measure of usage and volume. Quarterbacks that find themselves in more pass-heavy offenses will usually finish with a higher xTD value. As a result, it should not come as a surprise that Patrick Mahomes, Josh Allen, and Joe Burrow are the leaders in this metric, operating in offensive systems that prefer to pass the ball at a high rate.

When it comes to touchdown efficiency, how likely is it for a player to regress to the mean? Based on all quarterback seasons since 2010 (minimum 100 attempts), players who scored above their xTD value experienced an 80.1% decline in touchdown efficiency the following year. By contrast, those who scored below their xTD value saw their touchdown efficiency improve by an average of 108% in their next season. In other words, no matter where a player is located on this chart, expect some form of regression this year.

Negative Regression Candidates

Brock Purdy finds himself in impressive company, finishing with the 3rd most Touchdowns Over Expected among quarterbacks with a minimum of 100 attempts. If we convert his efficiency on a per-game basis, Purdy averaged an elite 0.45 TDOE. That would put him on pace to finish with 7.7 touchdowns above expected in 17 games. That would place him well ahead of Patrick Mahomes in the chart above. You might be wondering: can Purdy maintain that type of efficiency heading into 2023? Unfortunately, history would indicate that regression is almost inevitable. Among quarterbacks to finish between seven to eight touchdowns above expected, a staggering 94% regressed to the mean the following year. In addition, 57% finished at or below their expected touchdown value. Of those who managed to maintain positive efficiency, Aaron Rodgers had the highest TDOE in 2015 with +3.9, which was still a 49% decline from the prior season. Along with the concerns of regression, the lack of volume in the Shanahan offense could further limit Purdy’s upside as he was only on pace for 18.9 Expected Touchdowns in 17 games. While I am a believer in the 49ers’ offense, Purdy’s ceiling is likely limited to QB2 production unless he surprisingly maintains or even exceeds his efficiency.

Dak Prescott is another touchdown regression candidate after finishing fourth in Touchdowns Over Expected at +3.9. Similar to Purdy, he was also highly efficient despite missing multiple games. As a result, when we extrapolate his per-game average over a full season, Prescott would have scored 5.6 touchdowns above his expected value. That would rank right next to Mahomes in the chart above. Naturally, that should be a cause for concern as quarterbacks rarely maintain that type of efficiency year over year. However, unlike Purdy who was on pace to average less than 20 expected touchdowns (implying a low-volume offense), Prescott was in a much more favorable situation. In 11 healthy games, his xTD value was at 19.9. If we project that to 17 games, Prescott would have landed at 30.7 expected touchdowns. While that is far from elite, that would still have ranked QB10 last season. Keep in mind that his offensive coordinator – Kellen Moore – has since joined Chargers’ coaching staff. As a result, Prescott enters the season with some ambiguity. However, in a competitive division with one of the better wide receiver corps in the league, I expect Dallas to remain pass-heavy despite the change at OC.

Justin Fields started the year off slow, finding the end zone only three times in his first four games last season. However, starting in Week 5, Fields was one of the most efficient touchdown scorers at his position. In his final 11 games, he scored 3.2 touchdowns above his expected value. What is especially unique about Fields is that he averaged one of the highest xTD values on the ground among quarterbacks. Of his expected touchdown value, 35% can be attributed to his rushing opportunities. That is nearly identical to Jalen Hurts, who was the only quarterback to average a higher rushing xTD than Fields from Weeks 5 to 17. What sets Hurts apart from Fields is the drastic difference in passing volume. Not only do the Eagles average a higher pace of play, but they also pass the ball at an elite rate. In contrast, the Eagles ranked 9th in the league in neutral situation passing rate (56%), while the Bears ranked 29th at only 41%. So while his upside is already within QB1 range, the Bears will need to pass the ball at a higher rate for Fields to reach his true potential. With the recent addition of D.J. Moore, it is entirely possible that could happen as soon as this season, opening up the path for Fields to finish as the overall QB1 in fantasy football.

Positive Regression Candidates

Kenny Pickett had one of the most inefficient touchdown seasons by a quarterback. Not only did he finish with one of the lowest touchdown rates at 1.8%, but he also scored over 12 touchdowns below expected in his rookie year. Surprisingly, there was one other quarterback in my database that finished with a lower TDOE: Carson Wentz in his rookie year in 2016 (-15.9 TDOE). Even more shocking, Wentz’s efficiency completely flipped in the other direction the following year, producing elite QB1 numbers in 2017. While I hesitate to say that Pickett will have a similar trajectory, I am comfortable claiming that he should have a much more efficient campaign in 2023. Historically, we have only seen 10 quarterbacks finish a season with over eight touchdowns below expected in a single season. Every single one of them regressed closer to the mean the following year. Coupled with a very talented receiving corps and the additions made to the offensive line, I would not be surprised to see Pickett improve in his second year.

Justin Herbert had an equally inefficient campaign, scoring 10.3 touchdowns below expected in 2022. Intuitively, this should not come as a surprise as Herbert was 2nd in the league in total passing attempts, but only 11th in total touchdowns scored. It gets even more interesting when we split his production based on his opportunities inside and outside the red zone. Inside the 20-yard line, Herbert was actually one of the most efficient touchdown scorers at his position. He had an Expected Touchdown value of 16.8 and an impressive +3.2 Touchdowns Over Expected in the red zone. Outside of the 20-yard line, Herbert’s efficiency was much less encouraging. He was by far the most inefficient quarterback on those opportunities, scoring –13.5 touchdowns below expected. I expect these numbers to regress to the mean, especially since Herbert had been much more efficient in his first two seasons in the league (+4.9 TDOE). Keep in mind, the Chargers also added Kellen Moore as their offensive coordinator, who recently produced a top 10 offense in EPA per play as the OC for the Cowboys. Factor in the addition of Quentin Johnston and Herbert should be primed for a bounce-back season in 2023.

Despite a PCL sprain forcing him to miss the final five games of the season, Lamar Jackson continued to produce at a high level. When he was healthy, Jackson averaged the 6th most points per game (21.3) among quarterbacks. He was also the QB2 in rushing share at his position, accounting for nearly 33% of the Ravens’ opportunities on the ground. As you can see in the chart above, Jackson could have been even more productive. Surprisingly, his negative TDOE was entirely driven by his rushing production, scoring only three rushing touchdowns despite averaging nearly 10 rushing attempts per game. While I expect his rushing efficiency to improve this season, I also expect Jackson’s usage to evolve with their new offensive coordinator: Todd Monken. In fact, it would not surprise me if his passing attempts reached a career-high this season. Not only did they add Odell Beckham Jr. and rookie Zay Flowers to their offense, but their new OC has historically been a more pass-heavy play caller. In his last two opportunities as an OC, Monken was 4th (62.6%) and 20th (52%) in neutral situation passing rate, which would be an improvement from the Greg Roman era in Baltimore (49%). Therefore, expect this passing offense to be more creative and efficient in 2023, which could vault Jackson back into elite QB1 territory.