Standard Running Back Rankings

The most important part of a fantasy football season is the draft. Everyone starts off with high hopes and a blank roster. Everyone has the same information and, with the exception of the first round, the opportunity to get any player they want.

Seasons are won and lost at the draft. Picking up a guy in the 7th round who goes on to finish as an RB1, or nailing a late round sleeper pick could propel you into the playoffs. On the other hand, if you miss on a first or second round pick, you’re starting off at a big disadvantage.

What I’ve said isn’t new. Each year, dozens of cheat sheets, draft kits and rankings come out.

But most of these rankings look the same.

Sure, some rankings might have a guy going a couple picks, or even half a round earlier than another set of rankings. At the end of the draft, the difference might be a couple of rounds. But for the most part, the rankings are very similar.  Everyone agrees that there are four running backs that stand out from the rest. And after that is another tier of about a half dozen guys who will go in the back of the first and the early second. After that are rookies, guys who have struggled, and backs dealing with competition.
For the most part, these rankings are reasonable. Todd Gurley, LeVeon Bell, David Johnson and Ezekiel Elliott are all very talented running backs, and the experts who rank them are very knowledgable about fantasy football.

But they’re human, and all humans have biases. Young running backs with upside to hit 1000 yards get taken ahead of older running backs that already do. Guys with a lot of name value get taken ahead of where they should go. If you drafted a guy last year who busted, chances are you’re going to hesitate before drafting him this year.

But what if these running backs could be predicted by machine learning? The computer doesn’t care what name a player has, and it won’t overreact to how a player performed on their fantasy team last year, because it didn’t have a fantasy team. Instead, it will make impartial predictions based on production, opportunity, age, injury history, combine results, and the team around them.

I gathered data from the 2008 to the 2017 seasons and fed them in. I used six models: Linear Regression, Logistic Regression, K Nearest Neighbors, Random Forest, Decision Trees, and Linear Discriminant Analysis. I then had two different ways of turning the results from those models into rankings: A) averaging the rankings from the models, and B) feeding the results into a second linear regression model. When I tested these methods, both models performed about the same, so the results from the simpler model are my official projections.

With the details out of the way, here are my rankings for 2018 running backs in standard leagues! The first column is the rankings from the simpler model. The second column is the rankings from the more complex model. The next two columns are each player’s predicted carries, and a letter grade representing how efficient I expect them to be at turning carries into points. (I also aggregated my rankings with ADPs found at Fantasy Football Calculator, so if the rankings don’t match up, that’s probably why).