DFS players, when was the last time you jumped off a cliff? Presumably never, otherwise you likely wouldn’t be here reading a DFS article recommending players to fade.
But for the hell of it, really answer the question: Why haven’t I jumped off a cliff?
“Because it’s stupid.”
Be more specific.
“Because I’m scared I’d probably get hurt.”
“Because why the #@&% should I? Sure it could be fun for a minute but it’d be the last thing I’d ever do. The risk isn’t worth the reward.”
While you probably don’t realize it, your brain just used some variation of the formula that dictates most decisions we all make in our daily lives:
Range of outcomes = σp2 = w12σ12 + w22σ22 + 2w1w2 ρ1,2σ1σ2
This formula is tailored specifically for portfolio management, but it’s a general proxy for quantifying risk. Think of “risk” as in how wide or how narrow is our range of outcomes. Jumping off a cliff without a parachute has too wide of a range of outcomes for any of us. Yes, the ceiling is the thrill of weightlessness, but the floor is…well, sustaining serious injuries when we hit the literal floor.
Compare that to riding a roller coaster. The ceiling is the rush of a high-speed ride, and the floor is maybe feeling a little queasy by the end. It may not match the sensation of an unrestrained free-fall, but you’re also likely to finish the experience safe and sound. Compared to a cliff jump, riding a roller coaster has a lower ceiling and a higher floor. Mathematically, σp2 = w12σ12 + w22σ22 + 2w1w2 ρ1,2σ1σ2 would measure a narrower range of outcomes.
We can’t eliminate risk, but we can control it by adjusting correlation.
Risk always accompanies reward. There is no free lunch. There are no everlasting riskless profits. Risk can be reduced, sliced, diced, and parsed out, but it can’t be eliminated.
Think about it: if you took no risk, you’d never leave the house. Driving could mean an accident. Asking someone out on a date is out of the question because you might get rejected. You couldn’t play DFS because you might lose.
Okay, cool story bro. How exactly does any of this help me win the Milly-Maker again?
ρ1,2. That’s your answer. It’s the component in that crazy formula that stands for correlation. Correlation is the degree to which variables move together, measured on a scale of -100% to +100%.
For example, you walk up to a woman at the bar and ask, “How much does a polar bear weigh? Enough to break the ice.” And the woman walks away. You learn that dad joke pickup lines have about a 100% correlation with rejection, humiliation, and everlasting shame.
The realm of managing assets is a bit more nuanced. Generally speaking, combining assets that have a low correlation with each other narrows the range of outcomes. This process is better known as diversification.
By selecting a combination of players with minimal correlation (i.e., players in different games, no stacks, etc.), it effectively raises the floor of your lineup. The narrower range of outcomes also minimizes surprises. Obviously, this tactic is beneficial for cash games in DFS or if you’re a heavy favorite that week in seasonal.
The Science of the Stack
For GPP plays, we want to do the opposite of diversify, better known as portfolio concentration. We want the highest possible ceiling and we don’t care about the floor. Meaning, we want to widen our range of outcomes by combining assets with a high correlation.
Remember, range of outcomes = σp2 = w12σ12 + w22σ22 + 2w1w2 ρ1,2σ1σ2, where ρ1,2 is correlation. The higher this ρ1,2 value, the larger the value produced by this formula and the wider our range of outcomes. Wider range of outcomes means higher ceiling. It doesn’t guarantee a high score, but it makes it possible.
As you’ve correctly deduced, concentrating a portfolio is akin to stacking a lineup in DFS. Patrick Mahomes and Tyreek Hill’s fantasy point totals are generally going to move in the same direction most weeks (high correlation). Mahomes and OBJ’s point totals, on the other hand, are independent of each other (no correlation).
A key finding in digging into the numbers is that the correlation figure between a quarterback and receiver is driven largely by touchdowns. Yards and receptions are all fine and good, but TDs are the key ingredient to the stack recipe. Intuitively, it makes sense – yards and receptions are generally smoothed stats which are somewhat projectable, whereas touchdowns are random, clustered in frequency, and generate massive spikes in point totals. Clustered touchdowns, like Dalton/A.J. in Thursday’s tilt vs. the Ravens, win GPPs.
Cases in point
- A.J. Green and Andy Dalton’s weekly fantasy point totals since 2014 have shown a correlation of nearly 40%.
- During Allen Robinson’s 2015 breakout season where he amassed 1,400 yards and 14 TDs, the correlation between his fantasy point totals and Blake Bortles’s totals was a massive 50%.
- Brady and Gronk since 2014? 58%.
- Everyone and their brother would consider Doug Baldwin a superior receiver to Jimmy Graham. Last year, Wilson/Graham had a correlation of 58%. Wilson/Baldwin was 40%.
- Yardage and receptions weren’t even close, skewing heavily toward Baldwin. However, Graham had more touchdowns on the season and more games with multiple TDs.
You’ll notice the theme in the fades below that these pairings haven’t consistently generated clustered touchdowns.
Rather than firing blind, calculating correlations between various players and positions gives us a baseline starting point. We can discard the low correlations because as we’ve learned if doing the backstroke inside your own money bin like Scrooge McDuck is your goal, you want the highest ceiling possible.
This week’s Fade Route focuses on three stacks which have a lower ceiling than you’d assume. Fade these three stacks:
DFS stack fade #1: Matt Ryan and Julio Jones (2017 PPR fantasy points correlation = 31.5%)
Much has been made about the offensive dropoff between offensive coordinators Kyle Shanahan and Steve Sarkisian. What’s especially shocking is that the correlation between Ryan and Jones’s fantasy points in 2016 under Shanahan was a whopping 61.8%. That 61.8% figure is among the highest charted in the past five years. The change from Shanahan to Sark cut the correlation figure in half in just one year’s time. For those that skipped the prelude and jumped down to the player fades here – you missed a pickup line that works every time – this is the opposite direction we want to see that correlation figure moving in GPP formats.
As we saw again in 2018’s snoozefest of a season opener, it’s a real struggle to get Julio the ball in the red zone. According to ESPN Stats and Info, Matt Ryan went 1-9 with one interception in the red zone vs. the Eagles last Thursday, including 0-3 to Julio. Flashing back to last year, Ryan targeted Julio 19 times in the red zone, according to Pro Football Reference. The results weren’t pretty: five completions for one touchdown and one pick. That’s a 26.3% red zone completion percentage to his WR1 on the season. Julio’s 2017 season red zone fantasy point total: 14.3 in PPR. Baffling. Last week’s opener vs. the Eagles didn’t inspire much confidence either, with the Falcons pulling Jones off the field at the goal line in favor of the how-is-this-still-a-thing jumbo package.
Michael Salfino of the excellent Breakfast Table podcast postulated that perhaps Matt Ryan just needs to see his receivers open before he’s willing to pull the trigger. That might explain why Atlanta seems to be able to move the ball so easily between the 20s where wideouts can more easily create separation, and yet bog down in the more constricted red zone. So, maybe it’s Ryan, maybe it’s Sark, but regardless, the math says a Ryan + Julio stack in your DFS lineup is a losing bet.
The Cowboys were so bad that it’s probably fruitless to try and gauge the Panthers defense here in week two. If you want a piece of this Falcons offense, put Julio in without Ryan. His yards (and receptions in DraftKings) are still valuable, and a long distance TD is always on the table with Mr. Jones. However, given their red zone woes, this QB/WR combo likely won’t have the touchdown upside needed to justify a stack for the foreseeable future.
DFS stack fade #2: Andrew Luck and T.Y. Hilton (2016 PPR fantasy points correlation = 12%)
We know T.Y. isn’t a touchdown maker – he had six in 2016, Luck’s last full season. But a 12% correlation? Ouch. One would assume with the mind-meld these two share that their point totals would be a little more in sync. When expanding the sample size to examine 2014, when Luck was at the peak of his powers to-date, the correlation between T.Y. and Luck’s fantasy points only rises to 27%. T.Y.’s touchdown total in 2014? Seven.
Interestingly, the fantasy point totals between Luck and Donte Moncrief showed a stronger correlation than Luck and Hilton. Granted, Moncrief played seven fewer games than Hilton in 2016, so it’s always dicey trusting a smaller sample size for statistics like correlation. However, despite missing seven games with a broken collarbone that season, Moncrief still scored seven TDs, outpacing Hilton’s 16-game total of six scores. Again, it illustrates that touchdowns largely determine the ceiling of a stack.
Much like the difficulty in assessing the Panther defense’s true merit with playing the toothless Cowboys in week one, Washington’s win over the neutered Cardinals doesn’t tell us much. Hilton may square off against Josh Norman, but Norman’s not what he once was in his Carolina days (last year was his worst since 2013, according to PFF). It’s not hard seeing Indy racking up yards, but yards don’t win GPPs.
Instead of stacking, as with Julio, consider T.Y. as a solo play.
DFS stack fade #3: Matthew Stafford and…anyone
Finally healthy, Marvin Jones blasted off last season with 225 points on the season in PPR format. Interestingly, his correlation with Stafford’s weekly fantasy points was only about 23%, despite 9 TDs. It’s a curious anomaly, and in diving into the box scores, Stafford appears more the culprit than Jones. Stafford’s best two games book-ended the season, with 27 fantasy points in both week 1 and in week 17. In-between, he topped 20 fantasy points only twice and averaged just 15.67 FPTS per week.
Curiously, Stafford/Tate had a higher correlation than Stafford/Jones, though most of Tate’s production came within that mostly mediocre middle stretch. He also only posted five touchdowns on the season with no multiple-TD games.
The Golladay buzz intensified with Monday night’s game vs. the Jets, and rightfully so. Unfortunately, his sample size of games with Stafford is far too small to draw any statistical conclusions. The oft-quoted stat from the crew at ESPN shows Jones drawing exactly half the targets in games when Golladay plays vs. when Golladay sits (10.2 vs. 5.1 targets). These wideouts can be effective in their own way, but the one who has the clearest path to opportunity simply doesn’t post cluster TD games.
Stafford threw four picks in the home opener last week and sustained what looked like a scary leg injury. The Niners’ D-line is decent, posting 2.5 sacks last week vs. the vulnerable Vikes O-line. Detroit’s O-line profiles as one of the most athletic run-blocking units in the league, but that didn’t translate to protecting the QB in the Monday night outing. Adam Thielen torched the weak Niners linebacking corps last week which should set up nicely for Tate, but again, Tate isn’t a touchdown maker. San Francisco’s secondary did a decent job on Diggs, so it’s unclear if Jones or Golladay (or either) could be expected produce an outsized game on the road.
The bottom line is that with Jones and Golladay potentially offsetting each other, Tate’s low ceiling, and quite frankly, a low ceiling for the post-Megatron Matthew Stafford, any Detroit stack looks dicey this week.
That’s all for the Week 2 Fade Route. What does everyone else think? Let us know in the comments or @BeauNoes.