“It’s still wise to understand a team’s identity” – John Hansen, “2005’s Lessons Learned”
It's been a long time since John wrote that and started me taking a statistical look at the offensive identities of each year’s teams, to see what they would tell me about the coming year. This has gone from just an overview article to a complete series on each team. In a few weeks, I'll look at the 2017 coaches individually and what offenses they ran in their current and previous jobs to try to get some insight into what they’ll do this year – and what players will benefit or suffer from that identity.
But to start, I have to compile the data on what teams did in 2016. In this article, I’m going to share that data and a bit of analysis or observations on what happened last year.
I'm going to change how I displayed the data this year and go with more charts (pictures) vs. a giant table (numbers). And I'm going to spend time on how each aspect of identity translates (or doesn't) into fantasy value.
The most basic part of an offense's identity is how many plays it runs from scrimmage:
I tried to make the chart do most of the talking: the average team had 1023 plays from scrimmage in 2016; with a standard deviation (SD) of plus or minus 41 plays. Teams in green had at least one SD more plays than average; those in red were at least one SD below average.
Number of plays is somewhat connected with pace of play – the faster a team lines up and snaps the ball, the more plays it gets off. But it's not just that. Incompletions stop the clock, so teams that pass more tend to have more plays too. Good offenses stay on the field more, also translating to more plays. Good defenses get the ball back, more plays on offense. Bad defenses mean a team trails, therefore passes more, therefore, more plays.
You can see that some "green" teams were good offenses, like NO and ARI. TB was a good fantasy offense, maybe not as good for real football. BAL and PHI were below average offenses. HOU was terrible despite all the plays.
There is a lot more to fantasy scoring than just plays. The next chart shows how many total fantasy points (FP) each team accumulated rushing and receiving last year (10 yds = 1 FP, TD = 6 FP, receptions =1 FP).
Again, the chart shows the number of plays on the x-axis, with each labelled increment representing one standard deviation from average (1023). The y-axis depicts the total rush and receiving FP: the average team scored 1152. Again, each increment is one SD from average.
The dotted blue line represents the expected number of FP scored by a team, given a certain number of plays. The equation of the regression line is: Total FP = 1.2254 * Number of plays – 102.54.
Teams above the dotted line scored more FP than would be expected by the number of plays they had: these were relatively efficient offenses. Those below the line were inefficient, scoring fewer points than they would be expected to.
NO led the league in total rushing and receiving FP as well as plays. And LA was last in FP and second last in plays. But the correlations between plays and FP are not that strong.
Atlanta was #2 in FP despite fewer plays than average. Washington was #3 in FP, also with a below average amount of plays. MIA had by far the fewest plays, but only a little below average in generating FP. Houston, which was one of the leaders in plays from scrimmage, was near the bottom in in FP.
The R-squared number (0.1618) is the tip-off. Only about 16% of total rushing and receiving FP can be explained by how many plays a team runs. The average number of FP divided by the average number of plays is just 1.12 FP/play. So whether you use that number or the coefficient from the regression equation (1.2254) it helps fantasy production to have more plays, but not a lot. Remember that extra point or so per play is divided over a QB (his running), a couple of RBs, a couple more TEs, and several WRs. Not to mention the occasional Hungry Pig Right.
A quick scan of the teams above and below the dotted line seems to support this hypothesis: QB quality is a lot more important than number of plays in generating team FP. Even better if Drew Brees gets a lot of plays, but Brock Osweiler is a point-killer no matter how many plays he gets. And while it may be that Jared Goff was better off without many plays, as mentioned above, the quality of the QB probably is connected to how many plays an offense runs.
And by the way, wins - by themselves - explain little about a team's number of plays:
A quick look at the equation for the dotted regression line tells the story: # of Plays =1.4324 * Wins + 1011.4 with an R-squared of 0.0124. In English: every game a team wins tends to add about 1.4 plays to their bottom lin. And wins explain about 1% of the total number of plays a team runs. Winning teams DO tend to generate fantasy production, just not from play volume:
Again looking at the regression equation, FP = 19.368 * Wins + 996.32 or a win means about 20 more FP rushing and receiving. Not a huge amount, but not trivial. The R-squared (0.2443) indicates that wins explain about 24% of total rushing and receiving FP. I'd generalize about the teams with more than 10 wins: those above the regression line had good offenses (NE, OAK, PIT, ATL, GB), those below it had good defenses (KC, NYG, SEA) with a couple of exceptions (MIA and DAL).
Next I want to looking at rushing vs. passing plays as an aspect of a team's identity. I'll start with rushing plays:
The chart is laid out like the total plays chart.
Not surprisingly, Dallas led the league in rushing plays, with Buffalo close behind. The Cowboys were the only team 2 SD over average. The #3 team was a bit of a shock: New England. On the low-end, there's teams like BAL and GB that had trouble finding a healthy RB as well as a couple of bad team (CLE and LA). But total rushes can be a little tricky, because it includes QB carries, WR runs, and assorted hand-offs to other players. Here's just the RB carries, which is more important for RB fantasy production than total rushes:
Some of the teams that made the +/-1 SD list in total rushes fall out when looking at just RB carries: Baltimore, Buffalo, and San Francisco. The Bills and 49ers rushing numbers were inflated by a lot of QB carries. Buffalo still gave its backs an above average number of rushes but SF did not.
The big surprise in this chart is Seattle. Again, this was a team that had trouble keeping an RB on the field. But it definitely strayed from its past identity as a running team in 2016.
Houston is a bit of a surprise too. Of course the Texans ran a lot – they couldn't pass. But Lamar Miller's 268 carries don't seem like a dominant number. In fact, he was 6th in the NFL in rushing attempts. Since 2011, the RB with the 6th-most carries has averaged 271, so Miller is right there. But in the six seasons before 2011, the #6 back in carries averaged 315. The game has changed.
And if you want to understand part of the reason Todd Gurley disappointed, its in LA's low number of attempts. Sure, Gurley had 278, good for 5th. But when a team takes an RB that high, they want to give him 300+ rushes, like Ezekiel Elliott. The Rams just didn't have enough rushes to go around (yes, and some of that was Gurley's fault for not being more productive).
Let's compare the significance of total rushes and RB rushes for Team RB fantasy scoring (I think you know the answer):
I won't spend a lot of time on this chart. Note the regression equation and R-squared values, I'll come back to them after showing the chart with just RB rushes:
Here's the regression equations for the two charts:
Team RB FP = 0.5754 * Total Rushing Plays + 139.32
Team RB FP = 0.8904 * RB Rushing Plays + 59.362
Every rush a team gives its RBs adds about 0.9 FP to their total compared to just under 0.6 for each total rush by the team. While there is something to be said for running QBs helping RB productivity, it's still not as good as your RB ACTUALLY CARRYING THE BALL (I shouldn't have to write that). As long as the team is using its backs a lot, the QB taking some carries may help. But when the QB starts running instead of the RB, that is not good. Too obvious?
Not surprisingly, the R-squared of the first regression is just 0.1095: total team rushes explain about 11% of team RB FP. The number for the RB rushes regression is twice as big: 0.2229. That's still not huge: RB carries only explain 22% of total RB FP. Of course, that number would be higher in non-PPR formats. But it's important to remember that total RB carries are not the entire story. How a team divides them up is far more important than just volume. The R-squared between an individual RB's carries per game and his FP/G in 0.796: about 80% of individual RB FP/G can be explained by rushes per game. An RB on a team that runs a lot is likely to get more carries, but his share of carries on that team is far more important.
The next chart looks at the identity of 2016 teams in terms of the Rush percentage of its RB1 (Rush % RB1). RB1 is defined as the RB on the team who got the largest number of carries. So I'm talking about NFL teams Rb1s, not fantasy RB1s.
The Rush % RB1 is the percentage of the team’s total RB rushes claimed by its RB1. Note this is not necessarily be a good predictor of points game to game because it doesn't account for injuries, and we have to recognize that the identity of a team like GB or DET in 2016 is affected by multiple RB injures. So some additional knowledge has to be applied to these numbers:
The reason for the value of David Johnson and Ezekiel Elliott jumps out at you in this chart. Now, would their teams have given lesser backs as many touches? That's the point of subsequent articles when I delve into the identity of their coaches over multiple years.
We can also see why Frank Gore remained valuable. And we have questions raised about Todd Gurley: how does a back in 2016 get almost 90% of his team's RB carries rank only 25th in FP/G? (I'm not going to try to answer them fully here, I've already touched on them a bit).
How does share of RB carries translate to FP?
Most of the top-scoring RB1s also had a high share of their team's carries. And the highest scoring backs were "more efficient." That's a little misleading if you think of running back efficiency as yards per carry. In this context it's short-hand for more FP than would be expected based on the share of carries the back received. It's perhaps a better measure of involvement in the passing game: all seven "green" RB1s also had over 50% of their team's RB targets as well as an above average share of its rushes.
If you were to draw a line across the chart at the SF diamond for Carlos Hyde, you'd see that 16 of the 18 RB1s above that level of FP also had above average RB1 share. Only two backs below that line had an above average share without ranking in the Top 18 RB1s in total FP. (Bilal Powell was the highest ranking fantasy back as his team's RB2 – he was #18 in total points; the top-ranked RB2 in FP/G was Tevin Coleman at #14). And only two RB1s were above that line with a below average share of rushes.
From the regression equation for this chart, an increase in 1% share of RB carries is worth about 3 FP (RB1 Total FP = 321.08 * Rush % RB1 + 17.91; remember that Rush % is a fraction). Small changes in rush percentage are unimportant, but adding 10% share adds 32 FP – now we're talking real points. And with an R-squared of 0.4649, almost 50% of RB1 fantasy points can be explained by their share of team carries. Considering that ignores the passing game and TDs, that's a pretty good factor to hone in on, better than all the others discussed so far.
Part II will address the passing game part of identity.