# The Most Undervalued NFL DFS Correlations

Last week we released our new NFL Correlations page, which allows users to visualize positional and statistical correlations. Users can also view correlation data for specific seasons and teams, and they can even research how correlations change based on different Vegas data points like totals and spreads. Check out my tutorial video.

### Creating a Correlation Value Metric

I think the most valuable information gained from this new tool is not how correlated two positions are; it’s how owned two positions are together. Almost 100 percent of daily fantasy football users know that stacking a quarterback and his No. 1 wide receiver is advantageous and raises a lineup’s ceiling. They may not know it’s called “stacking” or that the exact r-squared value (the correlation) is 0.47 — but they know it’s a wise strategy. And the page confirms this: The tournament ownership correlation between those positions is 0.56, which is even higher than the r-squared of fantasy points. It’s useful and still probably wise to stack, but everyone knows about stacking.

That said, there are probably correlations that are less known by the public, or at least undervalued by the public. And since we have r-squared values for fantasy points, Plus/Minus, and tournament ownership, we can devise a formula to measure these discrepancies. (Also, Pro subscribers can review ownership trends across guaranteed prize pools of various buy-in levels via our DFS Ownership Dashboard.)

First, we need to put the correlation values on a non-negative scale; we’ll use 0-100 to make it easy. Here are the correlation values for actual fantasy points, per the NFL Correlations page . . .

. . . and here’s the same table scaled from 0-100:

Now we can do the same thing for both Plus/Minus correlations, and more importantly, tournament ownership correlations. For the purpose of the exercise, I’m giving low tournament correlation a high rating and vice versa. Basically, what we’re looking for is positional correlations that have a high correlation in terms of fantasy points and Plus/Minus but low correlation in terms of tournament ownership; the higher the rating the more undervalued the correlation would be.

Using that rating, here are the most undervalued positional correlations for both DraftKings fantasy points and Plus/Minus.

### Takeaways

The highest-rated correlation according to this exercise is the one between a quarterback and the opposing quarterback; second is between a quarterback and the opposing running back. The first one may not seem all that useful because you can’t roster two quarterbacks together on DraftKings or FanDuel, but what both examples show is that the high correlation between offenses is likely undervalued.

If you look at the ratings for the positional correlations above, they are around 50 or lower for all skill position players on the same team. Although the strongest correlation in terms of actual fantasy points is between a quarterback and wide receiver, that combination is also owned at a high rate. Quarterbacks and their second wide receivers, tight ends, and even defenses rate low in this exercise because of their popularity in tournaments. This isn’t to say you shouldn’t use a QB/WR1 stack; it is still valuable and raises the overall upside of your lineup, especially since there are so many teams and QB/WR1 combinations to roster. But on a general level — and this definitely applies to the chalky QB/WR1 stacks every week — this combination (due to its popularity) doesn’t provide a large edge in GPPs by itself.

But what is likely undervalued, at least according to the charts above, is game stacks — rostering multiple players from the same game on both sides of the offense. The correlation itself makes sense: If a quarterback is going to go off for a ton of yards and multiple touchdowns — a performance needed to win a GPP — the other quarterback will need to throw to stay competitive. Game stacks are intriguing in GPPs because they raise the overall correlation and upside of a lineup, but they aren’t owned nearly as much as other correlated groups.

In Week 1, you could stack the chalky Aaron Rodgers and Jordy Nelson and then roll it back with Doug Baldwin on the other side of the ball. It is likely that the ownership between Rodgers and Nelson will be significant — we project ownership for each main slate in our Models — but the ownership correlation between Rodgers, Jordy, and Baldwin will be low, despite the fact that they have very correlated outcomes. After lineups lock, Pro subscribers will be able to check out player combinations like this in our Contests Dashboard. Another example of a combination likely to have low ownership is Carson Palmer and Larry Fitzgerald rolled back with Detroit pass-catching back Theo Riddick.

We’ll continue to explore these GPP dynamics throughout the year, and I encourage you to use the other filters available in the NFL Correlations page. You can research how correlations are affected by different spreads and totals and even how correlations are different on different rosters. Correlations provide a huge edge in GPPs, as they provide free upside, and there are many secrets yet to be found within the FantasyLabs Tools.

Last week we released our new NFL Correlations page, which allows users to visualize positional and statistical correlations. Users can also view correlation data for specific seasons and teams, and they can even research how correlations change based on different Vegas data points like totals and spreads. Check out my tutorial video.

### Creating a Correlation Value Metric

I think the most valuable information gained from this new tool is not how correlated two positions are; it’s how owned two positions are together. Almost 100 percent of daily fantasy football users know that stacking a quarterback and his No. 1 wide receiver is advantageous and raises a lineup’s ceiling. They may not know it’s called “stacking” or that the exact r-squared value (the correlation) is 0.47 — but they know it’s a wise strategy. And the page confirms this: The tournament ownership correlation between those positions is 0.56, which is even higher than the r-squared of fantasy points. It’s useful and still probably wise to stack, but everyone knows about stacking.

That said, there are probably correlations that are less known by the public, or at least undervalued by the public. And since we have r-squared values for fantasy points, Plus/Minus, and tournament ownership, we can devise a formula to measure these discrepancies. (Also, Pro subscribers can review ownership trends across guaranteed prize pools of various buy-in levels via our DFS Ownership Dashboard.)

First, we need to put the correlation values on a non-negative scale; we’ll use 0-100 to make it easy. Here are the correlation values for actual fantasy points, per the NFL Correlations page . . .

. . . and here’s the same table scaled from 0-100:

Now we can do the same thing for both Plus/Minus correlations, and more importantly, tournament ownership correlations. For the purpose of the exercise, I’m giving low tournament correlation a high rating and vice versa. Basically, what we’re looking for is positional correlations that have a high correlation in terms of fantasy points and Plus/Minus but low correlation in terms of tournament ownership; the higher the rating the more undervalued the correlation would be.

Using that rating, here are the most undervalued positional correlations for both DraftKings fantasy points and Plus/Minus.

### Takeaways

The highest-rated correlation according to this exercise is the one between a quarterback and the opposing quarterback; second is between a quarterback and the opposing running back. The first one may not seem all that useful because you can’t roster two quarterbacks together on DraftKings or FanDuel, but what both examples show is that the high correlation between offenses is likely undervalued.

If you look at the ratings for the positional correlations above, they are around 50 or lower for all skill position players on the same team. Although the strongest correlation in terms of actual fantasy points is between a quarterback and wide receiver, that combination is also owned at a high rate. Quarterbacks and their second wide receivers, tight ends, and even defenses rate low in this exercise because of their popularity in tournaments. This isn’t to say you shouldn’t use a QB/WR1 stack; it is still valuable and raises the overall upside of your lineup, especially since there are so many teams and QB/WR1 combinations to roster. But on a general level — and this definitely applies to the chalky QB/WR1 stacks every week — this combination (due to its popularity) doesn’t provide a large edge in GPPs by itself.

But what is likely undervalued, at least according to the charts above, is game stacks — rostering multiple players from the same game on both sides of the offense. The correlation itself makes sense: If a quarterback is going to go off for a ton of yards and multiple touchdowns — a performance needed to win a GPP — the other quarterback will need to throw to stay competitive. Game stacks are intriguing in GPPs because they raise the overall correlation and upside of a lineup, but they aren’t owned nearly as much as other correlated groups.

In Week 1, you could stack the chalky Aaron Rodgers and Jordy Nelson and then roll it back with Doug Baldwin on the other side of the ball. It is likely that the ownership between Rodgers and Nelson will be significant — we project ownership for each main slate in our Models — but the ownership correlation between Rodgers, Jordy, and Baldwin will be low, despite the fact that they have very correlated outcomes. After lineups lock, Pro subscribers will be able to check out player combinations like this in our Contests Dashboard. Another example of a combination likely to have low ownership is Carson Palmer and Larry Fitzgerald rolled back with Detroit pass-catching back Theo Riddick.

We’ll continue to explore these GPP dynamics throughout the year, and I encourage you to use the other filters available in the NFL Correlations page. You can research how correlations are affected by different spreads and totals and even how correlations are different on different rosters. Correlations provide a huge edge in GPPs, as they provide free upside, and there are many secrets yet to be found within the FantasyLabs Tools.