Last year, Tre’Quan Smith scored 88.7 points in PPR leagues, finishing as the WR72. His preseason ADP was 219.2 (WR103). While he did finish 31 spots above his positional ADP, you wouldn’t exactly call him a league-winner.
Tyler Boyd, on the other hand, was a league-winner. His ADP was even higher than Smith’s at 226.7 (WR143). In fact, his ADP was so high that he was only drafted in 214 leagues (Smith was drafted in 1,556). He scored 221.1 PPR points and finished as the WR17. His win rate in BestBall10 leagues was 14.0%.
Smith’s win rate was 13.7%.
How is it possible that these two players had almost identical win rates when they had similar ADPs and Boyd outscored Smith by 132.4 points?
Here’s how: Smith had two unbelievable weeks. In Week 5, he scored 26.1 points. Six weeks later, he dropped 31.7. Those two weeks made up over 65% of his season-long total, and in many other weeks he scored exactly zero points. Among wideouts with at least 30 targets, he had the third-highest coefficient of variation (CV), a stat used to measure consistency. Boyd was much more consistent, as his CV was the 87th-highest out of 116 players.
Clearly, consistency (or in this case, inconsistency) matters. But, can we predict it?
For this study, I only included players from the last five years who finished the season with at least 60 targets. I did this for two reasons:
- It prevents players who only played a few games from skewing the results.
- I tried to only include players who were getting consistent volume on a weekly basis and exclude players who scored zero in an active week.
Furthermore, I only included players who stayed on the same team from one year to the next for all tests that examined the correlation between two variables from different seasons.
Question #1: Does aDOT drive wide receiver consistency?
Deep threats tend to get labeled as inconsistent whereas WRs whose targets are closer to the line of scrimmage are perceived as more reliable fantasy producers. Players like Will Fuller and John Brown are “best ball guys” because you never know when they’re going to break a 75-yard touchdown. No one is calling Jamison Crowder a best ball guy. However, there has been almost no correlation between average depth of target (aDOT) and CV over the last five seasons.
While there was no correlation for the sample as a whole, I wanted to specifically look at whether players like DeSean Jackson are less consistent than their lower-aDOT counterparts. So, I separated players into deciles by aDOT and found the average CV for each group.
For the most part, there does not seem to be a link between CV and aDOT. The sixth decile has a significantly lower average CV than the others, but that looks like an anomaly considering the fifth and seventh deciles both have an average CV that falls in line with the rest of the groups. Players with unusually high aDOTs do tend to be slightly less consistent, but that could be for a different reason that we’ll get to later.
Question #2: Is consistency consistent from season to season?
Put simply: No. CV in Year N explains 2.2% of CV in Year N+1. If you are using a player’s consistency in one year to predict their consistency for the following year, you are doing it wrong.
Question #3: If aDOT and previous-year CV don’t drive consistency, what does?
Fantasy points. We don’t have to project player consistency; we can just project player production and let consistency fall into place.
The relationship between CV and fantasy points becomes even clearer when you separate players by end-of-year finish.
Fantasy points are also slightly better than CV at predicting CV in Year N+1, but neither is very accurate. With an r-squared of just 0.05, 95% of CV is explained by factors other than previous-year fantasy points.
Earlier, we found that players with higher aDOTs had higher CVs in general. However, this result appears to be a product of fantasy production rather than actual inconsistency.
|Decile||Average CV||Average Fantasy Points|
As you can see, the deciles with the highest average CVs are among the lowest in average fantasy points. Given that there is no correlation between CV and aDOT, and there is a meaningful correlation between CV and fantasy production, this leads me to believe that aDOT is inconsequential when trying to project WR consistency.
Conclusion: In general, the best way to predict a player’s consistency is by projecting his season-long fantasy points total. In other words, consistency itself is not very predictable.
For the same reasons I outlined above, I required that all running backs included in this study had at least 100 carries or 30 targets in a single season. I chose those statistical thresholds, because they match the NFL’s requirements for a player to qualify for per-carry or per-target efficiency metrics. As with wide receivers, running backs had to stay on the same team for consecutive seasons to be included in tests that spanned across more than one season.
Question #1: Is consistency driven by player type?
First, I split the sample into three groups – three-down backs, two-down backs, and pass-catchers – based on the percentage of their opportunities that came through the air. For example, if a player had 50 carries and 50 targets, he would be defined as a pass-catcher because 50% of his opportunities were targets.
|Group||Percentage of Opportunities Coming Through the Air|
Three-down backs were notably more consistent than the other two groups, with pass-catchers barely edging out two-down backs for second place.
Logically, this makes sense: The production of pass-catchers and two-down backs is dependent on game script whereas three-down backs’ usage is more reliable. Furthermore, as Ryan Collinsworth detailed in Part 3 of his Passing Revolution series, most modern RB1s would fall into the three-down category, because they contribute equally in both facets of the game. So, it makes sense that the group that scores the most fantasy points would be the most consistent. Still, pass-catchers interestingly have a lower average CV than two-back backs despite scoring 20.4 fewer points per season on average.
|Group||Average Fantasy Points|
Question #2: Is consistency consistent?
At least compared to wide receivers, yes. Running back CV is much more stable year-over-year than wide receiver CV.
Question #3: Does that mean we should heavily consider previous-year consistency when evaluating running backs?
No. As was the case with wide receivers, running back consistency is largely driven by overall fantasy production.
Both PPR fantasy points and weighted opportunity are almost as good as CV at predicting CV in the following season.
This relationship becomes extraordinarily clear when you divide players into groups based on end-of-year position rank.
RB1s are significantly more consistent than RB2s, who are significantly more consistent than RB3s.
Conclusion: Running back consistency is driven by overall fantasy production. Although, pass-catching backs are more consistent than two-down backs because they have a lower average CV despite scoring fewer points.
The Big Picture
It may not seem like much, but freeing yourself from concerns about consistency and just attacking total points is a big deal. Stay tuned for Part 2 where we’ll look at a few players who are being under-drafted due to concerns about their lack of consistency.