True Fantasy Points: 2018 Quarterback Projections - Footballguys

A stats-based guide for separating skill from luck in quarterback projections.

It's time once again for our preseason series on True Fantasy Points (TFP), which enters its fourth season on Footballguys. Readers can consult previous years for a more thorough explanation of the methodology (2015, 2016, 2017), but the basic idea underlying the system is that, although most fantasy football players are aware of regression to the mean, most don't know how to adjust projections for small sample sizes mathematically. For example, most know that DeShaun Watson's stratospheric per-attempt stats last season must come back down to Earth, but most simply guess at how much they'll fall. (More on that later.)

With the hard part done (i.e., making the above calculations), the rest is easy. It's just multiplying these "true" rate stats by David Dodds' volume projections in order to obtain stat totals, and then applying Footballguys' standard scoring system to said stat totals. The final result is TFP, which can be thought of as an answer to the question, "How many fantasy points would a player score if we could mathemagically eliminate luck from the equation?"

But why ask the question in the first place? Well, it's so that you can properly value players in your draft(s). A player whose TFP is lower than Dodds' standard point projection is likely being selected earlier than they should be, and vice versa.

Speaking of which, below is a table showing Dodds' projections, TFP, and their difference for 25 starting quarterbacks returning to their 2017 team:

Dodds' Stats True Stats Points
Quarterback
Team Yards TDs INTs Yards TDs INTs Dodds TFP Diff
PHI 3579 26 9 3393 23 12 301.0 276.5 +24.4
DET 4031 26 11 3900 24 13 313.3 295.9 +17.4
LAR 4011 25 12 3773 24 13 304.4 288.2 +16.1
KC 3883 25 15 3764 23 13 304.6 291.6 +13.0
IND 3646 26 14 3527 24 13 306.8 294.3 +12.5
JAX 3469 23 13 3421 21 14 290.1 280.1 +9.9
MIA 3567 23 15 3522 21 13 273.4 265.2 +8.2
OAK 3727 24 12 3591 25 12 282.0 278.3 +3.6
ATL 4184 23 13 3989 25 12 304.2 300.8 +3.4
PIT 4135 29 13 4234 27 14 312.0 310.1 +1.8
NO 4229 29 11 4215 29 13 321.4 320.5 +0.9
DAL 3306 24 14 3433 22 11 298.7 298.4 +0.3
TB 3230 20 11 3284 20 12 250.5 250.9 -0.4
TEN 3377 23 14 3463 21 12 294.4 295.6 -1.2
CHI 3384 21 14 3511 20 13 280.2 282.5 -2.3
NE 4169 29 7 4137 30 9 326.2 329.0 -2.9
SF 4151 24 13 4245 24 14 310.1 314.0 -3.9
LAC 4077 28 15 4179 28 14 302.6 310.0 -7.5
CIN 3550 23 12 3728 24 13 278.0 288.4 -10.4
NYG 3733 24 13 3906 26 16 278.2 290.0 -11.8
BAL 3175 20 13 3459 20 13 238.2 250.4 -12.2
CAR 3437 23 15 3712 23 14 332.7 347.8 -15.1
HOU 3665 24 16 3749 26 13 326.4 341.7 -15.4
SEA 3637 25 13 3975 28 11 330.9 359.3 -28.5
GB 3939 32 10 4339 34 10 355.4 384.2 -28.8

Although the "True Stats" and "TFP" columns contain information you won't find anywhere else, the "Diff" column is where to focus attention for the purpose of identifying value. As the table proceeds from top to bottom, the "Diff" column proceeds from the most overvalued fantasy quarterback to the most undervalued fantasy quarterback according to what their "true" rate stats suggest. For instance, Russell Wilson is the second-most undervalued quarterback because his "true" rates suggests he would score 359.3 points across 513 projected attempts; whereas Dodds' projects only 330.9 points across that many attempts.

THE TRINITY OF 2018 REGRESSION TO THE MEAN

Three of the quarterbacks starting once again for their 2017 team have received particular scrutiny this offseason due to their performances over small sample sizes. DeShaun Watson only started six games, but his 9.3% TD% was so much better than the league average that everyone's identified it as unsustainable in the long run. Ditto Carson Wentz's 7.5% TD% in 2017, although at least we have his 2.6% TD% as a starter in 2016 to more clearly see regression to the mean on the horizon. And finally, there's Jimmy Garoppolo, who had a 2017 sample size akin to Watson's, but whose unsustainable performance was with respect to YPA, not TD%.

All of which is to say that separating the wheat from the chaff among among these three quarterbacks can provide an edge in 2018 fantasy drafts. And from the above table, it's clear that Watson and Garoppolo are the wheat, while Wentz is the chaff. With respect to Watson, after mathematically adjusting for regression to the mean, it turns out his True TD% still ranks fifth among these 25 quarterbacks.. Similarly, if he had qualified for the official rankings, then Watson's 8.3 YPA last season would have outpaced the league leader, Drew Brees. And yet, it turns out that, even after accounting for the small sample, Watson's True YPA only drops to an eighth-ranked 7.6 heading into 2018.

Garoppolo's 8.8 YPA would have outpaced Brees even more so than Watson's if he had amassed a qualifying number of attempts. But because both he and Watson have a similar number of attempts with Houston and San Francisco, respectively, it makes sense then adjusting for sample size results in Garoppolo's True YPA being slightly better than Watson's (7.7, ranked sixth).

On the opposite end of the spectrum, the math shows that Wentz's True TD% heading into 2018 is a 14th-ranked 4.6%, which represents a precipitous drop from his league-leading figure in 2017.

undervalued: Eli Manning

It's odd that both pundits and the public alike have soured so much on Manning, considering that his "dismal" 2017 came with a lame-duck head coach, no running game, and a primary target of rookie tight end Evan Engram rather than perennial Pro Bowler Odell Beckham. To boot, he's only one year removed from a 4,027/26/16 stat line and now has Saquon Barkley as a weapon in addition to Beckham, Engram, and Sterling Shepard. It's not surprising, then, that the math of regression to the mean suggests Manning's 2018 per-attempt rates will hew closer to his True 7.1 YPA and True 4.6% TD% than last season's 6.1 YPA and 3.3% TD%.

overvalued: Matthew Stafford

In 2017, Stafford massively outperformed his "true" stats across his board. Going into the season, he had a True YPA of 7.2, a True TD% of 4.4%, and a True INT% of 2.4%. That turned into 7.9 YPA, 5.1% TD%, and 1.8% INT% over the course of his next 16 games. What goes up is likely to go down. To his credit, Dodds does project a modicum of regression to the mean; just not as much as Stafford's "true" stats suggest.

Bringing these last two sections together, astute readers will have noticed that Stafford's TFP is only 5.9 points higher than Manning's, yet he's currently being selected nearly 70 picks earlier.