Read on Twitter
THREAD:
EPA/target is often used as the go-to efficiency metric for receivers and I don’t know if that is right. I’d like to float my alternative and see what thoughts people have.
EPOI (Expected Points Over Incompletion) is expected points added compared to an incomplete pass.
EPA/target is often used as the go-to efficiency metric for receivers and I don’t know if that is right. I’d like to float my alternative and see what thoughts people have.
EPOI (Expected Points Over Incompletion) is expected points added compared to an incomplete pass.
The denominator I’m using right now is team dropbacks. My goal in coming up with this was to contain all of the calculation within @nflfastR. Team dropbacks is probably the best *public* estimate of the number of opportunities a receiver has to make a catch.
I’ll start by defining my issue with EPA/target. When a receiver runs a route, one of the following basic events can occur:
1) reception
2) targeted, but not complete
3) pass thrown at a teammate
4) no true pass attempt (sack, scramble, throwaway, etc)
1) reception
2) targeted, but not complete
3) pass thrown at a teammate
4) no true pass attempt (sack, scramble, throwaway, etc)
EPA/target only takes into account only the first two groups. As a route runner, this suggests that you were at least open enough that the QB threw the ball to you, which probably means you did something good (or at the very least, not bad)
For #1, despite getting targeted and making a catch, a receiver could still be assigned negative EPA if they did not earn enough yards on the play. This could be because the pass was a short throw/checkdown and the pass catcher was schemed into this role pre-play.
#2 will always result in negative EPA. This punishes players for either being open, being targeted while covered, or being the focal point of the play. One could argue that none of these situations are explicitly better than not being targeted at all (#3 & #4).
My thought was that if receivers are not punished for plays where they cannot get targeted, they should not be punished for plays where they are targeted, but cannot complete a catch. One could resonably disagree with me on this point.
Furthermore, I wanted to make it so that a catch for positive yardage would not result in a negative value. To do this, I calculated the EPA of an incomplete pass on a play and subtracted it from the eventual EPA on plays with receptions. This is EPOI (Exp Pts Over Incompletion).
Turnovers are a complication. I felt that on a fumble the receiver should incur the impact from the loss of possession, but not from the fumble return. For EPOI, I assume the opposing team took 1st & 10 at the spot of the lost fumble. INTs are treated as incomplete passes.
Turnovers on downs make 4th down EPOI extremely valuable since the EPA for an incompletion on 4th down includes loss of possession. Obviously 4th down receptions are important, but I’m concerned that I might be over-stating their value with this.
Here is year-over-year stability for players who were targeted 50 times in consecutive seasons. EPOI/Team Dropback is more stable than EPA/target, but isn’t meaningfully more stable than average yards per game.
I will say that EPA/target may be useable for pass plays like quick reads or screens where the targeted player is unquestionably the focal point of the play. Those plays represent a more clear link to production and invested volume.
Everything here comes with two important caveats:
1) The public does not have great tools in general for pass catcher evaluation.
2) EPA is not a player metric, so applying any kind of expected points analysis that seeks to apply value to a single player is leaving out a lot.
1) The public does not have great tools in general for pass catcher evaluation.
2) EPA is not a player metric, so applying any kind of expected points analysis that seeks to apply value to a single player is leaving out a lot.
