# Using Air Yards to Better Predict Receiving Yards

The greatest trick the devil ever pulled was convincing fantasy footballers that efficiency is useful.

Unless you are talking about quarterbacks, I can’t really think of a good efficiency metric. When you test them to see if they predict anything, you always find that they come crashing back to the mean the next season like a garbage meteorite.

But maybe that is about to change.

## Efficiency metrics are terrible

The curse of lousy efficiency metrics really rears it’s head when you try and predict receiving yards. Yards per target is one of the worst. It’s about as useful as snake mittens, sporting an objectively shitty year-over-year correlation of just 0.26.

Catch rate is better, with a year-over-year correlation of 0.41. Still, if we multiply catch rate by targets to try and predict Y+1 receiving yards, we do worse than simply using previous year receiving yards.1 But this is a strange approach anyway; why are we using how many times a player does something to estimate how much of that thing they are doing. It’s too indirect. Something is missing. That something is air yards.

## Volume is king

Depending on the sample used, between 72 – 80 percent of a receiver’s same-season production can be explained just by looking at target and air yards volume. If we know how many air yards are thrown at a guy along with how many times (targets), we can make a pretty accurate guess at what his receiving yards will be.

The other 20-30 percent of a player’s receiving yards are explained by catch rate and yards after the catch. YAC matters for receivers who are good at it. Assuming equal air yards, a target to a good YACer is worth more receiving yards than a target to a not so good YACer.2

As noted above, catch rate has a year over year correlation of 0.41 for all WR with over 1000 air yards in a season. YAC yards per game has year-over-year correlation for the same sample of 0.59. So if you wanted to create an efficiency metric that was actually good, you would probably want to combine catch rate and YAC somehow.

If you divide receiving yards by air yards, that is basically what you get. Since receiving yards are just completed air yards added to YAC, when you divide receiving yards by air yards you are capturing a receiver’s ability to convert yards thrown at him and, as a bonus, YAC comes along for the ride.

I call it RACR (Receiver Air Conversion Rate).3 The year-over-year correlation for RACR is 0.52.

Finally, air yards per game multiplied by RACR predicts Y+1 receiving yards as well as simple receiving yards since they are equivalent.4 As I noted above, this is big news. Efficiency metric-based models are typically worse predictors of a stat than simply the stat itself.

## Conclusion

Receiving yards are the result of a function: Air Yards * RACR. By breaking receiving yards down into it’s base components we should be able to better optimize each term.

Air yards are the result of team air yards multiplied by a player’s air yards market share. Since air yards are more stable year over year than receiving yards, we should be able to derive better top down predictions, starting at the team level and working to the player level.

For player efficiency, we can probably optimize RACR using cluster analysis, which is a specialty of our very own RotoDoc. Stay tuned.

If you have a draft today, the actionable advice to take from this article is to target players who have high total air yards and targets. You can find a full sortable list in my previous article.

1. 0.423 r-squared vs.0.446 r-squared  (back)
2. But the overall effect is pretty small  (back)
3. Really wanted this to work out to be RICE, but you can’t have everything.  (back)
4. r-squared of 0.4747 vs. 0.4736 simply due to rounding error  (back)