NOTE: The pro projections below are first posted on Thursday, then updated on Saturday afternoon. On Thursday, commentary will be limited to early impressions. A more-detailed discussion of specific players will be added on Saturday afternoon.
Welcome back for Week 3 of Footballguys' new feature that uses a statistical model to project the roster percentages of "professional" tournament players on Draftkings. If you're curious about what goes into the model, click here.
As promised, this week's article will start off with a review of last week, identifying the model's hits and misses, as well as theorizing why the happened. For Week 2, actual roster percentages are based on lineups the pros entered in the low buy-in featured tournaments this author also entered (as well as the Millionaire Maker).
week 2 hits
Based on an 18.3% probability of achieving 4x value and the Browns defense being projected for less than one percent ownership, the model projected Alvin Kamara to be rostered at 18.5% by the pros, which was a lower rate than projected for the overall public. And indeed, Kamara was actually rostered in 18.6% of pro lineups, for an absolute error of 0.1%. This bit of depressed rostering paid off, as Kamara scored 17.9 points, which translates to only 1.9x value at a salary of $9,500.
Another running back that whose projected pro roster percentage was nearly a bullseye was Todd Gurley. The model projected him at 17.6%, which was far lower than what was expected of the overall public. Gurley ended up being rostered in 17.2% of pro lineups in this sample. Unfortunately, the pros didn't benefit as much on this call as much as they did on Kamara, as Gurley's touchdown-heavy performance produced 32.3 points, which translates to 3.5x value at a salary of $9,200.
The model also had two highly accurate projections at wide receiver. First, Quincy Enunwa appeared in 11.1% of pro lineups, whereas he was projected at 11.4%. Second, and far more surprisingly given how much he lapped the field in projected overall roster percentage, Antonio Brown appeared in 38.8% of pro lineups after being projected at 39.7%.
NOTE: So there was a bit of a snafu with the tables posted on Saturday. The "Overall" projections listed were actually Steve Buzzard's FanDuel roster percentages, not his Draftkings roster percentages. And since the pro projection model is based on this data, the "Pro" projections were also incorrect. Fortunately, after rerunning the model, it turns out that, except for Vikings defense, the player commentary/recommendations would have been the same regardless of this error. We have amended the tables to provide evidence that this is true. The "Original" columns are what was posted on Saturday. The "Corrected" columns are what the roster percentages should have been. Most importantly, we regret the error. Rest assured that it will never happen again.
week 2 misses
Model projections for quarterback, tight end, and defense were where most of the big misses occurred.
Starting at quarterback, at 12.3%, Ben Roethlisberger was projected for far higher rostering than the overall public, but he instead came in at 8.8% among pros. Of course, with Roethlisberger scoring 41.98 points for 6.1x value, pros almost certainly would have rostered him more if given a second chance. The same can't be said for the second miss at quarterback, Patrick Mahomes II II, who ended up in 9.3% of pro lineups, which was higher than the model's 7.7% projection.
One of the model's biggest misses of the week at any position was David Njoku. Specifically, he was projected for 5.4% among pros, which was only slightly higher than what was expected for the field. However, he ended up being rostered at a rate of 15.2%! This was likely due to late-breaking news about the impending absence of Josh Gordon. To wit, Jarvis Landry was also a huge miss -- 6.0% projected, 19.8% actual -- for likely the same reason.
Zach Ertz was another order-of-magnitude miss at tight end (3.8% projected, 8.8% actual), although this one is harder to explain. Perhaps it was based on Pro Football Focus' matchup chart, which gave Ertz a large advantage over Buccaneers safety Chris Conte. And what's even more frustrating about not having a ready explanation for this miss is that it paid off handsomely for the pros -- to the tune of 3.3x value (20.4 points at a salary of $6,100).
Finally, at defense, the Houston Texans appeared in a whopping 24.4% of pro lineups despite both the pro projection and overall projection being approximately 8.5%. But again, late-breaking news provides a likely reason: Blaine Gabbert was getting the start for Tennessee.
Now, let's move on to pro roster percentages for Week 3.
Below is the table for quarterbacks. In this and all subsequent tables, "Overall" is Buzzard's projection of roster percentage. "Pros" is the model's projection of roster percentage for Draftkings pros only.
Thursday Early Impressions
Matt Ryan and the two quarterbacks in Rams-Chargers are shaping up to be quarterbacks the pros will roster more frequently than the overall public. On the other end of the spectrum, Russell Wilson and Jimmy Garoppolo figure to be rostered less frequently. Although, in the case of Garoppolo, that's likely to change given the high -- and climbing -- over-under for Chiefs-49ers.
Andy Dalton ($5,700, 1.4% Overall Roster Percentage, 2.7% Pro Roster Percentage)
The pros being projected to roster Dalton at twice the rate of the overall public is due to factors unrelated to Dalton himself. Namely, the combined roster percentage of his wide receivers is 17.3%, which ranks eighth in the Sunday Main slate. Also, the Panthers defense is projected to appear in only 1.5% of rosters, which ranks fifth-lowest. If Dalton's wide receivers are going to do well and Carolina's defense is going to do poorly, it stands to reason that Dalton should be rostered more frequently.
Jared Goff ($5,900, 5.0%, 5.3% Pro Roster Percentage)
Goff is in an even better position than Dalton with respect to the model's other-position factors. He's one of six quarterbacks facing defenses projected to appear in fewer than one percent of GPP rosters. And at 30.9%, Goff's receivers have the second-highest combined roster percentage projection. Unlike Dalton, Goff also ranks ninth among quarterbacks in terms of his probability of achieving 4x value, at 23.9%.
Below is the table for running backs:
Thursday Early Impressions
Unlike previous weeks, the pros are pretty much in line with the public near the top of the rankings for Week 3. Farther down the list, David Johnson and Dalvin Cook have a large negative difference, the former presumably because of Arizona's overall offensive ineptitude and the latter because of his injury status. Besides those two, a couple of pass-catching No. 2 running backs near the bottom of the list have higher pro percentages than overall percentages: Austin Ekeler and Javorius Allen.
Corey Clement ($4,300, 21.2%, 26.5% Pro Roster Percentage)
At 41.2%, Clement has the highest probability of achieving 4x value among running backs on the Sunday Main slate. He also benefits in the model from facing a Colts defense projected to appear in less than one percent of lineups. If that isn't enough, he's also playing at home as a touchdown favorite with little, if any, competition for touches in a game with an above-average over-under against a defense that's allowed the eighth-most Draftkings points to opposing running backs, including a league-worst 11 receptions per game. Apologies for that run-on sentence, but Clement's matchup is that good and ticks all the boxes according to the model.
Dion Lewis ($4,500, 1.1%, 3.9% Pro Roster Percentage)
Lewis doesn't have a matchup nearly as great as Clement, but his 26.8% value probability does rank fourth on the slate. Also pivotal to the model projecting pros to roster him nearly four times as frequently as the overall public is that Tennessee's wide receivers are projected to appear in only 1.7% of rosters combined. As a 9.5-point underdog, the Titans offense will likely devolve into pass-heavy mode by the third quarter, thereby elevating Lewis into an every-down back role. And with Jacksonville's defense being stingier against opposing wide receivers than opposing running backs, Matt LaFleur's offensive game plan just might call for Lewis to assume said role even earlier.
Kareem Hunt ($6,000, 15.0%, 9.2% Pro Roster Percentage)
Kansas City is a 6.5-point home favorite, so you'd figure Hunt would not be a "fade" candidate amongst the pros. Unfortunately, he ranks dead last among running backs according to one factor important in the pro projection model: projected roster percentage of his team's wide receivers. Specifically, Tyreek Hill (28.8%) and Sammy Watkins (11.2%) combine for the second-highest combined projection so far in 2018 (behind only Pittsburgh's 44.8% in Week 2). And this doesn't even include Travis Kelce's 16.5% overall roster projection! It also doesn't help Hunt that the 49ers Front Seven is stronger than its secondary, and that said unit will have Reuben Foster back in the starting lineup on Sunday.
Below is the table for wide receivers:
Thursday Early Impressions
Julio Jones, in a likely Ryan-Jones stack, is chalk the pros are likely to be hammering this week. Devin Funchess, in contrast, isn't catching the pro's fancy. Corey Davis is a No. 1 wide receiver who the pros are poised to roster at a rate nearly twice the field. And as was the case with running backs, a pair No. 2s are likely to see a boost in roster percentage among the pros: Ted Ginn Jr and Calvin Ridley. Man, that Falcons-Saints game sure does seem to be popular according to the pro projection model!
Calvin Ridley ($3,700, 1.7%, 4.4% Pro Roster Percentage)
or Mohamed Sanu ($3,800, 2.6%, 4.1% Pro Roster Percentage)
Let's start with why the projection model gives these secondary Atlanta receivers a boost. First and foremost, they'll be facing a Saints defense projected to appear in less than one percent of GPP lineups. Second, Ridley has the third-highest 4x value probability on the slate (24.7%), while Sanu ranks 13th (17.5%).
Outside the model, two important factors which no doubt are driving their point projections are a) the Saints' dead-last ranking in Draftkings points allowed to opposing wide receivers; and b) the game's over-under, which is the second-highest on the slate. Not that it's advisable to read too much into computer algorithms, but what the model appears to be telling us is that Ridley and Sanu are ideal pivot plays away from Julio Jones should one choose to fade him.
Corey Davis ($4,400, 1.7%, 4.0% Pro Roster Percentage)
What Davis clearly doesn't have going for him is his matchup -- on the road against Jacksvonville's stout secondary. What he does have going for him, however, is the fact that his 4x value probability is a position-best 26.0%, and he is the only wide receiver projected to be rostered in more than one percent of GPP lineups. In short, the combination of his salary and his massive, Julio Jones-esque market share of air yards means that he -- alongside the aforementioned Dion Lewis -- is likely to accrue the lion's share of Tennessee's receiving stats in a game they're expected to be playing from behind the entire time.
Odell Beckham Jr Jr ($8,200, 10.3%, 6.5% Pro Roster Percentage)
Speaking of a massive market share of air yards, Beckham's also runs circles around his fellow Giants receivers. However, the difference between him and Davis is their cost. That $3,800 difference translates to Beckham ranking 46th in value probability among wide receivers rather than something closer to Davis' 1st. Otherwise, what lowers Beckham's standing per the pro roster percentage model is that Houston has the fourth-highest overall projection among defenses. This stat is also worse than Davis', as Jacksonville's defense ranks lower in overall projected roster percentage (6th).
Below is the table for tight ends:
Thursday Early Impressions
Continuing the theme, Austin Hooper is another player in Falcons-Saints game likely to be favored by the pros. In addition, it appears Eagles-Colts is shaping up to be a tight end friendly contest, what with both Jack Doyle and Zach Ertz registering as plus plays per the pros. Of course, Doyle's game status is in question at the moment, so his projection is likely to change between now and Saturday.
Eric Ebron ($3,400, 23.5%, 28.4% Pro Roster Percentage)
If there's any overarching theme astute readers have noticed while perusing the tables and reading the commentary these past few weeks, it's that the pros are drawn like magnets to situations in which a low-priced player will have a veritable monopoly on touches at his position due to injury. Hello, Eric Ebron. In terms of the projection model, Ebron (unsurprisingly) has the highest 4x value probability in the slate (38.7%) and is facing an Eagles defense (somewhat surprisingly) projected to appear in only 3.2% of lineups. Also working in his favor is the dearth of weapons at wide receiver in Indianapolis: They combine for only 12.4% rostering in aggregate, which ranks in the middle of the pack and is due almost entirely to T.Y. Hilton's 11.5%.
Zach Ertz ($6,800, 11.2%, 6.4% Pro Roster Percentage)
Ertz has gone from an early-week positive play to a late-week negative play for two reasons. First, his projected roster percentage for the overall public rose more than doubled from Thursday's 4.6%. Second, ditto the projected roster percentage for Eagles wide receiver Nelson Agholor (from 10.1% to 21.0%). With Ertz' 4x value probability staying essentially the same (11% on Saturday vs. 10% on Thursday), these developments were more than enough to massively reduce his projected roster percentage among the pros. And oh by the way, it also doesn't help that the weakness of the Colts' defense so far this season has been defending pass-catching running backs, not tight ends.
Below is the table for defenses:
Thursday Early Impressions
It's a wonder how the Raiders are projected to be so frequently rostered overall in a cross country road game against a Dolphins offense that hasn't been awful. And sticking with that game, it's a wonder how Miami's defense is projected to be so infrequently rostered when playing at home against a cross country opponent whose offense has been awful.
Minnesota Vikings ($4,300, 21.4%, 23.8% Pro Roster Percentage)
Because defensive scoring is so random from week to week, the public and pros alike tend to prioritize value at the position. And indeed, 4x value probability is why the overall public figures to roster the ultra-cheap Dallas defense so heavily in Week 3. This reality is why Minnesota's projection for pros is an anomaly worth highlighting: They're the chalkiest of chalk despite ranking seventh-lowest in value probability (29.0%). That's because they're in the platonic ideal of a matchup situation with respect to other factors, both in and outside the pro projection model. Within the model, they're going against Josh Allen at quarterback, who has a roster percentage projection less than one percent. Outside the model, Minnesota's defense is playing as a 16.5-point home favorite going against one of the consensus two-worst offenses in the NFL.