SportGrid Radio SportsGrid
NFL Free

DraftKings NFL Strategy: Player Level Stacking and Correlation

DraftKings NFL Strategy: Player Level Stacking and Correlation
Print Friendly, PDF & Email

[chatroll width=’100%’ height=’450′ id=’vMyVcgVwfmI’ name=’dailyroto-premium-football’ apikey=’niyjl891voszofbm’]


The most basic DraftKings NFL strategy is stacking. If you need me to explain what stacking is then you may have come to the wrong place. There are a lot of introductory articles you can read about stacking for free, what we are here to talk about today is a little more intricate. Last preseason I wrote a lot about this topic. I look at single stacks, double stacks, triple stacks. Last NFL season I was more stacked up than Floyd Mayweather heading to a craps table. And it worked to a degree. I wouldn’t say I definitively proved that a quarterback pricing goes up, you should increase the number of players you are stacking with him, but I felt like it was close enough. My favorite go-to stack on DraftKings for NFL DFS was the Quarterback – WR1 – WR2 – Opponent WR because I am a sucker for correlated plays and love a YOLO game stack. Many high stakes regulars also deployed similar tactics to win or attempt to win GPPs. That said there were certainly times as I loaded that combination into an optimizer to fire off some teams that it felt a little too much like a blanket rule I was following. In fact, much of the stacking narrative and articles written seem like a bunch of blanket rules and don’t necessarily take specific circumstances into account. The stacking conversation needs to change.

Correlation and Upside

What we are seeking in stacking is the presence of both correlation – our players’ scores increasing together (ideally at a high rate) – and upside – the ability to achieve a high score. The presence of correlation without upside in a stack presents little value, and the presence of upside with low correlation also does not benefit us in GPPs as part of a stack. The popular industry correlation matrices reference correlation between different positions. This is an elementary approach that doesn’t take into account player level nuance. To illustrate this let us look no further than the proficient pass offenses in Green Bay and New England, where quarterbacks Tom Brady and Aaron Rodgers are generally going off draft boards as the top players of their position. Many DFS players view the two as similar but the underlying numbers from last season are pretty different.

Grouping Brady Rodgers
Games Above 20+ 8 11
WR1 Games Above 20+ 3 7
WR1 Correlation 0.26 0.21
WR2 Games Above 20+ 1 5
WR2 Correlation 0.39 0.41
TE1 Games Above 20+ 2 1
TE1 Correlation 0.08 0.08

As you can see from the chart above only half of Brady’s “explosive games” were followed by an equally explosive game from one of his top weapons. Aaron Rodgers, on the other hand, boasted fairly similar correlation numbers but fed explosive games to his top weapons with more frequency. Despite similar correlation numbers, holding price, ownership, and match up equal, Aaron Rodgers is a better GPP target than Tom Brady because the ability to stack him offers more predictable upside. This take seems obvious, while the 2nd level of thinking shows that Brady is a quarterback who you can consider going more contrarian if you are going to stack him because he has shown the ability to spread the ball around.

So if despite the proficient passing attacks, Brady and Rodgers offered varying degrees of predictability, it certainly warrants a deeper dive at the position on a player level. In the chart below I identified the # of games each team’s “QB1” (as defined by total fantasy points) hit 20+ fantasy points, along with the corresponding values from the top options at each position.

The exercise left me with several takeaways heading into the season, some of which may be intuitive and others a little more nuanced.

  • Target Monsters Julio Jones, Antonio Brown and Mike Evans should always be paired with their respective quarterbacks. Julio Jones boasted a 0.58 correlation with Matt Ryan while Antonio Brown showcased a 0.51 with Ben Roethlisberger and Mike Evans a 0.54 correlation with Jameis Winston. The upside moved in parallel and while adding secondary pieces to the attack was an option you were rarely getting a big game from the QB without one of these targets. This especially holds true for DraftKings PPR scoring format.
  • Quarterbacks with rushing upside can be run naked or as solo stacks. While Dak Prescott offered a good degree of correlation with Dez Bryant, options like Tyrod Taylor and Colin Kaepernick had an individual upside that outweighed the upside of their passing attacks. When you combine that with a possibly cheaper price tag these are opportunities to gain leverage when a quarterback hits value but does it with his legs.
  • Offenses with diverse attacks warrant closer consideration. The attacks lead by Drew Brees and Tom Brady featured diverse performances even when the passing game was going off. The same can be said to a lesser degree with Kirk Cousins in Washington. These situations make the stacks less valuable because they are less predictable. The exception to this rule is if also causes deflated ownership on individual skill players.

Interesting Correlation News and Notes

While the upside is critical some of that also comes with variance. By understanding team and player level correlation for daily fantasy football, we can move past simple matrices. Below are some of the specific spots I found with interesting correlation numbers from the 2016 NFL season. You can view all the correlation at a positional level here.

  • Matt Ryan and Julio Jones (0.58) and to a lesser degree TE1. It will be interesting to see if that changes in a new system.
  • Cam Newton and Greg Olsen (0.48) and surprisingly Devon Funchess (0.51).
  • Andy Dalton’s surprisingly low correlation to AJ Green (0.25) last season
  • Terrell Pryor was a target monster (0.57) but will his role remain the same on a new team or mirror the spread out attack in Washington? Will Cleveland lock into a new primary receiver or are the personnel changes a changing of the guard?
  • Dak to Dez (0.59) or with his legs was basically the only way to look. This will be especially interesting with the Ezekial Elliott suspension during the first few weeks of the season.
  • Emmanual Sanders again (0.65) or a Thomas bounce back (0.29) coming?
  • Evenly split action in Detroit at WR1 and WR2 implies there is no clear favorite target for newly minted Matthew Stafford
  • Andrew Luck loves TY Hilton
  • Allen Robinson still featured in Jacksonville (0.40) but will target volume decline in the new offense?
  • Alex Smith offers minimal upside and struggling correlation, even when Tyreke Hill or Travic Kelce were having big weeks
  • Adam Thielen (0.52) the highest correlated Minnesota option surprised me to be sure
  • Odell Beckham (0.75) the key to Eli Manning’s success and highest correlated wide receiver in the NFL
  • NY Jets predictable (0.69 and 0.59) but in total rebuild at receiver with the departure of Decker and Marshall
  • In Oakland it was Derek Carr – Amari Cooper – Michael Crabtree double stacks for life. I’m all about that life.
  • Phillip Rivers loves him a WR1 (0.70) 
  • Russell Wilson was able to be run naked, with Doug Baldwin and not much else
  • Never stack Kaep, if he gets a job. Which he should, but we won’t go there. 

Hopefully, this article helps keep things fresh and remember not to follow blanket rules when creating teams, even if leveraging optimizers and correlation matrices. Each NFL team approaches game plans in their own unique manner, and so should DFS players. That is it for this week’s DraftKings NFL strategy, hit me up on twitter if you want to see anymore data.

@Drewby417 out.

NFL Free