"Correlation" is one of the key words in fantasy football among the galaxy-brainers. It's used often, which also means it's used loosely. Today's column is measuring just how much correlation players have with each other because correlation is a scale, not a yes or no thing.

Best Ball Correlation

When it comes to best ball, just how much correlation teammates have matters because we need to know how much we need to prioritize teammates in the draft room. Over the course of a season, teammates will be correlated with each other just because an offense as a whole can over- or under-perform our pre-draft expectations for a variety of reasons, so as a tie-breaker, it's fine to prioritize a teammate over another random player at average draft position (ADP).

BUT as we'll see in the column, this teammate correlation changes in a one-week sample, which definitely matters in these Underdog Fantasy best ball tournaments where the winner will advance in the regular season, advance in a mini-Week 15 tournament, advance in a mini-Week 16 tournament, and have to place at the top of the Week 17 tournament to take down the top prizes. Finding the balance of adding season-long correlation and one-week correlation requires finesse. In general, I'm a believer in adding as many mini team stacks as possible, especially with QBs, WRs, and TEs (not RBs). Some offenses are only good enough to support a QB with one of his pass-catchers. Others are good enough to support a QB with two of his pass-catchers. I have a YouTube Channel to discuss which offenses fit into each category.

Because most people understand the teammate correlation already, the new hot debate is how much to value Week 17 "bring back" correlation. For example, if I've drafted Lamar Jackson and Rashod Bateman (BAL), how much should I prioritize Chase Claypool and Pat Freiermuth (PIT) later on because the Ravens play the Steelers in the fantasy finals? In a vacuum, we're adding some correlation by selecting opposing players, but if some people are now going out of their way to stack Week 17 opponents, then the lack of uniqueness could easily counteract the bring back correlation if the correlation is minimal. It's not the industry consensus, but I think we're already at (and quite frankly beyond) that point for some of the bring back pairings. The only bring back options we should really care about are QBs and WRs, yet drafters are going out of their way to draft opposing RBs and TEs. We can't pretend that correlation is even between positions. We have to get into the details here to be optimal.

This is a pretty long column, so I'm adding a TL;DR table right at the top as a correlation cheat sheet. The key points are that this study is largely looking at ceiling outcomes, not correlation as a whole, because best ball is a format that chases week-by-week ceilings. The second key point in this data and in the table below is that a "WR1" label is that week's highest-scoring WR on a team, not the team’s WR1 in ADP nor the team’s WR1 in season-long points nor the team’s WR1 in weekly projections. Lastly, this data is from 2015-2021 and is only looking at Weeks 8-17 because I wanted it to reflect some of the late-season weather and recent-season offensive trends.

RB1 Correlation Details

Correlation: RB1 to WR1 (-0.02), to WR2 (-0.11), and to TE1 (-0.05) if we split the sample to RB1s with more than 15.0 half PPR points (aka ceiling games only). 

RBs are very slightly correlated to their teammate WRs on the whole, but this is not true at ceiling outcomes. When a RB scores more than 25 half PPR points, his highest-scoring teammate WR averages 14.6 half PPR points and has a 41% chance of eclipsing 16.0 half PPR points (which is essentially a weekly WR1 number). Those numbers are slightly higher when a RB only scores 15-24 half PPR points. In other words, there is either no correlation or potentially very slight negative correlation between a RB1 and his WR teammates during boom weeks. This especially true for a RB1 and the second-highest scoring WR teammate, likely because there are only so many yards and touchdowns to go around in a given week.

There appears to be a slight negative correlation between a RB1 and his teammate TE1. TEs average the most points and have the highest odds of scoring more than 16.0 half PPR points (an elite TE1 week) when his teammates’ RB1 only scores 10-14 half PPR points. When a RB scores more than 25 fantasy points, it’s very hard for a TE1 to reach a ceiling because TEs are the most tied to touchdowns.

In summary, the correlation between an RB and his teammates is essentially non-existent. Some pass-catching RBs could have slightly positive correlations with their WRs and their QB because of game script, but the bulldozing RBs ultimately need touchdowns, so the position is showing some slight negative correlation at the ceiling outcomes. For that reason, it’s rare for me to draft a RB on my pass-game stacks, especially when they are going early in drafts. As an example, my 2022 Joe Mixon shares will not be on teams with Ja’Marr Chase, Tee Higgins, or Joe Burrow, but I don’t mind Mixon with Tyler Boyd because the latter is much cheaper and both are correlated if Chase or Higgins miss time.

TE1 Correlation Details

Correlation: TE1 to WR1 (+0.02), to WR2 (-0.04), and to RB1 (-0.04) if we split the sample to TE1s with more than 10.0 half PPR points (aka ceiling games only).

On a whole, TEs aren’t very correlated with their WRs. It’s illogical for the WR1 data to be worst when a TE1 scores between 10-19 half PPR points but best when a TE1 scores an outlier 20+ half PPR points. Some of this is noise based on the Travis Kelce and Tyreek Hill sample, so my instinct is to treat the TE1 with WR1 data with a 10-foot pole. 

What definitely makes sense to me (and is confirmed by the data) is the TE1s being negatively correlated with his second-highest scoring WR teammate in a given week. When a TE1 scores 15+ half PPR points, his WR2 teammate only scores 7.3 half PPR points and only has a 4.7% chance of scoring more than 16 fantasy points, both are below average. There are only so many passing yards and passing touchdowns to go around, so it’s best to stack a TE with only one of his WRs.

Just like WR2s, it’s very clear that TEs are negatively correlated with RB1s at ceiling outcomes. A RB1 has a 41% chance of scoring 16+ half PPR points when his teammate TE scores fewer than 9.9 half PPR points, and those odds drop to 36% when his teammate TE1 scores more than 15 half PPR points.

In general, TEs are best stacked with their QB and one of their WRs. It doesn’t really matter which WR it is (it can be the WR1 or the WR2 in ADP), but the odds that you’ll want a QB1, WR1, WR2, and TE1 all on the same team in the fantasy playoffs are very low.

WR1 Correlation Details

Correlation: WR1 to WR2 (+0.16), to RB1 (+0.01), and to TE1 (+0.02) if we split the sample to WR1s with more than 15.0 half PPR points (aka ceiling games only).

What’s clear is that a WR1 on a week is positively correlated with his second-highest scoring WR teammate, even at relative ceiling outcomes. When a WR1 scores 25 or more half PPR points, his WR2 teammate averages 11.7 half PPR points. When a WR1 scores fewer than 20 half PPR points, the WR2 averages 7.5 half PPR points. This probably is because of game script or just offensive philosophy. 

The debate is at the most ceiling of outcomes, like Ja’Marr Chase’s Week 17 performance last year. When a WR1 pops off for 30+ half PPR points, do we want his WR2 teammate, like Tee Higgins in this example? While the average WR2 score is highest when the WR1 goes nuclear, the highest odds of scoring 16+ half PPR points is when the WR1 scores in the 25-29 half PPR range. In other words, Ja’Marr Chase teams are best off without Tee Higgins at Chase’s ultimate ceiling, but on a whole, Chase teams are fine with Higgins, too. As a guide, only 1.1% of teams had a pair of WRs each score more than 20.0 half PPR points in a game and only 0.2% of teams had a pair of WRs each score more than 25.0 half PPR points, so be price sensitive.

Onto RBs. The highest-scoring WR and his RB1 teammate basically have no correlation. I’m sure if I split the data into pass-catching RBs and bulldozing RBs, then we could find some very slight positive correlation between WR1s and pass-catching RBs. But for the most part, we’re not moving the needle much in either direction. 

Like mentioned in the TE1 correlation section, there’s not much WR1 and TE1 correlation. Perhaps there’s a tiny bit of negative correlation when a WR1 goes nuclear, but for the most part it’s fine to stack a WR1 with a TE1. You’re not going to make or break your roster either way.

RB1 Bring Back Correlation Details

Correlation: RB1 to opposing WR1 (+0.07), to opposing WR2 (+0.03), to opposing TE1 (+0.08), and to opposing RB1 (+0.01) if we split the sample to RB1s with more than 15.0 half PPR points (aka ceiling games only). 

It’s not the craziest thing ever by any means, but there is a very slight positive correlation between an RB1 and his opposing team’s WR1, WR2, and TE1. This makes a ton of sense! Our elite-scoring RBs are getting there because they are scoring touchdowns, which means the opposing team has higher odds of trailing, leading to more pass attempts. This is especially true for the bulldozing RBs whomst rely on touchdowns more than the pass-catching RBs. 

It’s also clear that at ceiling outcomes, an RB1 is slightly negatively correlated with the opposing team’s RB1. The odds an opposing RB1 scores more than 16.0 half PPR points are lowest when the RB1 scores more than 20 half PPR points. In fact, when a RB1 scores more than 30 half PPR points, the opposing RB1 only scores 13.7 points, which is notably lower than the 14.5 point average. Now, it’s possible that an elite bulldozing RB isn’t negatively correlated with an opposing pass-catching RB, but that’s pushing it. In general, it’s best to avoid RBs playing each other, especially if the field is jamming Week 17 correlation blindly (read: Christian McCaffrey with Leonard Fournette).

TE1 Bring Back Correlation Details

Correlation: TE1 to opposing WR1 (+0.00), to opposing WR2 (+0.04), to opposing TE1 (+0.04), and to opposing RB1 (+0.02) if we split the sample to TE1s with more than 10.0 half PPR points (aka ceiling games only). 

There is a very slight positive correlation between a TE1 and the opposing team’s WR1 and WR2. When a TE1 scores 10+ points, the opposing WR1 has a 47% chance of scoring more than 16 half PPR points compared to a 40% chance when the TE1 scores fewer than 10 points. This trend is directionally true for the opposing WR2, too, probably because the TEs are getting to high scores with touchdowns, leading the other team to throw more in catch-up mode.

What’s interesting about TE1 bring back correlation is unlike at WR, there appears to be no correlation between a TE1 and the opposing team’s TE1. This might be because TEs don’t offer as many big plays, which help speed up drives and lead to more touchdowns in general.

There appears to be a very slight correlation between a TE1 and the opposing RB1, but the data is all over the place at a TE1’s very ceiling outcomes. I can’t explain why that is to be honest. Ultimately, I view the RB1 bring back to a TE1 as nothing more than a tie-breaker in best ball. And even then, there are ownership concerns.

WR1 Bring Back Correlation Details

Correlation: WR1 to opposing WR1 (+0.09), to opposing WR2 (+0.10), to opposing TE1 (+0.01), and to opposing RB1 (+0.03) if we split the sample to WR1s with more than 15.0 half PPR points (aka ceiling games only).

Now it’s time for the good stuff. Opposing WRs are definitely correlated at extremely ceiling outcomes. When a WR1 scores 25+ half PPR points, the opposing team’s highest-scoring WR averages more than 17.0 half PPR points with more than 50% odds of scoring 16.0 half PPR points. This is because a long receiving TD can spark the opposing team to throw more or because two pass-heavy teams will run more plays as the clock doesn’t run as much.

For the most part, even when a WR1 has a good game, only one opposing WR will be fantasy relevant for the same reasons we went over in the WR1 Correlation section. In other words, the more mini WR stacks, the better. I’d much rather have two sets of a QB-WR1-WR2 with a WR1 bring back than one set of a QB1-WR1-WR2-WR3 with a WR1-WR2 bring back. The latter is too much.

Elite-ceiling WR1s have a smaller, yet still positive correlation with their opposing TE1 compared to the opposing WR1, likely because TEs are less prone to big plays. Drafting an opposing TE1 is better than nothing in theory, but you won’t win or lose because of it and the ownership overlap could make it actually a negative.

Finally, WR1s and opposing RB1s have a very slight positive (but confusing) correlation. The truly elite WR1 week (30+ half PPR points) has been relatively bad news for the opposing RB1, but opposing RB1s tend to score more points than normal when the WR1 is scoring 20.0 to 29.9 half PPR points. Maybe there's something to this, but I think it's best to keep it simple. There really isn't any correlation between a WR1 and an opposing RB1. If people are going out of their way to stack these two positions in best ball, it's best to fade it for uniqueness purposes.