San Francisco's Long-Range Forecast
As the 49ers ran for 232 yards and five touchdowns in a 51-13 rout of my beloved Panthers in Week 8 of 2019, all I could think about was the disaster the Panthers run defense had become. They finished the season with an 18.6% run defense DVOA that is fifth worst in the history of the stat. But as the season continued, I noticed that the 49ers kept running all over their opponents no matter the quality of their run defenses, and our weekly projections weren't catching up. Even without a dual-threat quarterback, the 49ers finished second with 2,305 rushing yards and first with 23 rushing touchdowns. And no matter which of Matt Breida, Tevin Coleman, or Raheem Mostert was the team's primary back of the week, he seemed to cap as a projected back-end RB2 with a meager forecasted touchdown total.
Coincidentally, it was that Panthers matchup that helped me solve the riddle. Four of the 49ers' five rushing touchdowns that week came from 10 or more yards away from the end zone. And by the end of the season, the team had an incredible 13 of those long scores. That is an outlier total. The Titans finished second with 10 rushing scores from 10-plus yards away. The Cardinals were third with eight. Only five teams had more than five.
It is important to our projections where players score their rushing touchdowns, because opportunity-adjusted touchdowns are a critical component. When he worked for Pro Football Focus, current ESPN fantasy analyst Mike Clay realized that where a player touched the ball was a better predictor of his future touchdown production than how often he actually found the end zone. That may surprise you because of player differences in size and strength, but it makes perfect sense when you look at the broader trends. Players who take a carry on the 1-yard line convert a touchdown 55.8% of the time, but that conversion rate falls to 39.8% from the 2-yard line and 33.7% from the 3-yard line, and continues to decline until the farther distances where players rarely score at all.
|Rushing Touchdown% by
Distance from End Zone, 2015-2019
Opportunity-adjusted touchdowns work by crediting players with expected touchdowns based on the average rates from whichever yard line they take their carries. No matter if a back weighs 240 pounds or 180, if he takes 10 carries from the 1-yard line, he earns (10 times 55.8%) 5.6 opportunity-adjusted touchdowns.
Bigger backs do score more touchdowns in practice because their coaches prefer to use them closer to the goal line. And it may be the case that that selection bias influences the scoring trends that make opportunity-adjusted touchdowns a better predictor of future scoring. But whatever the reason, they work. Over the last five seasons, opportunity-adjusted touchdowns have better predicted the following year's touchdown totals than actual rushing touchdowns 10 to five (with one tie) for the backs with the biggest differences in those totals and at least 100 carries in both seasons.
|The Biggest OppAdjTD and RuTD Differences and Which Projected Better, 2015-2019|
|Season 1||Season 2|
|Minimum 100 carries in each season|
Note that those projections aren't really projections, per se. They multiply the rates of opportunity-adjusted touchdowns per carry and rushing touchdowns per carry by each player's actual total of carries in the subsequent season. Those future carry totals aren't knowable ahead of time, but they provide the easiest means to illustrate opportunity-adjusted touchdowns' predictive power.
Also note that these cited opportunity-adjusted touchdown totals will be a little bit different than Clay's numbers, even if both looked at rushing alone. I apply some smoothing both across distances and across seasons to help combat small sample sizes.
Taken individually, Coleman and Mostert's opportunity-adjusted touchdown rates of 3.9% and 2.3% last year should presumably predict their 2020 scoring better than their actual rushing touchdown rates of 4.4% and 5.8%. But every 49ers rusher experienced a touchdown surplus last season, and their individual regressed touchdown projections added up to what I suspected was a radical underprojection of 14.0 rushing touchdowns for the 2020 team.
Every team gets a lot of carries from 10 or more yards away from the end zone. The 49ers had 448 of them last year that didn't end in a score, and those carries accounted for 3.9 opportunity-adjusted touchdowns that for most teams would counterbalance their handful of long carries that did find the end zone. But again, the 49ers hit paydirt on 13 long carries last season. And those plays created just 0.4 opportunity-adjusted touchdowns.
|49ers' Long Rushing Touchdowns with OppAdjTD, 2019|
The team did see a touchdown shortfall inside the 10-yard line, but all carries included, the 49ers scored 8.7 more touchdowns than opportunity-adjusted touchdowns would have predicted. That wasn't the biggest surplus -- the Titans had 9.3 extra scores -- but it was one of just four totals outside the band between -5 and +5.
Intuitively, I'm happy with the way our projections handle opportunity-adjusted touchdowns for those other outlier teams. Derrick Henry carried the bulk of the Titans' surplus with 9.5 opportunity-adjusted touchdowns versus 16 actual touchdowns. But you may be surprised to learn that, while scoring at a surplus, Henry provided much smaller surplus totals of 1.6, 0.1, and 2.5 touchdowns in his first three seasons. Yards after contact are not the touchdown cheat code you might expect -- just ask Nick Chubb and Leonard Fournette, who tied for second behind Henry (3.2) with 3.0 average yards after contact in 2019 and still fell short of their opportunity-adjusted touchdown totals. Chubb had 10.9 opportunity-adjusted touchdowns against just eight rushing touchdowns, and Fournette had 8.2 against just three. Both players should see a regression in those differences in 2020, and in Fournette's case that regression should pull the Jaguars out of their 7.7-touchdown shortfall that had them the lone outlier on their side of the spectrum.
The comfort I have in other teams aside, I'm still worried that the 49ers' touchdown luck was more than just luck. Their team surplus was spread over many different players, and all of them share Kyle Shanahan's preferred trait of incredible speed. Mostert's 4.34-second 40 time is second fastest of the 95 players who ran for a touchdown in 2019 and have a combine dash time in my database. Breida was not invited to his combine and Coleman was hurt for his, but they ran 4.38- and 4.39-second 40s at their respective pro days that would have tied them for fourth- and fifth-fastest. Even wide receiver Deebo Samuel ran a top-20 4.48-second time and scored three touchdowns with his legs in his rookie season.
Shanahan isn't a sharer. His 49ers collected 23 of the 93 rushing touchdowns scored by players with 4.48 speed or better last season. Of the remaining 70, wide receivers scored five and Christian McCaffrey, Ezekiel Elliott, and Saquon Barkley scored 33 -- and speed is hardly that trio's sole defining trait. We may not have a sense of whether players like Mostert and Coleman can consistently overachieve their opportunity-adjusted touchdown totals unless they consistently do it in Shanahan's offense over the next handful of seasons.
That said, I think it is an educated guess that they can. As such, during the course of writing this article, I manually increased projected touchdown rates for Mostert, Coleman, and Jerick McKinnon (who ran a sub-4.40 time at his combine and should replace Breida in the 2020 backfield) so that their respective forecasted rushing touchdown totals increased from 5.1, 4.4, and 1.2 to 7.1, 5.2, and 1.6; their positional rankings in PPR formats improved from 26th, 41st, and 78th to 24th, 38th, and 77th. Those changes also bumped the team's projected total to 17.2, which makes for a more palatable decline of 5.8. The 49ers may suffer some plexiglass regression after a single-season jump from 32nd (-22.4%) to 12th (-0.3%) in rushing DVOA. But Shanahan seems like he may have discovered a formula for a consistent surplus of long touchdown runs.