In my last four posts, I presented a series of analyses respecting the game-to-game consistency of air yards, or what Pro Football Focus calls average depth of target (aDOT). In a sentence, I found that aDOT stabilizes fastest for WRs (4 games), followed by TEs (9 games), QBs (10 games), and RBs (14 games). For these next four posts, I turn my attention toward the other main variable in the passing/receiving yardage equation, yards after catch (YAC). When I’m done, in addition to the raw reliability stats, we’ll also be able to answer a couple of questions about YAC vis-a-vis aDOT:
- For a given position, is YAC a more reliable indicator of “skill” than was aDOT?
- Is the rank-order of aDOT reliability between positions the same as that of YAC?
Because variety is the spice of life (or perhaps just to build suspense), I’m going to present my YAC analyses in reverse order: TEs first, then WRs, then RBs, then QBs.
Methods
As always, below is the standard procedure I use to determine a stat’s stabilization point; this time applied to TE YAC:
- I collected YAC data for all TEs that played at least 8 games for the same team from 2006 to 2016.
- Starting with TEs that played 8+ games for the same team, I randomly selected two sets of 4 games for each TE and calculated their YAC in both sets.
- I calculated the split-half correlation (r) between the two randomly-selected sets of games.
- I performed 25 iterations of Steps 2 and 3 so that r converged.
- I repeated Steps 2-4 in 8-game intervals, from 16+ games all the way to 72+ games.
- For each “games played” interval, I calculated
- the number of games at which the YAC variance explained, R2, would mathematically equal 0.5. ((The formula is (Games/2)*[(1-r)/r].))
- the “true” YAC for a hypothetical TE that’s had an observed performance of 5.00 YAC through that number of games. ((The formula is [(Observed Performance * Games) + (League-Average Performance * Stabilization Point)] / (Observations + Stabilization Point) ))
- I calculated a weighted average of the results from Step 6. ((Weighted by group size.))
Results and Discussion
Below is the familiar stability table, this time for TE YAC:
| Games | n | r | R2 = 0.50 | Avg YAC | Obs 5.00 YAC |
|---|---|---|---|---|---|
| 4 | 320 | 0.14 | 24 | 4.54 | 4.77 |
| 8 | 197 | 0.27 | 21 | 4.63 | 4.81 |
| 12 | 142 | 0.33 | 24 | 4.64 | 4.82 |
| 16 | 106 | 0.43 | 21 | 4.68 | 4.84 |
| 20 | 80 | 0.49 | 20 | 4.63 | 4.81 |
| 24 | 54 | 0.57 | 18 | 4.69 | 4.84 |
| 28 | 38 | 0.55 | 23 | 4.65 | 4.83 |
| 32 | 25 | 0.59 | 22 | 4.70 | 4.85 |
| 36 | 20 | 0.65 | 20 | 4.77 | 4.88 |
| Wtd Average | 22 | 4.62 | 4.81 |
Right off the bat, the “Wtd Average” row gives us a partial answer to our first question: Whereas TE aDOT stabilized in 8 games, TE YAC takes 22 games to stabilize. Given my sample average of 2.6 receptions, 3.8 targets and 18.2 routes run per game, this equates to about 58 receptions, 84 targets, and 408 routes. That’s … not a good sign. Extending the comparison a bit further by considering the five TE stats I’ve analyzed previously, it turns out that TE YAC is also less reliable than targets per route run (TPRR, 13 games), receptions per route run (RPRR, 14 games), and yards per route run (YPRR, 17 games), but actually more reliable than touchdowns per route run (TDPRR, 41 games). And of course, as I’ve already (repeatedly) revealed it to be the nut low receiving stat, TE yards per target (YPT, 52 games) brings up the rear.
One other result worth noting in comparison to previous analyses respects the Wtd Average YAC for TEs. Namely, if we assume that a TE’s incomplete targets don’t have a significantly different aDOT from their complete targets, ((This is an empirical question, but allow me to go with the null hypothesis until someone runs the numbers.)) then 4.62 YAC vs. 8.14 aDOT suggests that only about 36 percent of a TE’s receiving yards comes from what he does after the catch. This is by no means a mind-blowing finding, but it’s nevertheless a fundamental reality — a base rate if you will — that seems to go underappreciated in conversations about individual TE “skill.”
As always, based on the above results, I’ll end this post with a sortable, searchable table that converts Actual YAC into True YAC for TEs that played at least 25 percent of snaps in 2016 per Pro Football Focus:
| Player | Tm | G | Rec | Actual YAC | Rk | True YAC | Rk |
|---|---|---|---|---|---|---|---|
| Travis Kelce | KC | 16 | 85 | 7.7 | 5 | 6.4 | 1 |
| Martellus Bennett | NE | 16 | 55 | 7.5 | 6 | 6.0 | 2 |
| Rob Gronkowski | NE | 8 | 25 | 8.7 | 2 | 5.8 | 3 |
| Vance McDonald | SF | 10 | 24 | 8.7 | 2 | 5.7 | 4 |
| Levine Toilolo | ATL | 13 | 13 | 11.2 | 1 | 5.7 | 5 |
| Erik Swoope | IND | 9 | 15 | 8.7 | 2 | 5.4 | 6 |
| Lance Kendricks | LA | 16 | 50 | 5.4 | 13 | 4.9 | 7 |
| Brent Celek | PHI | 10 | 14 | 6.6 | 8 | 4.9 | 8 |
| Hunter Henry | SD | 13 | 36 | 5.5 | 12 | 4.9 | 9 |
| Ed Dickson | CAR | 10 | 10 | 7.1 | 7 | 4.9 | 10 |
| Jimmy Graham | SEA | 16 | 65 | 5.1 | 16 | 4.8 | 11 |
| Darren Fells | ARZ | 9 | 14 | 6.1 | 9 | 4.8 | 12 |
| Marcedes Lewis | JAX | 10 | 20 | 5.7 | 11 | 4.8 | 13 |
| MarQueis Gray | MIA | 8 | 14 | 5.9 | 10 | 4.8 | 14 |
| Vernon Davis | WAS | 15 | 44 | 5.1 | 16 | 4.8 | 15 |
| Dion Sims | MIA | 13 | 26 | 5.2 | 15 | 4.7 | 16 |
| Jared Cook | GB | 10 | 30 | 5.0 | 18 | 4.7 | 17 |
| Tyler Eifert | CIN | 8 | 29 | 5.0 | 18 | 4.7 | 18 |
| Josh Hill | NO | 8 | 15 | 5.3 | 14 | 4.7 | 19 |
| Zach Miller | CHI | 10 | 47 | 4.7 | 24 | 4.6 | 20 |
| Luke Willson | SEA | 10 | 15 | 4.9 | 21 | 4.6 | 21 |
| Nick O'Leary | BUF | 10 | 9 | 5.0 | 18 | 4.6 | 22 |
| Greg Olsen | CAR | 16 | 80 | 4.6 | 25 | 4.6 | 23 |
| Anthony Fasano | TEN | 9 | 8 | 4.9 | 21 | 4.5 | 24 |
| Brandon Bostick | NYJ | 8 | 8 | 4.8 | 23 | 4.5 | 25 |
| Virgil Green | DEN | 10 | 22 | 4.6 | 25 | 4.5 | 26 |
| Austin Hooper | ATL | 11 | 19 | 4.6 | 25 | 4.5 | 27 |
| Darren Waller | BLT | 7 | 10 | 4.2 | 31 | 4.5 | 28 |
| Eric Ebron | DET | 13 | 61 | 4.4 | 28 | 4.4 | 29 |
| Jacob Tamme | ATL | 8 | 22 | 4.3 | 30 | 4.4 | 30 |
| Kyle Rudolph | MIN | 16 | 83 | 4.4 | 28 | 4.4 | 31 |
| Garrett Celek | SF | 13 | 29 | 4.0 | 33 | 4.3 | 32 |
| C.J. Uzomah | CIN | 8 | 25 | 3.9 | 36 | 4.3 | 33 |
| Jack Doyle | IND | 16 | 59 | 4.1 | 32 | 4.3 | 34 |
| Will Tye | NYG | 16 | 48 | 4.0 | 33 | 4.3 | 35 |
| Jermaine Gresham | ARZ | 16 | 37 | 3.9 | 36 | 4.3 | 36 |
| Tyler Kroft | CIN | 6 | 10 | 2.9 | 55 | 4.3 | 37 |
| Clive Walford | OAK | 15 | 33 | 3.8 | 39 | 4.2 | 38 |
| Mychal Rivera | OAK | 10 | 18 | 3.4 | 50 | 4.2 | 39 |
| Delanie Walker | TEN | 15 | 65 | 4.0 | 33 | 4.2 | 40 |
| Brandon Myers | TB | 7 | 7 | 2.0 | 61 | 4.2 | 41 |
| Trey Burton | PHI | 14 | 37 | 3.7 | 43 | 4.2 | 42 |
| Jordan Reed | WAS | 12 | 66 | 3.9 | 36 | 4.2 | 43 |
| Coby Fleener | NO | 16 | 50 | 3.8 | 39 | 4.2 | 44 |
| C.J. Fiedorowicz | HST | 15 | 54 | 3.8 | 39 | 4.2 | 45 |
| Gary Barnidge | CLV | 16 | 55 | 3.8 | 39 | 4.2 | 46 |
| Richard Rodgers | GB | 15 | 30 | 3.5 | 48 | 4.2 | 47 |
| Demetrius Harris | KC | 12 | 17 | 2.9 | 55 | 4.1 | 48 |
| Tyler Higbee | LA | 11 | 11 | 2.2 | 60 | 4.1 | 49 |
| Daniel Brown | CHI | 6 | 16 | 2.8 | 57 | 4.1 | 50 |
| Julius Thomas | JAX | 9 | 30 | 3.4 | 50 | 4.1 | 51 |
| Dwayne Allen | IND | 14 | 35 | 3.5 | 48 | 4.1 | 52 |
| Antonio Gates | SD | 13 | 53 | 3.6 | 45 | 4.1 | 53 |
| Jason Witten | DAL | 15 | 69 | 3.7 | 43 | 4.1 | 54 |
| Ben Koyack | JAX | 9 | 19 | 2.6 | 58 | 4.0 | 55 |
| Zach Ertz | PHI | 14 | 78 | 3.6 | 45 | 4.0 | 56 |
| Jesse James | PIT | 16 | 39 | 3.2 | 53 | 4.0 | 57 |
| Dennis Pitta | BLT | 16 | 86 | 3.6 | 45 | 4.0 | 58 |
| Charles Clay | BUF | 15 | 57 | 3.4 | 50 | 4.0 | 59 |
| Ryan Griffin | HST | 16 | 50 | 3.2 | 53 | 3.9 | 60 |
| Cameron Brate | TB | 15 | 57 | 2.6 | 58 | 3.6 | 61 |
I’ve got some space available, so let me take this rare opportunity to highlight some specific nuggets of info elucidated in these True YAC rankings. First, Travis Kelce’s combination of large sample size (i.e., 85 receptions) and fifth-ranked Actual YAC vaults him into the top spot of True YAC for 2016. In contrast, although Dennis Pitta had the largest sample size in the league (i.e., 86 receptions), regressing his 3.6 Actual YAC to the mean results in a lower True YAC ranking (58th vs. 45th) despite the value actually increasing to 4.0.
Second, my current deep sleeper for fantasy football purposes, Erik Swoope, avails himself quite nicely here, just as he did with respect to aDOT. Previously, despite only 21 targets last season, regressing his second-ranked Actual aDOT resulted in a third-ranked True aDOT. Similarly in the current context, despite only 15 receptions and a much noisier underlying stat, Swoope’s second-ranked Actual YAC still only drops to sixth per True YAC after regression to the mean.
DT : IR :: TL : DR
In this first post on the reliability of YAC, I showed that it takes 22 games (or 58 receptions or 84 targets or 408 routes) to represent 50% skill and 50% luck for TEs. This is nearly three times as long as it takes for TE aDOT (aka air yards) to stabilize, as well as nearly twice as long as both targets and receptions per route run.