High-variance, volatile football players, the Will Fullers and Amari Coopers of our day, the risky bets — we all know those players, and most of all we all know about what they can do — for good and for bad.
When was the last time you trusted Will Fuller or Amari Cooper to go off the charts and they let you down? You probably don’t have to go too far back in time to find out. Just this past season, Fuller went all the way to reach 53.7 PPR-points in Week 5. He surpassed 11.1 points just once in the six games he played after that. The question is, can we use this volatility to our advantage? Let’s look at the concept from a more analytical perspective and see whether we can use it going forward in our fantasy football leagues.
How To Define Volatility
Any player with at least two games played during the season would have generated at least two PPR marks, which we can use to find an average. We can also get a sense of how volatile a player by calculating the standard deviation of that set of two games. What we get is the variance of those couple of performances compared to the average.
Above I’ve plotted Amari Cooper’s and D.J. Moore’s 2019 seasons on a game-by-game basis. When all was said and done, both Cooper and Moore had averaged the exact same 15.5 PPR/G, but as you can see they reached that point in very different ways. Cooper’s finishes fall in a wide range, while Moore was a steady performer.
If we look at their volatility marks over the season, that’s exactly what we see: Cooper’s volatility was at 11.6 while Moore’s finished at 5.8. In other words: Cooper was your 2019 boom/bust WR play while Moore fell into the safe-bet category of players.
How Does Volatility Relate To Fantasy Production?
The first question that comes to mind when trying to understand and assess volatility and its implications is how is it related to actual fantasy production. Are volatile players better overall performers than stable ones? Is volatility tied to average scoring in any way? Let’s see.
I will be using a dataset containing every game played by every QB/WR/RB/TE since 2000. There are 11,001 player-seasons in it. Of those, 9,221 included volatility data (min. 2 games played). To get things started, this is how each of those player-seasons went in terms of PPR (best players at the top) and VOL (most volatile players at the right side).
The relation between PPR and VOL here is quite high with an R-squared of 0.65. Of course, it’s easy to see in the above graph that the relationship is stronger at lower fantasy point levels. In order to avoid any potential results skewing we can focus solely on fantasy-relevant players (those average at least 10 PPR/G over their seasons with at least 10 games played). So, what happens if we apply that filter?
While there is still something, the relationship is much weaker (the R-squared value sits at 0.09 between PPR and VOL for fantasy-relevant player-seasons). While it’s true that higher fantasy scores correlate with volatility, once you get into the range of scoring that’s useful for fantasy owners, the significance of volatility decreases substantially.
It makes sense that we would see higher volatility levels at higher scoring averages: Given that fantasy points have a defined floor of zero1 but no defined ceiling (at least in theory), players who have higher PPR averages also have a wider range of possible scores.
Does Volatility Impact Positions Differently?
In order to drill down a little bit more, I have separated the players inside the dataset by position. Here is how they compare in terms of volatility and production:
Not a lot of changes when compared to the full picture:
- Quarterback PPR and VOL have R-squared value of just 0.05, by far the lowest among the four positions.
- Running back PPR-VOL relation goes up to 0.23, the highest among all four positions.
- Wide receiver and tight end PPR-VOL relations are both at 0.22.
How Stable Is Volatility From One Year To The Next?
Now, even if we assume some relationship between volatility and fantasy scoring, there is still a question of whether volatility is predictable. Are there are really volatile players, or is volatility an effect of randomness? To answer that question, we can look at how volatility changes from one year to the next one and try to find a relation between both values.
For the sake of comparison, let’s look at how the other main variable we are interested in — PPR points — correlates from one season to the next. We would expect good players to keep their level over their careers, no matter how volatile they are in getting their fantasy tallies. Here is how PPR in Year N relates to PPR in Year N+1 (data from all players with at least a game played since 2000 who averaged 10 or more points per game).
As expected, the R-squared yields a decent 0.25 value when looking only at fantasy-relevant players. About a quarter of the PPR-points in Year N+1 are explained by the prior season tally, which is reasonable when you consider we’ve eliminated all the lowest-scoring players who would artificially boost the numbers here.
Does something similar happen when looking at volatility, though? Do volatile players “stay volatile” from one year to the next?
The relation from one year to the next in terms of volatility is almost nonexistent. The R-squared of 0.06 is even lower than the one between PPR-VOL (0.09), which means volatility is not repeatable from season to season and therefore is not something we could chase even if we wanted to.
Volatility In Historical Context, And What To Do With It
The lack of stability from one season to the next doesn’t necessarily fit with our intuitions about certain players — the examples of Will Fuller and Amari Cooper I mentioned at the beginning jump out as players we all recognize as volatile. So why doesn’t this show up in the data? It’s possible looking at one season worth of games is too small of a sample to give us a real sense of how volatile a player actually is.
I have trimmed the data set to just player-seasons from 2010 on with at least 10 games played and players with at least three seasons of NFL experience. Here are a few charts with volatility and average scoring measured over the course of a player’s career. The colors mark how “stable” a player is in a diverging scale going from green (most stable) to red (most volatile).
The players we expect to see on the far right side of the graph — the Amari Coopers — do in fact show up there, suggesting our intuition is right. And indeed, players who score a lot of points are more volatile overall than players who do not at the career-level, which, as suggested above, makes perfect sense.
For a slightly different perspective, here is the same data, split by position, showing only players who averaged at least 10 PPR per game over their careers:
The thing to note about these charts is that, while high-scoring players tend to be more volatile than low-scoring players, broadly speaking, the top scorers at all positions have a wide range of volatility outcomes. The takeaway is that you shouldn’t shy away from getting players with reputations for volatility on your roster — they tend to be among the highest scorers. But at the same time, you shouldn’t chase volatility, because it’s probably impossible, and even if you could, top fantasy scorers are found at many volatility levels.
Image Credit: Andrew Dieb/Icon Sportswire. Pictured: Amari Cooper.
- Not exactly true, as they can get into the negative side, but that doesn’t happen very often. (back)