A few weeks ago, I wrote an article about correlation anomalies – the idea of players correlating positively with one of their teammates even when Rotoviz’s correlation matrix says they shouldn’t. The primary example I cited at the time was with New Orleans, where I found a positive correlation between Brandin Cooks and Mark Ingram even though the correlation chart shows a -0.07 correlation between WR1 and RB1.
Matt Harmon had an interesting tweet the other day that got me thinking about this a little more. Here’s the tweet:
There are several teams this year where two players are commanding at least 50% of the total passing targets. I think it will be really interesting to see how the above listed players will correlate with each other. WR1 and TE1 has a negative correlation of -0.08 and yet, for the teams above, it kind of seems like if the passing game is hitting on all cylinders, both players are going to do well.
The below chart requires a little explaining. Basically, it shows how “Player” performed when “Teammate” exceeded their projection in a game. A positive Plus/Minus means both players did well in the same games. A negative Plus/Minus means one did well while the other had a poor performance.
| Player | Teammate | Results | Average +/- |
| Fitzgerald | Brown | 5 | -1.34 |
| Brown | Fitzgerald | 6 | +5.28 |
| Ginn | Olsen | 4 | +2.03 |
| Olsen | Ginn | 4 | +3.66 |
| Eifert | Green | 2 | -6.54 |
| Green | Eifert | 4 | -6.55 |
| Bennett | Jeffery | 2 | +1.4 |
| Sanders | D Thomas | 4 | -0.39 |
| D Thomas | Sanders | 4 | -2.63 |
| Hurns | A Robinson | 5 | +7.2 |
| A Robinson | Hurns | 5 | +5.16 |
| Marshall | Decker | 7 | +4.05 |
| Decker | Marshall | 5 | +4.94 |
| Cooper | Crabtree | 5 | +4.84 |
| Crabtree | Cooper | 4 | +12.41 |
| Evans | V Jackson | 2 | -13.31 |
| Reed | Garcon | 2 | +3.18 |
Even on teams where the majority of the total targets are funneled through two players, there is not 100% correlation. Some specifics:
• Fitz/Brown looks worse than it is. Against the Ravens, John Brown exceeded his projection by around four points, while Fitzgerald underperformed by almost 11 points. Outside of this result, Fitz generally checked in right at his projection when Brown played well.
• John Brown did perform well when Fitzgerald performed well, however. It seems as Fitzgerald goes, so goes the Cardinals passing game as a whole.
• If you exclude the Panthers’ game against the Saints where Olsen scored 36.4 points and Ginn scored 13.3, the players would have actually had a negative correlation.
• AJ Green and Tyler Eifert have one of the largest negative correlations on the board. There has not been one game this season where both players have exceeded value.
• Demaryius Thomas slightly exceeded value in two of Sanders’ best games but completely tanked in the other two, leading to an overall negative value.
• As far as WR1-WR2 correlations go, I don’t think you’ll find a stronger one than Allen Hurns & Allen Robinson. Each player had one game where they just about hit value on the nose when their teammate had a good game, but other than that, they have been destroying value together.
• Brandon Marshall and Eric Decker have exceeded value together in over half of the games the Jets have played this year. The numbers aren’t quite Hurns-Robinson good, but this is another highly positive correlation.
• Crabtree’s two biggest games of the year occurred in weeks where Amari Cooper produced one of his four best scores.
Although I think there are probably some useful nuggets in the above chart, the purpose of this article isn’t really to highlight the individual. Looking at tight ends and wide receivers, Olsen/Ginn and Reed/Garcon did well together, while Eifert/Green did not.
If you take every WR1-TE1 correlation, I think you probably get a negative number once again this year. But a lot of tight ends around the league just aren’t great playmakers in the passing game. For the Jets, Jeff Cumberland and Kellen Davis have both been virtually non-existent in the passing game this year. What if the Jets traded for Greg Olsen tomorrow? I have a feeling that would change things a bit.
If you look at every WR1-WR2 correlation around the league this year, I’m convinced you’ll get a positive number there. But there are some teams where the relationship has seen one receiver do well while the other struggles.
I think the main point here is that you should try to take a narrowing down approach when constructing your DFS teams. The correlation chart is one of the first pieces of information to consider while you are thinking about things on a broad level. Then as you narrow in on the matchup, maybe you think about the opponent’s “Opponent +/-” rating, then the Vegas score, then check the weather, then whether there are any injuries that change things, etc.
As you go from a broad perspective to a more narrow one, you want to check off as many boxes as possible. If the correlation chart says there should be a negative correlation between two players you are considering, maybe you don’t check off that box. But there are still many other boxes you could check off for the stack and to rule the stack out on such a broad, general level is usually going to be a mistake.
