Things have changed a lot over the 20 years I've been playing fantasy football, but one constant has been the importance of running backs. The problem, of course, is that the proliferation of committees, the increasing rarity of true workhorse backs, and the popularity of flex spots has made it downright hellish trying to figure out who to start each week outside of the top 10 or so. And, unfortunately, as was the case a couple of weeks ago when I introduced my system for projecting quarterbacks, the best we're able to do is explain about 20 percent of the variability in week-to-week RB scoring. The good news, though, is that 20 percent is nothing to scoff at, to the point that it makes this whole endeavor certainly worthwhile.
how to project weekly points for running backs
The analysis that led to this RB system followed the same general principles and methods I used for quarterbacks. I got data for all RBs who had at least 125 touches and ranked in the top 48 at any point in the season from 2007 to 2011. (Note: I didn't include fullbacks unless they were hybrids like Marcel Reece. You're never starting a fullback otherwise.) All factors were calculated to be what they were at the time. For instance, the row associated with Week 5 of Frank Gore's 2007 season used his 13.1-point average going into the game against Baltimore rather than the 12.7-point average that he had at the end of the season. Again, the purpose of doing this kind of thing was to make sure I was predicting his Week 5 score using the information that was available at the time (e.g., average going into the game), not retroactively explaining how he achieved that score using information (e.g., end-of-season average) that already has it baked into the cake. Speaking of scoring, I used Football Guys' standard: 0.1 points per rushing or receiving yard and 6 points per rushing or receiving touchdown.
After gathering all the data, I ran a series of multivariable regressions to find the optimal prediction equation, and then tested the accuracy of that equation on the qualifying running backs in 2012. So, without further ado, onto the results.
First up, the things that didn't turn out to be important for projecting weekly RB scores. Once again, things like home/road and bye weeks aren't predictive: RBs playing at home or coming off of a bye don't score more than RBs playing on the road or not coming off a bye. Also a repeat of the QB analysis, RBs who are listed as probable don't score less than RBs who aren't on the injury report at all.
A couple of other unimportant factors were a bit more surprising. Chief among them is the over-under. You would expect that higher real-world scoring means higher fantasy scoring across the board. But, while that indeed was the case for QBs, it wasn't for RBs. The most likely explanation, at least to me, is that high-scoring NFL games come by virtue of fireworks through the air, not on the ground. If a game ends 41-35, it's probably because the QBs combined for 700 passing yards and 6 touchdowns. Any stat-padding for RBs was probably the result of burning clock in the fourth quarter. The other factor that made no difference was the back's Speed Score, which is Football Outsiders' measure of size-speed combination. Although being big and fast is an indicator that reliably separates great backs from the rest of the pack, it's of little use for identifying great weekly scores.
As far as the factors that do help projections, they're as follows:
- Age -- Younger RBs score more on a weekly basis than older RBs.
- Average -- RBs with a higher average going into the game score more than those with a lower scoring average. And, just as a reminder, I count David Dodds' preseason projection as a "game" here.
- Opponent's Average -- The more fantasy points that a defense has allowed to RBs in past games, the more they will allow in the upcoming game. Of note here is that I also include the defense's preseason projection as a "game." You might ask, "What preseason projection? Dodds only lists how many fantasy points the team defense will score per game. There's no projection for points allowed to running backs!" Excellent question. Although it's true that I don't know the points allowed number directly, it turns out I can use regression analysis to estimate it pretty well from the stats that make up fantasy scoring for team defenses. And since I'm such a nice guy, here's the equation if you ever want to calculate this yourself: Projected Points per Game Allowed to RBs = -2.17 + (.033 * Projected Yards Allowed per Game) + (.408 * Projected Points Allowed per Game).
- "Questionable" Injury Status -- On average, RBs lose about 2 fantasy points when they're listed as "questionable" on the injury report.
- Implied Win Probability -- As a reminder, this comes from the game's point spread. RBs playing on teams that are more likely to win their game that week score more than those on teams who are less likely to win. That this factor is important should surprise no one. I will add, however, that the effect for RBs doesn't have a curve in it like it does for QBs; there's no point of diminishing returns here.
- Promotion/Demotion -- This is analogous to the "new starter" factor I use for QBs. It's purpose is to fill in the blind spots of Dodds' projections because there are instances where he could not have known a starter would get hurt, and an unknown backup would become the starter. Keep in mind, however, that I use this very conservatively. In today's NFL, RB roles are dynamic from week to week, and often depend on specific in-game circumstances. For instance, LeGarrette Blount's weekly touch percentages so far this season have gone like this: 19%, 19%, 41%, 31%. Did he get promoted two weeks ago? No. In Week 3, the Patriots were up big in the second half. In Week 4, Stevan Ridley was dinged up, and Bill Belichick chose to ride the larger back against a soft Falcons front seven. Blount's "promotion" in my system only arrives this week because Ridley is out for Sunday's game.at Cincinnati. He will be going from clear backup to clear starter, and that's the information I want to capture when applying this factor.
Speaking of touch percentage, you might be surprised to see it omitted from the above list of important factors. The reason is that the information we get from touch percentage is already incorporated into one of the other factors: the RB's scoring average going into the game. For the more stats savvy readers out there, the other issue is that the correlation between touch percentage and scoring average is 0.79, which means that using both in a regression analysis would introduce the problem of multicollinearity.
How well does this system work? Here's a table showing the average errors and accuracy rates for predicting whether the RB would score above or below his average going into the game:
| Period | Error | Right? |
|---|---|---|
| Wk 1 | 5.4 | 44% |
| Wk 2 | 4.7 | 63% |
| Wk 3 | 4.5 | 63% |
| Wk 4 | 5.9 | 48% |
| 2013 | 4.9 | 56% |
| 2012 | 5.2 | 56% |
| 2007-2011 | 5.5 | 57% |
Compared to the QB table I posted on Thursday, the RB system is slightly more accurate in terms of error points, but it's slightly less accurate in terms of the binary plus-minus prediction. Most importantly, however, its accuracy has been every consistent over the past seven seasons, which is what you want to see in a predictive system.
week 5 rb projections
Here are my projections for the top 48 running backs going into Week 5, minus C.J. Spiller, Fred Jackson, and Chris Ogbonnaya, each of whom played on Thursday night:
| RB | Tm | Avg | Matchup% | Matchup%LO | Matchup%HI | Points | PointsLO | PointsHI | Direction? |
|---|---|---|---|---|---|---|---|---|---|
| LeSean McCoy | PHI | 17.4 | 2.6% | 2.0% | 3.1% | 15.8 | 12.3 | 19.3 | LOWER |
| Jamaal Charles | KC | 17.7 | -0.8% | -1.7% | -0.1% | 15.2 | 11.6 | 18.7 | LOWER |
| Matt Forte | CHI | 15.9 | -0.3% | -0.4% | -0.2% | 14.0 | 10.5 | 17.4 | LOWER |
| Reggie Bush | DET | 16.7 | -5.3% | -8.2% | -2.8% | 13.8 | 10.0 | 17.6 | LOWER |
| Marshawn Lynch | SEA | 15.5 | -3.1% | -4.9% | -1.5% | 13.5 | 10.0 | 17.0 | LOWER |
| LeGarrette Blount | NE | 4.7 | -2.0% | -3.0% | -1.2% | 13.0 | 8.4 | 17.7 | HIGHER |
| DeMarco Murray | DAL | 13.2 | -3.1% | -6.7% | -0.1% | 12.4 | 8.8 | 16.0 | LOWER |
| Arian Foster | HOU | 13.3 | -1.2% | -4.6% | 1.6% | 12.4 | 8.8 | 16.0 | LOWER |
| Rashad Jennings | OAK | 4.0 | -8.6% | -14.6% | -4.7% | 11.7 | 6.8 | 16.7 | HIGHER |
| Eddie Lacy | GB | 8.2 | 13.8% | 10.1% | 16.8% | 11.7 | 7.9 | 15.4 | HIGHER |
| Knowshon Moreno | DEN | 10.4 | 5.1% | 2.6% | 7.4% | 11.5 | 8.1 | 14.8 | HIGHER |
| Frank Gore | SF | 11.7 | 2.4% | -0.5% | 4.8% | 11.4 | 7.7 | 15.1 | LOWER |
| Giovani Bernard | CIN | 10.7 | -3.5% | -4.9% | -2.3% | 11.2 | 7.8 | 14.7 | HIGHER |
| Trent Richardson | IND | 10.5 | -5.5% | -8.0% | -3.5% | 10.9 | 7.4 | 14.4 | HIGHER |
| Darren Sproles | NO | 11.6 | -3.6% | -5.1% | -2.4% | 10.7 | 7.2 | 14.2 | LOWER |
| Johnathan Franklin | GB | 7.1 | 15.1% | 11.0% | 18.6% | 10.7 | 7.1 | 14.2 | HIGHER |
| Joique Bell | DET | 11.4 | -6.9% | -10.7% | -3.7% | 10.6 | 7.2 | 14.0 | LOWER |
| Daryl Richardson | STL | 7.1 | 11.8% | 8.0% | 14.8% | 10.5 | 6.9 | 14.0 | HIGHER |
| Ryan Mathews | SD | 9.1 | 1.0% | -1.5% | 3.3% | 10.1 | 6.9 | 13.3 | HIGHER |
| Ray Rice | BAL | 8.3 | 5.1% | 3.1% | 6.8% | 10.1 | 6.9 | 13.2 | HIGHER |
| Lamar Miller | MIA | 8.8 | -4.4% | -7.0% | -2.2% | 10.0 | 6.6 | 13.4 | HIGHER |
| Bilal Powell | NYJ | 10.0 | -9.9% | -15.3% | -5.3% | 9.8 | 6.3 | 13.3 | LOWER |
| Bernard Pierce | BAL | 6.8 | 5.3% | 3.3% | 7.0% | 9.6 | 6.3 | 13.0 | HIGHER |
| Danny Woodhead | SD | 8.8 | 1.1% | -1.6% | 3.4% | 9.6 | 6.3 | 13.0 | HIGHER |
| Ronnie Hillman | DEN | 6.2 | 6.3% | 3.2% | 8.8% | 9.4 | 6.0 | 12.8 | HIGHER |
| DeAngelo Williams | CAR | 10.0 | -7.2% | -10.9% | -4.2% | 9.3 | 5.8 | 12.9 | LOWER |
| Maurice Jones-Drew | JAX | 7.1 | 3.8% | -5.1% | 10.6% | 8.8 | 5.1 | 12.5 | HIGHER |
| Isaiah Pead | STL | 4.5 | 14.2% | 9.6% | 17.9% | 8.7 | 5.3 | 12.0 | HIGHER |
| David Wilson | NYG | 5.4 | 4.0% | 3.1% | 4.8% | 8.6 | 5.4 | 11.8 | HIGHER |
| Chris Johnson | TEN | 8.4 | -9.0% | -13.0% | -5.7% | 8.5 | 5.2 | 11.8 | HIGHER |
| Rashard Mendenhall | ARI | 7.6 | -7.3% | -10.2% | -4.8% | 8.5 | 5.4 | 11.6 | HIGHER |
| Ben Tate | HOU | 6.5 | -1.8% | -6.7% | 2.3% | 8.4 | 5.2 | 11.7 | HIGHER |
| Montee Ball | DEN | 4.6 | 7.2% | 3.6% | 10.1% | 8.2 | 5.0 | 11.5 | HIGHER |
| Kendall Hunter | SF | 5.2 | 3.5% | -0.7% | 7.0% | 8.0 | 4.9 | 11.2 | HIGHER |
| BenJarvus Green-Ellis | CIN | 6.7 | -4.9% | -6.9% | -3.3% | 7.9 | 4.8 | 11.0 | HIGHER |
| Jacquizz Rodgers | ATL | 5.6 | -5.4% | -17.0% | 3.1% | 7.9 | 4.2 | 11.6 | HIGHER |
| Jason Snelling | ATL | 7.0 | -5.5% | -17.7% | 3.1% | 7.7 | 3.9 | 11.6 | HIGHER |
| Daniel Thomas | MIA | 5.5 | -6.0% | -9.3% | -3.1% | 7.3 | 4.3 | 10.3 | HIGHER |
| Bryce Brown | PHI | 3.1 | 5.6% | 4.4% | 6.7% | 7.3 | 4.2 | 10.3 | HIGHER |
| Andre Ellington | ARI | 4.9 | -8.8% | -12.4% | -5.7% | 7.1 | 4.0 | 10.1 | HIGHER |
| Pierre Thomas | NO | 5.4 | -5.6% | -7.9% | -3.7% | 6.9 | 3.8 | 10.0 | HIGHER |
| Zac Stacy | STL | 1.2 | 17.8% | 12.4% | 22.1% | 6.9 | 3.6 | 10.2 | HIGHER |
| Chris Polk | PHI | 2.7 | 6.1% | 4.6% | 7.4% | 6.7 | 3.8 | 9.6 | HIGHER |
| Benny Cunningham | STL | 0.9 | 18.8% | 13.0% | 23.4% | 6.5 | 3.3 | 9.8 | HIGHER |
| Christine Michael | SEA | 3.4 | -6.4% | -10.3% | -3.2% | 6.5 | 3.4 | 9.6 | HIGHER |
By now, you should be familiar with how to read the table, but here's an example just in case there are any newcomers out there. For Jamaal Charles, the "Points" column says he's projected to score 15.2 points against Tennessee, which is lower than his current average of 17.7 points per game. The system is 90 percent confident that he will score between 11.6 points and 18.7 points. As a proportion of his projection for the week, matchup-related factors (i.e., the spread and Tennessee's average fantasy points allowed to RBs) lower his projection by 0.8%, with a range of -1.7% to 0.0%. All in all, it generally looks like Charles' matchup this week isn't going to matter much for his scoring, but his current average is so high that regression to the mean says he'll probably score below it. (I know. I'm a broken record when it comes to that idea.)
Now for a few matchups I would like to highlight:
Eddie Lacy
Lacy returns from his concussion with a great matchup and great circumstances in the Packers' backfield. James Starks has been ruled out for Sunday's game, and, while Johnathan Franklin tore it up when Starks got hurt against Cincinnati, that's just the thing: He only got touches because the two guys ahead of him on the depth chart were unavailable. Also helping Lacy is that the Lions defense currently averages 22.3 points allowed per game to RBs. And, in terms of the spread, Green Bay being a 7-point favorite implies that they have an 80.7% chance of winning the game. That bodes well for a ton of clock-killing carries in the fourth quarter, and we all know Mike McCarthy goes super-conservative with a lead.
Am I worried that he might get re-concussed during the game? Yes. Is that something we can predict with any degree of certainty? Not really. For the purposes of this system, Lacy's listed as "probable," and I told you earlier that such RBs don't score fewer points than those who aren't on the injury list altogether.
LeGarrette Blount
As I mentioned earlier, Blount gets awarded a "promotion" tag this week. Historically, RBs who get promoted average about 6 points more than their average going into the game. That more than makes up for a slightly negative matchup at Cincinnati. Surprisingly, the stout Bengals defense is allowing 16.2 points per game to opposing RBs, which only ranks them a squarely-in-the-middle-of-the-pack 16th. With Stevan Ridley out, Blount figures to be the one who gets most of those points. At the very least, the world's worst-kept secret is that he will be getting all the carries, while Brandon Bolden gets all the receiving targets.
RAshad Jennings
Jennings is another beneficiary of a promotion. Darren McFadden is currently listed as "doubtful," but such players suit up approximately 0% of the time, and there's no reason for Oakland to rush him back when (a) he's a notorious slow healer, and (b) they can afford to hold him out until after their Week 7 bye. The problem for Jennings is that he has one of the worst matchups of the week, as the Raiders are 3-point underdogs against a Chargers defense that ranks sixth in fantasy points allowed per game to RBs (14.3). His saving grace, however, is that in addition to McFadden being out, Marcel Reece is "questionable." It should be very clear at this point what that means for Reece's output this week.
Reply Hazy Running back(s) of the week --everyone playing in st. louis
Hey, whaddyaknow! The Jaguars game is a total crapshoot from a matchup perspective! Thursday saw Sam Bradford being awarded Reply Hazy Quarterback of the Week, and the same explanation applies to the Rams' RBs. Of course, there's also the added complication of not having any clue how Brian Schottenheimer and Jeff Fisher will be allocating touches on Sunday. Daryl Richardson said he was demoted, then Fisher said he wasn't. There are rumblings that rookie Zac Stacy will see significant action for the first time in his young career. Isaiah Pead remains a talent St. Louis selected in the top 50 of an NFL draft, but is reluctant to use in any meaningful way on offense. And, of course, with all three RBs figuring to play some role this week, the fact that Jacksonville ranks 24th in fantasy points allowed to RBs (19.4 per game) is less important when that feast will be getting divvied up like fresh roadkill found by a pack of vultures.
For the opposition, Maurice Jones-Drew's downtrodden fantasy fortunes have been inextricably tied to the epic failure of his team as a whole. MJD used to be the RB who seemingly excelled in spite of his team, but at 28 years old and quickly approaching "broken down old man" status, those days are long gone. The good news -- if there is any -- is that the Jaguars don't have to face 25-point underdog territory until next week, so there's a chance they actually keep the game close this week. MJD and his fantasy owners would welcome it. But like I continue to say, an excess of uncertainty warrants an excess of caution. If you're a desperate 0-4 or are a huge underdog in your fantasy matchup this week, it's correct to play a high-variance strategy. If not, I would leave MJD (and the St. Louis RBs) on your bench if I were you.
