Matchup Zone: Early Projections

OK, so here's the deal. One of the issues with having stats-based projection models that rely on the point spread, the over-under, and the injury report is that each one of those things change as the week progresses. In the case of the injury report, one can only guess at a player's game status until Friday evening for Sunday games and Wednesday evening for Thursday games. In Houston this week, Ben Tate practiced in a limited fashion on Wednesday, while Arian Foster didn't practice at all. Both backs were limited in practice today. Will they both be listed as questionable come Friday afternoon? Who knows? Meanwhile, the Vegas lines for Sunday night's Texans-Colts game are also hard to pin down. For instance, the sports book at the Wynn Las Vegas opened on Sunday night with Houston favored by 1 point, but they now have the Colts as a 2.5-point favorite four days later. (No doubt that's due, in part, to the aforementioned injury situations.) That's enough to move the projections several tenths of a point for players on both teams.

So, the solution I've come up with is that I'm going to post early projections on Thursday (with minimal commentary), and then post final projections as part of Football Guys' Sunday morning content (with much more commentary). That way, you'll have projections for the Thursday night game if you need them, and you'll have a preliminary idea about the players suiting up on Sunday and Monday. Then, on Sunday morning, you'll be able to see how things have changed since the Vegas lines stabilized and the injury reports came out.

week 9 EARLY tight end projections

I'm going to start with tight ends because this is the first time I list them this season. To come up with a formula, I used the same methods I've used previously for quarterbacks and running backs:

1. Gathering game-related information (as they were at the time) for tight ends who ever ranked in the top 24 in any week from 2007 to 2011.
2. Using linear regression to identify a formula that includes the most predictive factors for weekly tight end scoring.
3. Testing that formula on an analogous data set from 2012.

It turns out that weekly TE scoring can be projected pretty accurately by three things: the TE's average going into the game, the point spread, and the over-under. The higher the TE's average, the more he scores. The bigger a favorite his team is, the more he scores. The higher the over-under of his game, the more he scores. Although each of these were statistically significant predictors, it's worth pointing out that the TE's average is doing most of the heavy lifting here. But that's to be expected given what we know about the randomness of TE scoring (and VBD) compared to other skill positions.

With respect to factors that didn't make the cut, home-road and opponent fantasy points per game allowed once again were not as predictive as the spread and over-under. Surprisingly, injury status wasn't predictive either. Just because a TE is listed as questionable or probable doesn't mean he's going to score below his previous average.  And one final non-predictive factor was the notion of a backup TE being promoted to the starting lineup. We've seen this be important for projecting QBs and RBs, so the finding for TEs may seem odd.

With a little bit of thought, though, I think the answer is fourfold. First, except for the cream of the crop, TEs score far fewer points than QBs or RBs, so any benefit for going from backup to starter would be small, if it exists at all.Second, unlike injury replacements at QB and RB, a new starter at TE isn't as likely to be guaranteed touches, even when that specfic offense features the TE. New England is a good example here. When Tom Brady got hurt in 2008, Matt Cassel was guaranteed X number of passes per game. When Shane Vereen got hurt earlier this year, those touches were guaranteed to go to LeGarrette Blount (and later Brandon Bolden). However, when Rob Gronkowski and/or [murder case pending] were injured (or under arrest) over the past year, the Patriots simply stopped throwing to the TE. Remember Zach Sudfeld? The third reason why being a "new starter" isn't much of a boon for TEs is because teams generally don't stash talent on the bench at that position. It's not rare at all for teams to have a talented backup at QB (e.g., Colin Kaepernick behind Alex Smith last season) or at RB (e.g., Zac Stacy behind Daryl Richardson this season). In contrast, if a backup TE has enough talent to start, he will start. And if there's a nominal starter ahead of him on the depth chart, well, the team will simply use two-TE sets more often (See the 2011 New England Patriots, 2012 Indianapolis Colts, and 2013 Cincinnati Bengals, among others). Finally, you'll recall that the main reason for looking at "new starter" as a predictive factor is that it fills in the blind spots of David Dodds' preseason projections, which I include as a "game" when I calculate the player's average each week. Well, thanks to there being few, if any, TE talents riding the pine unnecessarily, there are far fewer blind spots. Every once in a while, you'll come across an exception like Charles Clay, who's averaging 7.4 points per game after Dodds projected 2.5 points per game. However, a Clay-esque exception is few and far between, to the point that there's no overall benefit leaguewide.

So, without further ado, here are my early TE projections for the top 24 TEs heading into Week 9:

It's been a while, so here's a reminder of how to read the table. Take a look at the row for Jimmy Graham. It says his projection for this week's game is 11.4 points, with a lower bound of 8.1 points and an upper bound of 14.7 points. Because his average going into the game is 15.3 points, that means the system thinks he will score lower than his current average. Finally, the system also thinks that Graham's matchup against the Jets accounts for 3.8% of his projection, with a lower bound of 1.9% and an upper bound of 5.4%. Essentially, what the projection system is saying is that Jimmy Graham is Jimmy Graham, and so he's going to score like Jimmy Graham -- matchup be damned.

Graham's projection provides a nice segue to reiterating the point I made earlier: Unlike the projections for QBs and RBs, the vast majority of TE projections view the matchup as having very little impact. Whereas we routinely see -- and will see very shortly -- the matchup percentage columns around 20% for those positions, only 2 of the 24 TEs listed here even make it to double digits. Again, that's simply a byproduct of how much more a TE's average affects his weekly scoring than does a QB's average or a RB's average.

All of that said, Greg Olsen does have a nice matchup this week, what with Carolina currently being a 7.5-point favorite with an over-under of 43.5.

WEEK 9 EARLY QB PROJECTIONS

Here's the table for the top 24 QBs:

Not a lot in the way of surprises here. Among the elite QBs, Aaron Rodgers and Tony Romo have great matchups as of right now, and that probably won't change come Sunday. Among those farther down the totem pole, Tom Brady is a QB who the system thinks will outscore his average, which is very un-Brady like and has been all season.

WEEK 9 EARLY RB PROJECTIONS

Finally, here's the table for the top 48 RBs:

You will notice the system has Daryl Richardson ranked higher than Zac Stacy. Yeah, ignore that. It's only because Stacy has a chance of being listed as questionable on the injury report, so I included that in the formula as it stands right now. If he ends up getting listed as probable instead, then his projection will be right around Arian Foster's. Such is life in the early projection game!

Aside from that, just like Romo and Rodgers were the projected stars at QB, Eddie Lacy and the newly returning DeMarco Murray are the RBs to have this week.

See you Sunday for final Week 9 projections -- with more commentary.