Fantasy Football Lineup Strategy: Running Backs

By Kurt Bullard

For most fantasy football fanatics, lineup tinkering is an every week occurrence come Sunday morning. Before heated matchups and a Sunday spent sitting on the couch watching football, these managers take one last look at their lineups, perhaps making a change or two before submitting the lineup and letting the Fantasy Gods take over.

One of the most common reasons for lineup tinkering is when one of your better starters is put up against a tough defense, while someone on your bench is up against one of the softest defenses in the league, just begging to be put in the starting lineup to run over a helpless defense.

In FXX’s The League, tinkering often leads more to trouble than it does to good. A Sunday spent on your smartphone sweating over your lineup generally bears no fruit. But is that the actual reality in fantasy football? Should a manager go with his entrenched starter facing, for example, the Seahawks D, or should he make a call to the bench to replace him with someone going against, per se, the Bears?

As a starting point for this question, I looked exclusively at running backs during the 2014 regular season. I didn’t look at quarterbacks and wide receivers quite yet because it’s a little tougher to use broad defensive stats to measure the challenge, let’s say, a wide receiver faces on any given Sunday; while the defense as a unit could be subpar, that defense may have a shutdown corner who takes out a team’s No. 1 threat each game. So, for the time being, the analysis is only to be done on running backs.

I looked at the top 100 scoring running backs from last season for each game that they played in. For each game, I looked at how they played compared to their season average by calculating the difference between that week’s points and their per-game average during the year.

I then regressed these residual week totals against the defensive DVOA of the team as well as the difference between the offensive and defensive DVOA of the relevant parties to see if playing a bad defense was actually a boost to a running back, and, if so, how much of a boost it was.

For example, Lamar Miller, who averaged 11.6 points per game last season, played the Pats in Week 15 and scored 5.7 fantasy points, which was 5.9 lower than his season average. I regressed -5.9 points against the Pats’ defensive DVOA , .102 (defensive strength and defensive DVOA are negatively correlated), and the difference between the Dolphins offensive strength and the Pats’ defensive strength (-.034+.102 = .068).

The regression returned the following results:


As is shown here, the defensive DVOA is a significant variable in predicting how a running back performs on a week-to-week basis if you know, on average, how good of a performer is. On the other hand, the difference between the strength of an offense and strength of a defense was not significant – it only mattered how good the defense was in an absolute – not relative – sense.

However, when looked at in deeper depth, the regression tells a different story. The effect of the defensive strength of an opponent is actually not that large on average. The expected points gained from starting a running back going against the league’s worst defense last year instead of a league-average D would have been 1.54 points on average (9.47*.163). That’s not a huge jump and not a reason to panic over who to start one week.

Perhaps more importantly, the defensive strength of a team only explains 1.38% of the variation in the data. In other words, on an individual week-by-week basis, knowing the opponent’s DVOA on defense would tell you very little about how the running back actually did that week. Whether it be other factors – weather, specific scheming that week, or just general randomness in performance, defensive strength tends to get clouded out.

So go on, Ruxin and the like, keep on tinkle tinkering. But don’t overthink it.

About the author


View all posts


Leave a Reply

Your email address will not be published. Required fields are marked *