A minor projections correction

Just documenting for transparency's sake

When I told you guys in multiple ways that I don’t do nearly enough spot-checking of the projections, I meant it.

I was looking over some of the final numbers as I work on my rankings, and noticed Drake Maye came out pretty low. I’m pretty into playing him as an upside option, but I also knew that during the projection process I felt good about his numbers because the team win total is one of the biggest improvements over last year’s numbers, and they are expected to score a bunch more points.

I was looking at it, and surprised I only had him for 7.0 yards per attempt. I went and looked and saw he was at 6.7 last year, but that still felt like a modest bump given the offense is expected to improve in the market numbers, and also just that he’s going into Year 2 with a probable better offensive scheme and projected improved skill position group (even stuff like TreVeyon Henderson adding catch-and-run explosiveness is a boost to Maye’s YPA).

Turns out I had made a decimal mistake on WR5 Kayshonn Boutte’s efficiency, giving him a RACR of 0.057 instead of 0.57, which equated to a yards per target on his thin slice of targets of just 0.8. But then, while looking at New England more closely, I did bump Kyle Williams a bit (he’s apparently started well in camp, and I probably was a little weak on his initial projection), and also Stefon Diggs, Rhamondre Stevenson, and the two lead TEs just a touch each, to sort of move Maye’s YPA up even more than just the adjustment to Boutte’s typo did. It only got Maye to 7.4 YPA, but that felt better situated among the other QBs that finished in his projection range, and how I feel about those different guys.

This is a great example of how I must have just been feeling a certain way when I did New England’s projection, and wasn’t real aggressive with how Maye might positively influence the numbers for other guys. I do obviously believe that’s possible with his skillset. A funny thing I noticed as well was that despite Maye’s adjusted projection coming out similar to Caleb Williams’, that Williams was still solidly ahead, mostly because of a bit better passing numbers and similar rushing. But I don’t really think of them as similar in terms of rushing ability — I think Maye has real skill there — and when I compared my underlying numbers for those two, I had them both at a regressed 5.8 yards per carry figure. That was fine for Caleb, who was at 6.0 last year, but was a pretty aggressive regression of Maye’s 7.8 YPC last year.

Again, another great example of where I wasn’t consistent. Yes, Maye is unlikely to post a 7.8 YPC again. But I absolutely would want him a half yard better than Caleb, if I’m going to look at their final projections next to each other.

The point here is to say that in each team projection, for the way my sheet builds out, there are a minimum of 104 inputs. Sometimes if there is a QB2 or a deeper skill position group, there are more. Many of those are fairly inconsequential, and I don’t think there are a bunch of mistakes like the Boutte issue because I do have checks and balances in place (that appears to be something I adjusted after I’d already been checking Maye’s final efficiency numbers, but then didn’t catch the typo and how it pushed down Maye’s numbers).

But even if there aren’t errors, there is the possibility that one or more of the teams I did included issues like me projecting Maye for the same YPC as Caleb as rushers. I don’t think either is meaningfully wrong, but as far as a direct comparison, I did want Maye’s rushing to look a little better. So I did also make that adjustment, and I also noticed I had an absurdly low designed run rate for Maye (kneels count toward this, and his designed run rate should have risen merely by the win total rising, so I’m almost certainly still too low), and suddenly I’d shifted some team-level play volume from runs to passes (like 0.2 per game), which also shifted some of the TD distribution from runs to passes. I’ve had to update every sheet on the projections page to account for these minor tweaks to the New England projection, because I just wasn’t as sharp that day or whatever.

My bigger point than documenting that the projections that were not going to change have now changed a bit (but not to account for things like injuries in recent days), is more that there are undoubtedly other little technical issues I haven’t caught. As I scrolled through just now to make sure I didn’t have anything that stood out like Boutte’s 0.8 YPT should have, one other thing I noticed was how I did quickly work through RB receiving efficiency sometimes. You’ll see a lot of RBs with either 5.0, 5.5, or 6.0 yards per target, and a bunch with a 0.1 aDOT, which is just my default. But others do have a real aDOT and some air yards, and what I’m saying is the guys I didn’t project for RB air yards aren’t really in every case players I wouldn’t actually believe have that potential. That’s an example of another area where my projections are almost certainly inconsistent team to team, is what I’m saying.

I would suspect sites that have more than one person working on projections to be able to catch these types of things better. Sites like ETR do an awesome job modeling various things as their starting point so it’s not just starting from some league-wide default for each of the 100+ levers on every team page.

But I’m pretty damn anal, and I’d probably suggest I’m not the only one missing this type of stuff out there, and these types of things influence other peoples’ projections, as well. And the key note here is as I caught these things and made some adjustments, just shifting Maye’s YPA from 7.0 to 7.4 and his YPC from 5.8 to 6.2, plus the small shift in TDs that I guess bumped his pass TDs a little at the expense of a minor amount of team rush TDs that would cost him in a smaller way (it would be divvied up, and cost the RBs the most, but we’re talking about like 0.1 TDs), the net result was that Maye does shoot way up and wind up right next to Caleb Williams, something like six or seven QB slots higher than where he was up until I started making these adjustments.

And that’s the point of this update. If I can find some inconsistencies that — because these projections all come out fairly closely — will shift a guy from looking like an avoid at ADP up to a good value, that’s a pretty clear indication to me to just look at the end results as ballpark, and not exact.

Alright, that’s all for now!