Receivers experience dramatic differences in the context of their attempted catches, but from a fantasy perspective, some of them wash out over time. Slot receivers tend to gain fewer yards on their catches, but they also tend to see more targets and catch a higher percentage of them. Deep threats see fewer targets and catch even fewer of them, but when they do, they accumulate chunk yardage and frequently score touchdowns to help make up for their intermittent production. Context is critical to understanding which receivers are actually helping and hurting their teams, which in turn influences their future volume of targets. But in the short term, a fantasy player could judge a receiver by his basic stats and rates and likely wouldn’t miss the mark too severely.
Things become much more difficult with running backs. Not only do backs play more varied types of roles with more diverse impacts on their potential for fantasy production, they also succeed and fail in concert with their blockers. Receivers can draw top cover corners or see some poor throws from their quarterbacks, but those rarely cost them more than a handful of catches over the course of a season. In contrast, a back who plays behind a poor offensive line will suffer that disadvantage constantly, and rates like yards per carry will punish him whatever his true share of the responsibility.
In an attempt to move past imperfect measures like yards per carry, I have created an expected rushing model that is similar in nature to my expected receiving model. It starts by establishing the difficulty of every rushing attempt based on six play factors.
Distance from the end zone
Just like for receivers, backs produce fewer yards the closer they make their attempts to the end zone. Not only do carries there have a yardage cap at the front plane of the end zone, they also have to face a defense that is compacted by the lack of vertical space on that end of the field. That makes it so even carries in the 3-4 yards from the end zone range average less than 2 yards per attempt.