Why rush attempts matter, and receptions do not
Posted by Jason Lisk on September 22, 2009
This one is my theory on "Why rushes matter and receptions do not" in measuring running back workload and future injury risk.
I often see people complain that measuring workload by excluding receptions is inappropriate. I think this is up for debate, because it is not an apples to apples comparison. Rush attempts are far more likely to result in tackles (rather than runs out of bounds), tackles involving multiple tacklers, and tackles involving really big defenders.
But setting that aside, I think there is a far more significant reason why rush attempts matter, and receptions do not, when measuring workload. Take a look at these real games turned in by NFL running backs.
Running Back A: 29 rush attempts, 102 rushing yards, 0 receptions, 0 receiving yards, 2 touchdowns
Running Back B: 23 rush attempts, 117 rushing yards, 6 receptions, 52 receiving yards, 0 touchdowns
Now picture those games in your head, based only on the running back statistics. What type of backs are they? And more importantly, how did the game proceed?
These two games occurred almost exactly thirteen years apart. Running Back A is Ickey Woods in the AFC Championship game played on January 8, 1989 against the Buffalo Bills, in a game that the Bengals won to advance to the Super Bowl, 21-10. The Bengals led or were tied the whole game, and Woods' fourth quarter touchdown plunge sealed the victory. Running Back B is Priest Holmes in the final game of the 2001 season (played January 6, 2002) against the Seattle Seahawks. Holmes got some carries early, but the Chiefs fell behind 14-0 in the second quarter, and the Chiefs would play from behind the rest of the game, ultimately losing 21-18.
Which leads me back to why rush attempts matter in measuring workload, and receptions do not. It has to do with correlation with winning. High rush attempt games are highly correlated with teams that won the game. Why should that matter? Because teams that are winning tend to run the ball heavily late. The distribution of runs is not uniform throughout the game. The difference between a 20 carry back and a 28 carry back may be 8 carries, but it's probably not 2 carries every quarter. The 28 carry back probably has a quarterly distribution more like 7-5-6-10, while the 20 carry back is more uniform. The high workload rushing attempt guy, then, is getting a greater percentage of carries in the fourth quarter, when he may be tired. Those carries are more likely to come consecutively, when his team wants to run clock, and the other team knows they are going to run the ball. Running the ball is a risk every time, running the ball fatigued, though, is a greater risk to cause damage that will show up in the near future.
Let's confirm that my Ickey Woods/Priest Holmes example is not just an isolated one. Here, for example, are the winning percentages for all running back games since 1960, where a running back had between 28 and 30 touches (rushes + receptions), sorted by the rush attempt percentage for those touches:
%rushes total win pct ==================================== >90% 593 0.823 >80% 460 0.726 <=80% 128 0.539 ====================================
Or we could look at it another way, by holding the rush attempts constant but seeing how the number of receptions changes the winning percentage. Here are all backs who had exactly 25 rushes, sorted by number of receptions:
recept. total win pct ==================================== 0 to 1 228 0.803 2 to 4 218 0.766 5 plus 60 0.683 ====================================
So, we see that the percentage of rushing touches is an indicator of whether the back played in a win (and presumably, thus, with the lead late). Thus, a few more of those 25 carry backs who had a lot of receptions may have actually had a less harmful game, if they were not running the ball directly in to the teeth of the defense repeatedly in the fourth quarter.
The original idea for why attempts matter more than receptions was planted by this post by Chase Stuart a few years ago. Here's what Chase said then:
I don't think there's anything groundbreaking in the data, although it's nice to get some empirical evidence. If a team is giving its RB thirty-five or more carries in a game, chances are that: 1) that team is winning and/or going to win; and 2) it's a pretty close game. Thirty-three of the 59 games were decided by seven points or less, and only six involved 20-point victories. It makes sense that you'd only keep riding your star RB if it's a close game.
And the "keep riding your star RB" line is why I think that when we measure workload in an effort to study future injury risk, it's rush attempts that matter and receptions are extraneous information. But I'm a little slow on the draw, and it's taken me a while to sit down and formulate my thoughts. Now that we have this great game finder index and individual game data back to 1960, I figure it's time to dig back into the issue. After all, you don't want to hear any more of my musings, you probably want to see some hard data.
Well, if its getting a high workload while icing a lead that matters most in increasing injury risk, then we should see that backs who are the exception, and get a high workload rushing game while losing, stay healthier than their similar workload comrades on winning teams. And we can sort that out using the game finder. In this first case, I took all backs who had 25 or more carries in a playoff game since the 1977 post-season (hence, year N+1 would be 1978 or later). I sorted them by whether the back got those high carries in a game his team lost or won. The reason I settled on 25 before I began the research and actually looked at the data, by the way, was to be consistent with this series of posts from two years ago and this one from after the 2007 season.
81 different running back post seasons appear on this overall list. Some backs appeared on the list with more than one qualifying high carry game. If the back had a 25+ carry game in a loss as well as a win in the same postseason, he was placed in the "win" group. Backs in the loss group only had 25+ carries in games their team lost that post season.
I list the number of players in each group, average games played the next year (GP), percent of backs in that group who played every game the next season, and the percent of backs who missed at least half the games (M-8+), and those who missed at least 3/4 of the games the next season (M-12+). Here is a summary of the results:
Type No GP All M-8+ M-12+ WIN 67 12.3 0.27 0.18 0.13 LOSS 14 14.9 0.64 0.00 0.00
Okay, so far, it seems like the high carry games in wins might be more costly than in losses, but we only have 81 cases. Let's look at the final week of the regular season, same methodology.
Type No GP All M-8+ M-12+ WIN 108 12.9 0.43 0.15 0.08 LOSS 22 14.0 0.55 0.05 0.05
Similar story here, though not as pronounced. The one guy who missed over half the season after having a high carry game following a loss, by the way, was a dude named Scott Lockwood, who had 35 career carries with 30 coming in one game. A check of that game shows that the loss in question was in overtime and the Patriots led the entire game.
Let's pool that data, clean it up and eliminate duplicates (guys who appear on multiple lists). If a guy had 25+ carries in a loss in one group, and 25+ carries in a win in another, he was counted in the "win" group for pooled data. If a guy appears in both the week 16 and playoff groups with 25+ carries in a win, I removed the duplicate so he was only counted once (there were no duplicates on the loss lists). Here are the results:
Type No GP All M-8+ M-12+ WIN 159 12.7 0.37 0.16 0.09 LOSS 31 14.2 0.58 0.03 0.03
We are dealing with only 31 guys in the pooled Loss group, but they stayed healthier than their winning counterparts with similar rush totals, and the Loss group's early season injury rates are in line with previous research on running backs with moderate starting rush totals at the end of the previous season. The difference between the groups is mostly due to the really early serious (mostly knee) injuries from the high workload, winning game group. Going back to 1978, we find that several other notable injuries appear from this group. Curt Warner was the old school Jamal Lewis, as Warner and Lewis were by far the most heavily worked rookie running backs over the last month of the regular season and into the playoffs. These promising runners combined for 1 game and 10 carries in their second seasons. William Andrews was a star in the early 1980's for the Falcons. His career effectively ended with an offseason knee injury before the 1984 season, and he closed the 1983 season with a 29 carry game in a win. Former Charger Gary Anderson closed the 1988 season with 60 carries in consecutive wins to close the year, and missed the entire 1989 season with a knee injury.
I don't believe in curses or myths. And I don't necessarily believe in magic numbers, like 25 rush attempts or 28 rush attempts, where a back suddenly becomes more likely to get hurt, even though I've cited those as points where the injury rates increase. And I don't believe that every running back who has a high carry game gets hurt, far from it in fact, most go on just fine, its just that the number of "mosts" is less than other backs used more cautiously.
What I do believe is that running fatigued late in a game has the potential to cause injury, and backs who are playing for winning teams late in a game, and who have already played a full game, are more likely to be pushed over that line. Why do rush attempts matter and receptions do not? It's mostly not because of the different risk in injury occurring on the respective types of plays. It's because of what high rush attempts represent and typically say about how the back was used.
September 22nd, 2009 at 8:34 am
Whether you believe in the curse of 370 put forth by Football Outsiders or not (and I know that PFR respects the work at FO and vice-versa), this data reinforces their work. I think that what you are showing here, Jason, as well as other previous posts on RB over-usage, is that the injury rate for your RB dramatically increases if you are using him on consecutive carries, esp. in the 4th quarter. PFR has concentrated on individual games or several games in a row, whereas FO's research focuses on the entire season. I also think that NFL coaches recently have either a) come to the same conclusion based on their own research or b) studied both groups' research and realized that the conclusions were correct. I think this is why we see LOTS more multiple back systems (which, of course, frustrates fantasy owners, but is better for the team/players).
September 22nd, 2009 at 4:58 pm
Very interesting study! A couple questions.
Did you compute the average carry totals for the 25+ carry win group vs. the 25+ carry loss group? I would think that the win group backs would more often have 30+ carries.
Why did you limit this study to games in the post season and the last week of the season? Perhaps if you took the backs with any regular season games of 25+ carries, you'd find a relationship between whether their teams won and future games missed.
September 23rd, 2009 at 12:14 am
Interesting.
A little situational imagination aided by last night's game (Colts over Dolphins) makes me wonder how many of those losses were ones where a team with a lead was running the ball to milk the clock (i.e., same type of running as the backs on the winning team), only to see the defense cough up the game in the last minutes. With four minutes left last night, was Ronnie Brown running less hard and in a less punishing situation than a back who won the game?
September 23rd, 2009 at 4:41 am
Dave, how do we decide whether a defense coughs up the lead or the other team's offense comes up with a good scoring drive or play? Maybe the size of the lead that the defense is protecting could make a difference?? Just a thought.
September 23rd, 2009 at 5:05 am
Joseph, I respect the work of everyone who works in the area of football research, though we may not always agree on specifics. The Curse of 370 bothers me more because it does allow for some criticism about the methodology (not being consistent on years until injury, not differentiating between injury and "loss of effectiveness", selecting 370 instead of 360 for effect, and carving out an Eric Dickerson exception) by others, to the point that smart people don't accept it. Even though I think the main idea behind it will prove correct.
On the other hand, with small sample sizes, the data still shows that high season carry backs decline more than those right below them, even if the results aren't statistically significant. Why don't we have statistically significant sample sizes? After all, almost every team gives 350 carries or more to their backs in a season. Because teams rest their backs. I don't think this is a situation you demand a p value of .05 as an organization to make a decision on how to handle them.
September 23rd, 2009 at 5:11 am
Jim A, I didn't, but I can go back and do that. I suspect that the Winners will have a slightly higher average per game. However, the highest carry game in history (Jamie Morris) came in a loss in game 16.
I limited the study for now because of my belief that it is end of the year workload that is more important for the next season. If I used week 9 games, you might say, yeah, but how did they do over those last two months. Here, we go directly from what happened at the end of one, to measuring the next.
I do plan on digging deeper though, and have to some extent. It's not an easy program or quick search, but we can find every back who had a game with 26 carries and 2 receptions, and then could measure whether that back got hurt within 50 touches, 100 touches, etc. If we look at injuries within X touches across groups, it shouldn't matter as much when we measure from (week 3 or week 16).
September 23rd, 2009 at 5:13 am
Dave and Bob S., I agree with you completely. Some percentage of those losses were games the team was leading late. This is a preliminary view. As always, onward and upward with the research. I want to dig into the play by play in more recent seasons, also on the to do list
My guess though, is that very few of the 26 carry, 0 reception games were come from behind wins.
September 23rd, 2009 at 6:55 am
Interesting theory about end of season vs. in season workloads. I would have guessed that you'd see more of an effect in season because the offseason gives a high-workload back six months to rest and recover/rehab from any nagging overuse injuries. It seems strange to me that the biggest effect of workloads on future injuries would contain an artificial six month delay.
Just curious for anyone who knows. Do the increased injury rates for baseball pitchers who regularly exceed 100 pitches primarily occur in the following season also? Or is there also an in season effect?
September 23rd, 2009 at 7:46 am
Great work, JL. Very interesting!
September 23rd, 2009 at 8:57 am
You have no idea the things I scream at Brad Childress when the Vikings have a 10+ point lead late in the fourth quarter, and Childress decides it's a good time to let Adrian Peterson ram his body into waiting defensive fronts, when he has Chester Taylor standing on the sideline next to him. Very angry things.
September 23rd, 2009 at 8:59 am
You stated that "Running the ball is a risk every time, running the ball fatigued, though, is a greater risk to cause damage." Is there statistical evidence that more injuries occur in the fourth quarter than the first three quarters? I tried to google any studies, but couldn't find any.
September 23rd, 2009 at 9:30 am
PV-
I actually don't believe in running back platoons. I think star running backs can handle 20 carries on a regular basis. I do, however, think teams should platoon late with a lead. In baseball, you have to pull your starter to bring in the closer. You don't have to do that in football. But I think teams should have "closers". Typically, more plodding, sure handed veterans who can take 50% of the carries when leading or running the ball heavily late. So yes, I agree with you, Taylor should get carries late. The difference between Taylor and Peterson is less pronounced in that situation, I trust Taylor to be almost as good at picking up 3 yards and keeping the clock running. Leverage Peterson early, when the outcome is in the balance, preserve him late by using a closer platoon.
September 23rd, 2009 at 9:37 am
Steve Schwarz, I should have been more exact. I don't know if there are any studies, and if I were to guess, there aren't more injuries late.
What I mean by my statement is that it is a greater risk to cause damage, which will show up in the near future. (i.e., in the form of a visible injury within a game or two) Not that the back will get hurt on the 27th carry. Think of it this way, if I am running fatigued, I may get less bloodflow to the lower extremities. My ligaments, and the muscles they are connecting, may not get the oxygen and are weakened. Does that mean the ligament will snap immediately? No. It may take 10 sharp cuts on it, or 20, and that injury may randomly appear to occur on the 11th carry of a game a week later.
September 23rd, 2009 at 2:12 pm
Reading this conversation gave me a couple of thoughts, too. First, the games that you looked at, it's highly likely that the player getting many carries late in a game did the same thing throughout the season. Historically, coaches have run the ball with a lead, and if they have a RB that has proven reliable, that's who they stick with. The second thought is that I think it's a great point to have a player who can move the pile for late in the game who may not be your primary back later in the game. LaRon McLain was pretty good in that role for the Ravens last year, and Bettis was awesome in that role for the Steelers his last 3 years. I thought that the Bus was a huge luxury on a team, because although his body couldn't take 16 games of 20-25 carries anymore, he could come in late in the game and eat the clock. And when the RBs ahead of him were injured he could start a few games and still carry the load.
September 23rd, 2009 at 2:57 pm
Jim A (comment #8), sorry, just now seeing that one.
Pretty much every early season serious injuries that showed up on that list is a torn knee ligaments. Ligaments don't heal on their own. If the damage was caused in January, the injury may not occur until the back begins cutting laterally alot. He's probably not doing that in March. But 1 week, a month, or 6 months later, the damage is still there.
But let me be clear. I'm not saying that it takes 6 months or a season for damage to occur. I'm saying that the injury occurs soon after the high carry game. In my earlier study from two years ago, the backs with the most carries through the first 6 weeks got hurt at a higher rate in games 7-10. It just so happens, that "soon after," for games in the postseason and game 16, isn't until the next pre-season and start of the regular season.
September 23rd, 2009 at 7:47 pm
Not a fan of the 370 rush "curse", not when FO conveniently decided upon that particular number. By doing so they ignored Curtis Martin's seasons of 368, 369, and 367 carries. There were a few other backs who didn't fall apart after seasons in which they carried the ball 360 something times.
Furthermore, I don't agree with FO as to their statement that only Dickerson survived a 370+ campaign. Unfortunately, with an infant son I don't have the time right now to get into the arguments again.
September 23rd, 2009 at 7:53 pm
http://www.pro-football-reference.com/play-index/psl_finder.cgi?request=1&match=single&year_min=1920&year_max=2009&season_start=1&season_end=-1&age_min=0&age_max=99&league_id=&team_id=&is_active=&is_hof=&pos_is_qb=Y&pos_is_rb=Y&pos_is_wr=Y&pos_is_te=Y&pos_is_rec=Y&pos_is_t=Y&pos_is_g=Y&pos_is_c=Y&pos_is_ol=Y&pos_is_dt=Y&pos_is_de=Y&pos_is_dl=Y&pos_is_ilb=Y&pos_is_olb=Y&pos_is_lb=Y&pos_is_cb=Y&pos_is_s=Y&pos_is_db=Y&pos_is_k=Y&pos_is_p=Y&c1stat=rush_att&c1comp=gt&c1val=360&c2stat=rush_att&c2comp=lt&c2val=369&c3stat=&c3comp=gt&c3val=&c4stat=&c4comp=gt&c4val=&order_by=rush_att
So Terrell Davis had 369 carries in 1997. He was then spectacular in 1998.
I see the names Walter Payton and Emmitt Smith there. God forbid they had a few more carries in those particular seasons.
Adrian Peterson had 363 carries in 2008. Good thing he didn't have another 7 carries over the course of the season. This saved him. Heck, if he had 370 last year, he'd probably be injured right now.
September 25th, 2009 at 10:56 am
No offense, but this says nothing about whether or not receptions take their toll on workload. To do that you would have to compare running backs with 25+ carries to other backs with 25+ carries and receptions, then compare their injury rates.
I also think people overstate the case that receptions count less than carries. Quite often players get tackled as they go out of bounds, receivers get hit on drops and tipped passes, carries go out of bounds (think sweeps and long rushes), and some devastating hits can come on wheel routes. Think Reggie Bush on his hands and knees after getting flattened in the NFC Championship game a few years ago.
January 13th, 2010 at 11:25 am
This is what I think. Everyone talks about defense. The whole evolves around the offensive and defensive lines. In the pros, clock management determines games. The point scoring line is the offensive line. You have 4 15 minute quarters to win in. What am I getting at? The bigger, faster, and more athletic, defensive lines get, the bigger and more effective offensive lines must be. Winning is about scoring points. There are statistics for everything,but, scoring points is the only one that is important. The key is to play a balanced game and to control the ball. That will manage the clock.