Yards Created: Five Year Review

The NFL Draft is right around the corner, which means that we have a new crop of running backs to analyze. This is now my sixth year charting rookie running backs for Yards created and, if you’re new here, welcome!

Yards created is a statistic I came up with in 2016 and is the number of yards a running back creates on his own after the offensive line has (or has not) opened up a hole for the runner to get through.

So, let’s say a particular play gains 10 yards on the ground. If the offensive line opens up 3 yards of space and the running back gets through it, makes a linebacker miss, and gains 7 more yards — I chart that play as 3 yards blocked for the line and 7 Yards Created for the running back. Conversely, if the back is contacted three yards behind the line of scrimmage but manages to shake the would-be tackler and gain 10 yards for a first down, that would go down as -3 yards blocked and 13 Yards Created.

Not only do I chart runs, but I also log missed tackles forced (as a runner and receiver), receptions and routes run, and pass protection. When I first started doing Yards Created, I intended it to be used as a descriptive measurement. We always hear analysts say something along the lines of “he doesn’t create much yardage on his own” or “he forces a ton of missed tackles” — with very little data to backup their argument.

The one thing I did not expect was just how predictive Yards Created would become.

Over the last five years, Yards Created has correctly identified running backs like Christian McCaffrey, Saquon Barkley, Dalvin Cook, and Ezekiel Elliott, and Leonard Fournette as standout prospects that ended up going early in the draft. It’s also highlighted premier backs that were under-the-radar (at the time) like Alvin Kamara, Kareem Hunt, and Kenyan Drake. Each of these eight running backs rank inside of the top-12 RBs in my database in Yards Created per rush attempt (YC/A).

McCaffrey, Barkley, Cook, Elliott, and Kamara make up the top-5 backs in fantasy points per game since 2016 while Hunt is the RB12 and Fournette is RB15 in this span.

Of course, there have been some misses too. Darrell Henderson is second in YC/A and will almost definitely not live up to that billing with the emergence of Cam Akers. Unfortunately, it also looks like Anthony McFarland is going to be a swing and a miss after not being able to get on the field in his rookie season despite middling talent on the depth chart. McFarland is sixth in YC/A, just behind Elliott and Barkley. My numbers were similarly high on Joe Williams and D’Onta Foreman — both rank top-12 in YC/A in the database — but they saw their careers cut short because of injuries.

On the flip side, Yards Created was low on Derrick Henry (53rd-of-57 RBs in YC/A) and David Montgomery (45th) and both have made those numbers look very bad. The Henry miss especially hurts. To be fair, Henry’s low YC/A can somewhat be explained by usage. Nearly 75% of his carries at Alabama went in between the tackles and he gained just 3.3 YC/A on those attempts. (For reference, just over 62% of all college carries go up the middle). However, when Henry ran off-tackle, he was unstoppable — generating a monster 7.0 YC/A on his outside carries. That was the third-best figure in the 2016 class behind Elliott and Drake. He was always going to be an outlier.

Recapping the one-off hits and misses is fun, but just how predictive is Yards Created over the full sample? Well, I ran the numbers to find out below.

But first, I need to make it clear that Yards Created is not a magic bullet and that some of the success the YC has had might be a product of still not having enough data. Five years might seem like a long runway, but separating what is signal and what is noise is always a balancing act.

Because football is such a small sample, context-driven sport and there are hundreds of factors influencing each play, it is impossible to have a single statistic or a model that can dramatically increase your probability of making better predictions. In fantasy, the strike zone for success is being correct anywhere from 55-60% of the time over the long run. Hitting on 65% of your picks in a single season would be incredible (and not likely repeatable). Anyone that thinks that they can hit on 75% of their picks or sells themselves as such isn’t someone you want to take advice from. The idea is to be less wrong than our opponents and to maximize gains when we’re right.

Just how predictive is Yards Created?

To find out, I ran a simple linear regression test on career PPR points per game vs. Yards Created per attempt. The full sample is 57 running backs (the 2021 class was obviously not included). Then, I went through various college statistics — like rushing yards per game and receptions per game — along with a few combine metrics and tested them to see how well they predicted future career fantasy points.

So, how well do Yards Created stack up?

YC is more predictive than all of the college counting stats and combine metrics

(Notes: Correlation measures the relationship between two variables like how well rushing yards per game explains future career fantasy points. The higher the number, the stronger the relationship. All Yards Created data was tested against 2016-2020 NFL seasons. All of the college statistics vs. NFL output testing spans 2000-2020. The study only included running backs that were drafted. I also made the logical assumption to only include running backs that had at least 200 career NFL touches for the study. All of the college/NFL stats were sourced from Sports Reference.)

I again want to reiterate that there are two massively different sample sizes at play here. While there are just 57 RBs that I’ve charted over the last five years, the NFL sample of backs that I tested included 274 RBs. So, essentially the college counting stats vs. future production study included nearly five times more players than the Yards Created one. Ideally, I’d have way more Yards Created data to test but the small sample of work has certainly been promising.

Summary

• Yards Created are more predictive of future fantasy points than all college counting stats like receiving and rushing yards per game, but not by a massive margin.

• So far, YC/A does a two-times better job explaining future fantasy points than college YPC.

• College receiving yards and receptions does a better job at predicting future output than rushing yards. This shouldn’t be a surprise given how modern NFL backs score their points in the passing game.

• Make sure you realize that these stats all have fairly weak relationships with future success. Correlation of 0.8 would be amazing, but it’s impossible. This goes back to what I was saying earlier about the nature of the sport: It is very random and we need to embrace that.

• The Combine is pretty much worthless — or, at least, the testing is. The Combine still has value for teams because it allows them to meet with all of the invitees on a level playing field and go through questions, medicals, etc. But by and large, NFL Combine metrics do not materially predict future fantasy success. They’re fun to talk about! But noisy all the same.

• In fact, the 40-yard dash is negatively correlated to career NFL fantasy points per game. You’d be better off just picking names out of a hat than using 40 times to select your fantasy RBs.

• I did look into Speed Score’s relationship with career fantasy points and the correlation was much better (0.05). Speed score adjusts 40 times for weight, the heavier you are and faster you run — the better. Saquon Barkley and Jonathan Taylor both crushed Speed Score. However, my theory on why Speed Score is better than 40-yard dash times is because weight matters at the position. You don’t see many 185-pound RBs dominating in fantasy.

• Basically, the 40-yard dash is a barometer. “Can you run faster than a 4.7? Good! You’re probably fast enough for the NFL!” Burst inside of 2-5 yards is what matters most. And the vertical and broad jumps (somewhat) measure burst. Maybe that’s why they’re more useful than 40-yard dash times. Alvin Kamara crushed in the vertical/broad jumps.

• Lastly, if the Combine does a poor job of predicting running back success… beware of the Pro Day numbers that are coming out this year. If the Combine is done in a controlled environment and still provides a very weak signal (at best), just how bad must the Pro Day numbers be when it’s just a bunch of coaches using their stopwatches on fields that may or may not be doctored?

I’ll come back to this article in a few years and refresh the data for Yards Created. Hopefully it continues to do fairly well. In the meantime, be on the lookout for profiles on the 2021 class!