Advice

Depth-Adjusted Efficiency — 2016’s Best and Worst

Yesterday I showed which players were 2016’s most- and least- efficient pass catchers from the perspective of yards per target over expectation (YPTOE). YPTOE situationally adjusted a player’s efficiency based on down, distance, and yard line. Today, I will depth-adjust those numbers, because not all passing depths are created equal.

Methodology

The methodology behind this was quite basic. I simply regressed each player’s average air yards per target (aYPT) against YPTOE, to find how far players were above or below the regression line. I call this depth-adjusted YPTOE (da.YPTOE). A more accurate, and better version of this would be to do this on a pass-by-pass basis for each player, rather than with each player’s average depth, but that will have to come in some future research. That will allow us to account for the distribution of target depths along with the down, distance, and field position at which that target depth occurred. For example, a pass 12 yards into the air from the opponents 12 yard line is far different than a pass 12 yards into the air on 1st and 10 from a player’s own 20 yard line. However, today we’ll just work with averages, which still gives us a very strong picture of which players stood out in 2016. YPTOE and aYPT express a near-linear relationship, with YPTOE decreasing as target depth increases. That makes sense, because not only does catch rate decrease, but yards after the catch also tends to decrease with increased target depth. aYPTvYPTOE It’s interesting to see how the shorter depths were more tightly grouped around the trend line than longer depths. A simple explanation is that shorter passes have a more consistent catch rate across players, whereas deeper targets have a much higher variance in catch rate. However, simply taking each player’s da.YPTOE1 doesn’t do us justice. That’s because it’s far easier to put up a big number above expectation at a higher depth, since at the shorter depths things are so much more consistent. While raw da.YPTOE is great for finding actual depth-adjusted efficiency I also looked at something else. Suppose we wanted to ask the question “is Jarvis Landry better at catching his distribution of passes than DeSean Jackson is at catching his distriubiton of passes,” then da.YPTOE wouldn’t get us there. That’s because Jackson has a YPTOE of 1.82, while Landry had a 1.74 value, but it’s far easier to put up a 1.82 da.YPTOE on deeper passes, because the spread is larger. As it turns out, that 1.82 da.YPTOE value is only in the 80th percentile of players with an average target depth of about 13.8 yards. However, Landry’s 1.74 is actually in the 90th percentile of players that have a target depth of approximately 6.5 yards. We can translate those percentiles to z-scores, which I call Z.da.YPTOE in the table below. Both metrics are useful. I think da.YPTOE is better for fantasy purposes, because it answers the question which players provided the most raw yardage value over depth expectations. Z.da.YPTOE tells us who the most efficient players were in their roles. In other words, the latter might be useful for scouting collegiate prospects2 or for evaluating NFL talent in free-agency.3 So, with our depth-adjusted values in hand, let’s see which players topped 2016’s depth-adjusted efficiency, and which ones wound up in the cellar.
  1. Which is calculated by taking the player’s actual YPTOE value minus the trend line YPTOE value for his average depth of target  (back)
  2. If we ever get air yards data for them.  (back)
  3. Even then, as I talked about in yesterday’s article, we’d have to make an adjustment for QB quality as well.  (back)

Subscribe to the best value in fantasy sports

You're all out of free reads for now and subscribing is the only way to make sure you don't ever miss an article.

By RotoDoc | @RotoDoc | Archive

Comments   Add comment

  1. Really surprised to see the big 3 in GB all with 0 to slightly negative numbers, especially considering QB play might help those. Any thoughts on why that might be?

  2. Yes, it is. I have a much larger data set with multiple years. I just showed 2016.

    YPTOE also correlates strongly with @friscojosh's RACR metric, and when we depth adjust both (YPTOE linearly, RACR has a reciprocal relation with depth), they are basically the same thing (R^2 like 0.85)

  3. Yeah, Rodgers didn't have the most efficient year in terms of depth-adjusted efficiency. His super short passes were well above league average (which is why TyMo looks good), but the rest were medicore.

    @friscojosh showed me a graph of depth vs RACR against league average, and basically backs up what my findings found here.

  4. Sorry, this is Rodger's career YAC/catch rate and RACR.

    Picture is the same in 2016 tho, just larger error bands.

Discuss this article on the RotoViz Forums

5 more replies