Two players from the same NFL team
Posted by Doug Drinen, Exclusive to Footballguys.com
It's an extremely common question, both before and during the season: if my #1 receiver is, say, Roy Williams, and if Calvin Johnson is atop my cheatsheet when it comes time to draft my second receiver, should I skip him in favor of a non-Lion? More generally, what about pairing two NFL teammates at different positions? If you have a draft slot near the middle of the round this year, you could be in position to grab Joseph Addai in the first round and fellow Colt Reggie Wayne in the second. Or, you could be eyeing Addai in the first round and Peyton Manning in the second. Or maybe Marion Barber in the first and either Terrell Owens or Tony Romo in the second or third. T.J. Houshmandzadeh and Chad Johnson would make a good WR duo in rounds three and four.
Or would they?
Many people might tell you that they prefer Houshmandzadeh to, say, Marques Colston and Chad Johnson to Plaxico Burress, but that they'd rather have Colston/Burress as their starting WR pair than Houshmandzadeh/Johnson. Two wideouts from the same team is putting too many eggs into a single basket. Too inconsistent. Too variable. Just as you must diversify your investment portfolio, you must diversify your fantasy football team. Or so the argument goes.
I've written about this a few times before. The first was in 2003. In that article, I looked at all pairs of same-team WRs who both finished in the top 20 in fantasy points. I then found all receiving duos who scored a comparable number of total fantasy points as the same-team duos and discovered that, more often than not, the same-team duo was more consistent from week to week than the comparable different-team pairs. In a followup article, I found similar results for QB/RB and RB/WR pairs.
Three years later, I wrote a short article on the topic for ESPN magazine. Here is a clip of that:
Mathematicians use a thing called "covariance" to describe whether two quantities vary together or vary inversely. Yeah, it's a fancy word, and it's a bit messy to compute, but that's what we're here for. All you need to know is that a positive covariance means that, generally, the two things move together and a negative covariance means they tend to move in opposite directions. Carson Palmer and Chad Johnson had a positive covariance last year: when Carson was up, Chad generally was too, and vice versa. But interestingly, Rudi Johnson and Chad Johnson had a negative covariance. Generally, Rudi's strong games coincided with Chad's weak games, and vice versa.
In the last decade, there have been 20 instances of a top-10 running back and a top-10 wide receiver playing for the same team in every game of weeks 1 through 16 (I'm throwing out week 17 because most fantasy leagues don't use it and because starters often rest). In 16 of those 20 cases, the two players' weekly fantasy scores had a negative covariance. As mentioned above, Rudi Johnson and Chad Johnson of last year's Bengals were one of the 16. The overall effect is that a Rudi-Chad pairing would have made your team more --- not less --- consistent than a different-team pair of similar quality.
As you can see from the table below, the majority of QB/RB and WR/WR pairs behaved similarly.Pairing Result -------------------------------------------------------------------------- Top 10 RB with Top 10 WR 16 out of 20 had negative covariance Top 10 RB with Top 10 QB 16 out of 29 had negative covariance Top 20 WR with another Top 20 WR 12 out of 21 had negative covariance
Bottom line: in most cases, starting two guys (other than a QB/WR pair) from the same NFL team is a conservative play, not a risky one. It's likely to make your team more consistent.
In these two studies, I have seen enough evidence to believe strongly that a same-team RB/WR, RB/QB, or WR/WR combination consisting of good players will usually (not always, but usually) make your team more consistent from week to week than a similar-scoring different-team pair. (Note that QB/WR pairs are very different. They will, in the vast majority of cases produce a postive covariance, and hence make your team less consistent.)
But there's a problem here. Both studies above have considered only pairs of players that were known after the fact to have had good seasons. If we knew that Roy Williams and Calvin Johnson were both going to be top 20 receivers this year, then I'd be willing to bet that they, as a duo, would make your team more consistent than a similar scoring different-team pair. But we don't know in August that they're both going to be top 20 receivers. It is, in fact, not unreasonable to suspect that they're chances of being top 20 receivers could be tied together. If Kitna gets hurt and Stanton isn't ready, it hurts both of them. If you own Williams and Johnson, and the Lions' offensive line suffers a rash of injuries, it hurts two members of your team instead of just one.
In other words, maybe drafting two players from the same team makes you more likely to have two underperformers.
Or maybe it doesn't. Maybe the two players tend to provide a sort of insurance for each other. Maybe if Williams gets hurt, it will open the door for Johnson to have that ridiculously great season many feel he's capable of. Maybe if your Terrell Owens investment goes south because of injury or general Owensishness, it makes the Cowboys a run-heavy team, which increases the value of your Marion Barber investment. Marvin Harrison and Reggie Wayne 2007 provide an example of this. Obviously, Marvin Harrison didn't turn out to be a good pick. But Harrison / Wayne turned out to be better than Harrison / Holt or Harrison / Fitzgerald. Why? Lots of reasons, obviously. But one of which might be that Wayne's numbers were improved by the absence of Harrison. Wayne, as it turned out, served as Harrison injury insurance. And vice versa. In this case at least, it seems that drafting the same-team pair made you less likely to have two underperformers.
As is often true, a reasonable common sense case can be made, with examples cited, for each of two opposing viewpoints. So as I often do, I'm going to look at the historical data to see which way, if either, the actual data leans.
Let's start by looking at WR/RB pairs.
The first thing I did was to find all instances from 2000 to 2007 where a single NFL team had an RB and a WR who were among the top ten in terms of positional ADP (fine print). There were 27 such pairs, from Jimmy Smith / Fred Taylor in 2000 to Torry Holt / Steven Jackson last year.
For a concrete example, let's focus on the 2004 Seattle Seahawks. They had Shaun Alexander, whose ADP was RB4, and Darrell Jackson, whose ADP was WR10.
The next step was to identify all players who were within two ADP slots, one way or the other, of each of our two players. In the example above, Alexander was RB4, so I looked at RB2 (Tomlinson), RB3 (Ahman Green), RB4 (Alexander), RB5 (Portis), and RB6 (McAllister). Darrell Jackson was WR10, so I looked at WR8 (Mason), WR9 (Joe Horn), WR10 (Jackson), WR11 (Steve Smith), and WR12 (Coles). Then I looked at all combinations --- one RB and one WR --- of players from those lists and recorded the fantasy point total of each duo. That's 24 pairs, plus the Alexander/Jackson pair we're interested in. The median of the other 24 pairs was 360 fantasy points. That's what you might reasonably expect if you drafted a similarly-ranked pair of players to Alexander and Jackson. Alexander and Jackson themselves totalled 468 fantasy points, so you would have done well to take the Seahawk duo.
Of the 27 pairs, the same-team duo did better on 14 occasions, and worse on 13 occasion. The same-team pairs averaged a total of 383 fantasy points. Their different-team comps averaged 380. Essentially no difference.
But that's not really what we're interested in. What we're interested in is the spread of outcomes. We want to know if the same-team pairs ended up at that average via lots of very high scores and lots of very low scores, or if their range of scores was more or less the same as the different-team pairs' range was.
The answer turns out to be the latter. The most common mathematical tool for measuring the spread of a collection of numbers is the standard deviation. The same-team pairs had a standard deviation of 104 points. The comparable different-team pairs had a standard deviation of 107 points. Again, no difference. In general, there is no reason to conclude that a same-team WR/RB pair is more variable --- more boom-or-bust --- than a similar-scoring different-team pair.
And it turns out the same is true of QB/RB pairs and WR/WR pairs.
There were 38 instances of a top-10 ADP QB and a top-10 ADP RB playing on the same NFL team. The same-team pair outscored the median of their different-team comps 22 of the 38 times. The same-team pairs averaged 502 points vs. 500 for their comps. The standard deviations were 127 (same-team) and 124 (different-team). Again, no significant difference.
There were 20 instances of two top-20 ADP WRs on the same NFL team. The same-team pairs outscored the median of their different-team comps exactly half the time. The averages were 309 (same-team) to 302 (different-team), and the standard deviations were 73 (same) and 79 (different).
Bottom line: the historical data strongly suggests that you not skip over a player on your cheatsheet just because you already have another player from the same NFL team. As always, though, the particular situation you're considering is not a composite of historical situations. It's got its own pecularities, and you might have good reasons to believe that it's different. As usual, I view this study merely as a burden-of-proof shifter. That is, it should only define your default mode of thinking, your starting point. Instead of assuming that a same-team pair has more variability than a different-team pair and asking, "should I do it anyway?", you should instead start by asking, "In general, the same-team pairing is NOT more variable, so is there any particular reason why, in this case, it would be?"
Fine print: from 2005--2007, I used the last FBG composite ADP before the start of the regular season. For years prior, I used myfantasyleague's ADP data for redraft leagues drafting on August 25th or later.