With the start of the NFL season just around the corner and draft season upon us, I thought I would look into what statistic best predicts the future performance of NFL running backs for Fantasy Football.
In the same way that Fantasy Baseball players use FIP to predict the future performance of a pitcher’s ERA, we want to find a statistic that is a better predictor of success than a running back’s Fantasy points scored from the previous season. Whatever the statistic is, it’s ability to be a better predictor of future performance than past performance will result from its ability to strip any noise that is inherent from past production.
By no means did I cover all statistics, so the author of this article is aware of how dogmatic the title of this piece is, but I tried to cover as many metrics that were readily available through Pro Football Reference as possible.
Population and Methodology
The population of this study consists of all running backs that started a minimum of three games in a single season from 1989 to 2014. It’s necessary to set a minimum level of games started to weed out players that had no opportunity to perform to their true talent level.
In order to test which metric is the best predictor of future success for running backs, we need to look at a player’s previous season total for a particular metric and see how it correlates (r) with their Fantasy points scored for the following season (e.g. we would look at a player’s rushing touchdowns for 2013 and see how rushing touchdowns correlate to a player’s Fantasy points scored for 2014).
This approach gives us 1,229 pairs of seasons in which a running back started at least three games in each.
|Previous Season Statistic||Correlation (r)|
|Fantasy Points (Minus TDs) per Game||0.75|
|Fantasy Points (Minus TDs and Add Fumbles) per Game||0.75|
|Fantasy Points per Game||0.74|
|Fantasy Points (Minus TDs and Add Fumbles)||0.73|
|Fantasy Points (Minus TDs)||0.73|
|Fantasy Points (Add Fumbles)||0.72|
|Fantasy Points (Add Fumbles)||0.72|
|Rushing Touchdowns per Game||0.61|
|Rushing Attempts per Game||0.58|
|Receptions per Game||0.45|
|Receiving Yards per Game||0.44|
|Yards per Carry||0.35|
|Receiving Touchdowns per Game||0.17|
Most of the metrics that I tested should be self explanatory, but I’ll go over some of them now. This link provides a good explanation of how these numbers should be interpreted.
There is a common discourse in the fantasy community that “you shouldn’t go chasing touchdowns,” and this research suggests that that narrative is in fact true.
The best predictor of future success for running backs proves to be the previous season’s Fantasy points-per-game, with the points gained from touchdowns backed out.
Second Best Predictor:
Just behind that is a player’s fantasy points-per-game, minus points gained from touchdowns and adding points lost from fumbles.
When I started the project, this is the statistic that I assumed would be the best predictor of future performance, however this stat becomes difficult to test because Pro Football Reference only has data on fumbles-lost since 2012.
As a result, when Pro Football Reference lists a player’s Fantasy points scored for a previous season before 2012, they assume that half of a player’s fumbles were lost to the other team and adjust their Fantasy points from there.
In all reality a player may have fumbled five times and never lost the ball to the other team, but Pro Football Reference penalizes him for two-and-a-half fumbles when they calculate his fantasy points. This is actually a fair assumption, because we know that a team’s ability to recover a fumble is entirely random and not a skill, which lead me to want to test this stat in the first place (i.e. because the amount of points a player loses from fumbles is mostly random, I thought that if you added back in the points lost from fumbles, you could better predict Fantasy points scored for the following season).
Instead of fumbles lost to the opposing defense, we are left to test fumbles (that is, fumbles divided by two) in general when we look to weed out variables. Oddly enough we still get the second highest correlation of our group of statistics when we add the points lost from fumbles back into a player’s points per game, minus touchdowns.
Applying What We’ve Learned
|Player||Fantasy Points||Fantasy Points per Game||Fantasy Points (Minus TDs)||Fantasy Points (Minus TDs) per Game||DIFF Points per Game||DIFF RK|
The chart above shows what the 2014 scoring leaders at running back look like when we back out their touchdowns and judge their performance on a per game basis.
Obviously, players like Marshawn Lynch and DeMarco Murray—the league leaders in rushing touchdowns—are hurt quite a bit, but, because their total production wasn’t completely dependent on touchdowns (i.e. a large percent of the points they produced came from the yards they gained, and not just touchdowns), their overall rank doesn’t change much.
Players like Matt Asiata and C.J. Anderson are hurt the most, because an appreciable amount of their production came from end-zone celebrations.
(What follows was added to this article on August 27th to account for what some commenters noticed (i.e. that Fantasy points per game minus touchdowns fails to account for players that were active in games but saw little involvement in the offense for that game))
However, when it comes to C.J. Anderson, the former statement is a half truth, which reveals a shortcoming of Fantasy points-per-game, minus touchdowns. We use games played as a way to adjust player performance to take into account differences in playing time between players, however just because a player played in a game, it doesn’t necessarily mean that a player was heavily involved in the game. For example, C.J. Anderson played in played in 15 games in 2014, but he didn’t see more than five carries in a game until mid-way through the season.
To account for this and have a better proxy for playing time, we can use total touches (rushes plus receptions) instead of games played when we adjust for playing time differences. The table below shows Fantasy points-per-touch, minus touchdown points for all running backs with 120 or more touches in 2014.
|Player||Age||Tm||Fantasy Points per Touch (Minus Touchdowns)||Touches|
As we can see, if we adjust performance by the times a player touched the ball, and not the amount of games they played in, we get a much more accurate depiction of a players true talent (C.J. Anderson ranks fourth in the latter metric).
While we have learned that the best predictor of future performance for Fantasy running backs is fantasy points scored per game, minus touchdowns, we have also learned that this metric tells somewhat of a fib and that total touches is better than games played if we want to adjust for differences in playing time.
- Week 13 NFL Breakdown: Joe Flacco Shouldn’t Make You Think Twice - December 7, 2016
- Week 12 NFL Breakdown: Mark Ingram Rolls Against the Rams - November 30, 2016
- Week 11 Breakdown: DeVante Parker Could Be Turning the Corner - November 23, 2016