**Fantasy Football Analytics » R | Fantasy Football Analytics**, and kindly contributed to R-bloggers)

The popular wisdom is that “massive workloads” result in a decline in running back productivity in the following season. For example, anecdotal evidence using Corey Dillon (2004) and Jamaal Anderson (1998) are often presented as sure-fire confirmation of running back regression. The most recent example debunking this line of thinking was found in High-Volume Running Backs, Rotoworld, August 24, 2014. In the article the concept of running backs going into decline following a high-rushing-volume season was revisited and partially rebuffed.

The analysis recognized selection bias due to health concerns and productivity, but indicated the results were “*overwhelmingly in support of high-volume running backs regressing in the year following a heavy workload*.” For example, the article notes that running backs coming out of a 300+ carry season experienced:

- 15% decline in games started
- 23% decline in touchdowns
- 22% decline in rushing attempts
- 25% decline in rushing yards

An important consideration is that these statistics reflect the average of the sample. For example, the average carries in the year following a 300+ carry season was 261 carries, approximately 21% below the prior year’s campaign.

However, not all players face a decline in productivity following a 300+ yard campaign, and many players are highly productive for multiple seasons. As a result, it is important to characterize the distribution of player performance **over their playing career** (and following a 300+ carry campaign) when seeking to answer the question: do massive workloads equal big declines in running back productivity?

**Reviewing the Distribution of Player Workload (300+ Carry Rushers)**

In order to expand the High-Volume Running Backs, Rotoworld, August 24, 2014 analysis and examine the entire distribution, player performance data was collected from Yahoo Sports from 2003 to 2013 (VBA script). Next, the data was filtered to examine only players who had 300+ carries during a particular season. For each player, the year in which the player first crossed the 300+ carry threshold was identified. Then the seasonal deviations from the first 300+ carry season was calculated, e.g., negative (positive) value if the season was prior to (after) the 300+ campaign. (see RB Data)

Finally, the data was loaded into R to estimate the linear relationship (using Ordinary Least Squares, OLS) between the seasons removed from the first 300+ carry season and:

- Deviation in games played
- Deviation in total carries
- Deviation in total yards from scrimmage (rushing)
- Deviation in yards per game
- Deviation in yards per carry
- Deviation in touchdowns (rushing)

**Impact on the Performance Metrics in the First Nearby Season**

In order to determine how NFL running backs performed in the season immediately following a 300+ carry campaign the data was parsed to include only the deviations associated with the 300+ carry season and the season immediately following the 300+ campaign (first nearby season). The data was plotted and OLS estimates of performance metrics during (year 0) and after (year +1) were modeled using the year deviation as the conditioning information. Figures 1 through 5 identify the scatterplot and OLS results (R jitter term included to define node separation).

**Figure 1. Deviation in Games Played and Deviation from First 300+ Carry Season**

**Figure 2. Deviation in Carries and Deviation from First 300+ Carry Season**

**Figure 3. Deviation in Yards from Scrimmage (Rushing) and Deviation from First 300+ Carry Season**

**Figure 4. Deviation in Touchdowns and Deviation from First 300+ Carry Season**

**Figure 5. Deviation in Yards per Carry and Deviation from First 300+ Carry Season**

Results of the single-season view indicate that **on average** running backs do regress in some statistical categories in the year following a 300+ carry campaign. High volume running backs **statistically** receive **fewer opportunities** in the running game in the year following a heavy workload. OLS coefficients indicate that in the year following a 300+ carry campaign running backs:

- Play 3.2 fewer games (statistically negative, P-value <0.000, R-squared 0.24)
- Have 179 fewer carries (statistically negative, P-value <0.000, R-squared 0.51)
- Gain 837 fewer rushing yards on the season (statistically negative, P-value <0.000, R-squared 0.48)
- Gain 40 fewer rushing yards per game (statistically negative, P-value <0.000, R-squared 0.39)
- Gain 6 fewer touchdowns (statistically negative, P-value <0.000, R-squared 0.17)

However, there remain exceptions to the rule. Scatterplot observations above the x-axis represent players who improved following a 300+ carry campaign. Thus, while on average, results indicate the potential for a reduced workload, empirical data does not support it with 100% certainty –* there will be exceptions to the rule*. Additionally, these results *do not* indicate regression in physical abilities. In fact, no statistical sign of physical regression is found. The OLS parameter estimate on yards per carry (Figure 5) was not statistically different from zero, indicating that there is no significant difference in yards per carry following a 300+ carry campaign.

Players can be just as productive, but the data indicates they are simply not given the same workload. The drop-off in workload can be attributable to a number of factors. For example, coaching philosophy (more rest to maintain productivity), substitute players in the offense (wide receiver, and quarterback productivity), game outcomes (rushing less when playing from behind) may all impact the number of times a player’s “number is called” in the offense. These results suggest that while legitimate concerns exist over health (# games does statistically decline), if running backs remain healthy their yards per carry productivity remains in line with their historical performance — before and after the 300+ carry campaign.

What fantasy owners should consider, as a result of this analysis, is that a player’s number may not be called as often in the season **immediately** following a 300+ carry campaign. But what about in the following season, the year after, and the year after that? Let’s look at the numbers.

**Bouncing Back – Big Workloads Can Get Bigger **

Results of the analysis are clear: *There is a statistically significant, and negative, trend on most of the output metrics following a player’s first 300+ carry campaign*. How does this impact high volume tailbacks such as A. Peterson, A. Morris, A. Foster, C. Johnson, D. Martin, F. Gore, L. McCoy, M. Lynch, M. Forte, R. Rice, and S. Jackson? In order to determine if the drop-off after a 300+ carry campaign means a player will continually regress season-over-season multiple seasons of player data was analyzed. Data was collected for all 300+ carry rushers, then their game and season level performance metrics were evaluated to determine if a pattern was obvious. I seek to answer the question: How well does a high volume running back perform two, three, or even four seasons after eclipsing the 300+ carry mark?

Data for 41 players spanning 10 NFL seasons were plotted and OLS estimates of performance metrics as a function of the deviation from the first 300+ carry season were estimated (R jitter term included to define node separation). Analyzing figures 6 through 10 we see that not all variables are statistically significant and the fit of the OLS regression lines are relatively poor in all five models (low R^{2}). Additionally, the scatterplot results confirm that it is possible for players to bounce back after a 300+ carry campaign. **Massive Workloads Can Get Bigger. ***Due to the poor fit of the model it is likely that other conditioning information may better explain how a running back performs in the years following a 300+ carry campaign (e.g. player age, role in the offense due to substitute players, coaching philosophy and strategies, weather conditions, and health). *

Positive increases in games played, touchdowns, total carries, total yards, yards per game, and yards per carry were observed in the data.** **Scatterplot observations in the upper right-hand quadrants of figures 6 to 10 are circumstances where a player’s performance improves year-over-year as the distance from the first 300+ carry campaign increases.

For example, in seasons following a player’s 300+ carry campaign

- 35% of players played more games than in the previous season
- 43% of players had more carries, gained additional yards, and had higher yards per game statistics compared to the previous season
- 42% of players had a higher yards per carry compared to the prior season
- 44% of players scored more rushing touchdowns than in the previous season

**Figure 6. Deviation in Games Played and Deviation from First 300+ Carry Season (Multi-Season)**

**Figure 7. Deviation in Carries and Deviation from First 300+ Carry Season (Multi-Season)**

**Figure 8. Deviation in Yards from Scrimmage (Rushing) and Deviation from First 300+ Carry Season (Multi-Season)**

**Figure 9. Deviation in Touchdowns and Deviation from First 300+ Carry Season (Multi-Season)**

**Figure 10. Deviation in Yards per Carry and Deviation from First 300+ Carry Season (Multi-Season)**

Results of the multi-season analysis indicate that **on average** running backs do regress in some statistical categories (carries, yards, and touchdowns) in the years following a 300+ carry campaign. To determine how responsive the dependent variables are with respect to the distance from the first 300+ carry campaign OLS model coefficients are presented. Model results indicate that for each season beyond a player’s first 300+ carry campaign:

*The player plays 0.20 fewer games (not statistically significant, P-value 0.10, R-squared 0.01)*- Total carries for the season decline by 14 (statistically negative, P-value <0.000, R-squared 0.09)
- Total yards for the season declines by 61 yards (statistically negative, P-value <0.000, R-squared 0.08)
*Yards per carry decline by 0.03 (not statistically significant, P-value 0.22, R-squared 0.01)*- The player scores 0.47 fewer touchdowns (statistically negative, P-value <0.000, R-squared 0.04)

Extrapolating out to five years, and including only the statistically significant results, after a player has the first 300+ carry campaign, OLS model results suggest:

- Total season carries would decline by 70
- Total yards for the season would decline by 300
- The player would score 2.5 fewer touchdowns

There is no statistical evidence that players consistently play fewer games, or that players have a lower yards per carry average. The only conclusive result is that opportunities decline. Opportunities could decline for a variety of factors (described above). Nevertheless, in the years following a 300+ carry campaign players continue to suit up and many maintain the high level of efficiency when utilized. It’s worth noting; however, that many players also regress.** **Scatterplot observations in the lower right-hand quadrants of figures 6 to 10 are circumstances where a player’s performance declines as the distance from the first 300+ carry campaign increases. From the data it is apparent that, on average, a larger percentage of players regress in years following their 300+ carry campaign. This is a reality of football, right Cap Rooney?

**Conclusion**

The key takeaway of this analysis is that (i) players do regress in the nearby season following a 300+ carry campaign, but (ii) players have the potential to bounce back, and massive workloads can get bigger. However, as a general rule of thumb running backs tend to regress as they move beyond their first 300+ carry campaign. The overall regression in player performance is less likely to be attributable to whether or not a player eclipsed 300+ carries in a season (and how long ago it was) and is more a function of a variety of factors such as player age, role in the offense due to substitute players, coaching philosophy, game strategy, weather conditions, and health.

These results have significant implications for players on the tail-end of their career. In the 2003-2013 sample, the maximum number of seasons played for players eclipsing the 300+ carry mark is 10 years. The in-sample average of seasons played is 6 years. This means that many players are out of the league within 6 years, and all are out of the league within 10 (at least empirically). This is information fantasy managers should consider carefully as they plan their draft strategy and invest their early picks on a running back. For example, it’s worth noting that Adrian Peterson is entering his 8^{th} season, Arian Foster is entering his 6^{th}, Frank Gore is entering his 10^{th}, Marshawn Lynch is entering his 8^{th}, and Steven Jackson is entering his 11^{th}. Will these players continue to out-perform their peers and get heavy workloads? Alternatively, both Alfred Morris and Doug Martin are entering their 3^{rd} season. Given, these players only recently had a 300+ carry campaign are these less risky running backs and more likely to have a “bounce back” year?

These are questions each fantasy manager must answer when considering which players to draft. In this post I have presented empirical evidence which suggests that NFL running backs **statistically** regress following their first 300+ carry campaign; but that many running backs have bounce back potential. Massive workloads can get bigger, and as many as 40% of players improve their productivity in the out years following a 300+ carry campaign – indicating that performance regression is not certain year-over-year.

**The R Script for nearby season analysis is below:**

https://github.com/New10/RCode/blob/master/Big_Workload_Nearby.R

**The R Script for entire analysis is below:**

https://github.com/New10/RCode/blob/master/Big_Workload.R

**The Visual Basic Script for entire analysis is below:**

https://github.com/New10/Visual-Basic/blob/master/YahooScrape

The post Do Massive Workloads Equal Big Declines in Running Back Productivity? appeared first on Fantasy Football Analytics.

**leave a comment**for the author, please follow the link and comment on their blog:

**Fantasy Football Analytics » R | Fantasy Football Analytics**.

R-bloggers.com offers

**daily e-mail updates**about R news and tutorials on topics such as: Data science, Big Data, R jobs, visualization (ggplot2, Boxplots, maps, animation), programming (RStudio, Sweave, LaTeX, SQL, Eclipse, git, hadoop, Web Scraping) statistics (regression, PCA, time series, trading) and more...