Written By: Carlo Duffy
If there is one play that defines Super Bowl LII, it has to be the “Philly Special,” one of the gutsiest play calls in Super Bowl history. To remind you of the play’s greatness (coming from a diehard Eagles fan), let me set the scene. It’s close to the end of the second quarter, the Eagles lead 15-12, and they are only one yard away from scoring a touchdown. But it’s fourth down. Eagles head coach Doug Pederson, like he has all season long, takes the aggressive route by going for it. Facing the New England Patriots’ unrelenting offense, he can’t afford to settle for field goals.
To make the stakes even higher, Pederson rolls the dice with the ultimate aggressive play call: a reverse pass from 3rd-string tight end Trey Burton (who has never thrown an NFL pass) to backup quarterback Nick Foles (who has never caught an NFL pass). How riskier can you get?
The Eagles’ gamble ended up paying off. The Pats’ defense was caught off-guard, Foles was left wide open in the end zone, and the Birds extended their lead to 22-12 coming into halftime.
This play turned out to have a huge impact in the game. Foles’ touchdown catch provided the cushion needed for the Eagles to end with an eight-point victory. But what if Philly failed to score? New England would have stolen the momentum, and who knows how the rest of the game would have went.
Whether the play succeeded or failed, I imagined that Philly Special would easily have one of the highest magnitudes of Win Probability Added (WPA) among pass plays by non-quarterbacks. In other words, I’m guessing that of such pass plays, the Philly Special would have highly affected the possession team’s chances of winning, for better or worse. Which led me to wonder: What were the most impactful passes by non-quarterbacks in the 2017 NFL regular season?
Before jumping into the plays with the top 5 highest WPA magnitudes, let’s first explore Philly Special’s impact in Super Bowl LII by looking at the two teams’ Win Probabilities during the game.
This was generated from the Super Bowl play-by-play data from the nflscrapR package in R that was created by CMU alumni (more info here: https://github.com/maksimhorowitz/nflscrapR). Here, the black dashed lines represent the ends of each quarter, and Philly Special is denoted by the purple vertical line set just before the middle black dashed line (which denotes halftime). It looks like my intuition was right: there is a steep change in Win Probability for this play. In fact, the play resulted in a WPA of around 0.127, good for the third highest WPA of all plays in the Super Bowl (#2 was Tom Brady’s deep pass to Danny Amendola late in the first quarter; #1 was Nick Foles’ game-winning TD pass to Zach Ertz in the fourth quarter). No surprises here.
So how does Philly Special compare to non-QB pass plays in the regular season? Before finding out, let’s also explore those regular season plays, whether or not they were successful.
For the next couple graphs, the nflWAR package in R (created by Ron Yurko, check it out here: https://github.com/ryurko/nflWAR) was used to filter out the quarterbacks from the data for all pass plays.
The result of this would be all the pass plays by non-quarterbacks. In the distributions above, blue represents plays that hurt the possession team’s probability of winning (WPA < 0), while orange represents plays that helped the possession team’s probability of winning (WPA > 0). Two observations stand out: (1) the spread for the orange distribution is wider than that for the blue one; and (2) although there are more non-QB pass plays that hurt a team’s winning probability (24) than those that help (14), on average, plays that help have a higher WPA magnitude. In other words, although successful plays appear harder to predict in impact than unsuccessful plays, on average, successful plays had more of an impact than unsuccessful plays. Perhaps this is a sign that the risk of a non-QB pass play failing is not as bad as a coach may think; that is, a non-QB pass play’s reward may outweigh the risk.
I also considered whether a play’s success depends on the distance of a pass.
The same color scheme was used as in the previous figure, with the same criteria for coding. The difference is that these distributions plot plays’ Win Probability Added through the air (airWPA), which gives the hypothetical WPA from the distance travelled in the air. That is, airWPA ignores the play’s actual results to help answer our additional question. Here, there is a good amount of overlap between the two distributions, but noticeably, more successful plays seem to be associated with longer attempts. This may suggest that coaches shouldn’t be afraid to go for the big play when the game is on the line.
During the regular season, many non-QB pass plays had little impact, but here are the top five ones with the highest WPA magnitudes (Spoiler alert, all 5 plays were successes).
[WPA = 0.133] Week 16, Buccaneers vs. Panthers
Panthers punter Matt Palardy’s pass attempt draws a penalty for the first down:
(We couldn’t find video footage, but here’s Panthers head coach “Riverboat” Ron Rivera’s rationale behind the play call.)
[WPA = 0.142] Week 13, Colts vs. Jaguars
Jaguars punter Brian Nortman’s pass results in a first down and allows for a drive-capping touchdown:
[WPA = 0.148] Week 12, Buccaneers vs. Falcons
Falcons wide receiver Mohamed Sanu throws a 51-yard strike to Julio Jones for the touchdown:
[WPA = 0.151] Week 2, Redskins vs. Rams
Rams punter Johnny Hekker’s pass catches the Redskins defense off-guard for a big gain:
[WPA = 0.198] Week 5, Vikings vs. Bears
Bears punter Pat O’Donnell throws a pass for a 38-Yard Touchdown:
Acknowledgment: Thanks to Ron Yurko for his assistance with utilizing nflscrapR.