How Injury Affects Weekly NFL Player Performance in Fantasy
You know the drill: Adam Schefter tweets out a confirmation that a player who is on the injury report is active and will play. Within minutes, a bevy of fantasy experts retweet the news, adding something like "deploy him as you normally would in your fantasy lineups".
But should you really deploy him as you normally would?
We’re officially in the age of data, and fantasy football is of course a data-driven game. But one area where we still haven’t fully utilized the power of data to make decisions is injuries.
Those pesky injury tags pop up every week, but what exactly do they mean from a statistical standpoint?
How much of a percentage dip in production should we expect when a player plays hurt?
And what is the difference between playing with a questionable (or doubtful) tag versus playing with a probable tag?
Let's find out.
The NFL Eliminated Probable Tags for 2016 -- What it Means
Before we jump into the study, it's important to discuss the ramifications of the league eliminating probable tags. As of 2016, there will only be questionable, doubtful, and out.
First, let's review what the injury tags used to mean:
- Probable: 75 percent chance of playing
- Questionable: 50 percent chance of playing
- Doubtful: 25 percent chance of playing
With the new rules in 2016, here is what the injury tags now mean:
- Questionable: "Uncertain" to play
- Doubtful: "Unlikely" to play
The NFL's reasoning was that players listed as probable play at a much higher rate than 75 percent of the time, and so eliminating the probable tag will provide more clarity as to which players are truly in danger of missing the game. The odds that rule change will have the intended effect are shaky at best. The rule change could backfire -- as the new touchback rule looks like it might be doing -- and NFL teams could simply lump in players they would have formerly listed as probable as questionable.
How does this affect us as fantasy owners?
There are two scenarios that could play out, but before I get to that, I want to give a preview of the results of the study. I found that for every position besides quarterback, players saw a decrease in production both when probable and when questionable or doubtful, with running backs and wide receivers seeing a greater decrease when questionable or doubtful than when probable. With that being said, here are the two scenarios:
Scenario 1: Teams could remove players they would have formerly listed as probable from the injury report.
While it may seem counterintuitive at first, this is actually the less ideal scenario for fantasy owners. Remember, while players listed as probable almost always play, they suffer a decrease in production compared to playing when absent from the injury report entirely. If these players disappear from the injury report, we have no way of spotting these players who are likely to suffer a production drop-off.
Scenario 2: Teams could start listing players they formerly would have listed as probable as questionable.
Although this may muddle things in terms of who is actually going to play, it would still be the more ideal scenario for fantasy owners because at least we would know all the players that will possibly play hurt and adjust their projected performance accordingly. I broke down the injury performance data in probable and questionable/doubtful, but if this scenario played out, a projection for a player listed as questionable could simply be adjusted by the average dip in production for all injury tags. It wouldn't be quite as exact as being able to apply different percentage decreases depending on probable or questionable, but it would be a whole lot better than the former scenario, where a bunch of players that we won't be able to identify beforehand (because they're not quite banged up enough to be listed as questionable) will likely see dips in production anyway.
It is anyone's guess as to how this plays out, so the best course of action will probably be to apply the percentage decrease from all injury tags collectively when a player is listed as questionable, rather than assuming that all probable players will simply disappear from the injury report.
Update: 4for4's John Paulsen correctly points outs that I'm looking at the ideal scenarios more so from a DFS perspective. From a season-long perspective, obviously having a bunch of players listed as questionable is a headache, because you may be unsure if they are going to play by the time you have to make lineup decisions. Paulsen also provides a possible solution to differentiating between the "true" questionable players versus the questionable players that would have previously been listed as probable: If a player is listed as questionable but is practicing in full by the end of the week, he would have in all likelihood have been considered probable under the old system.
I looked at all player seasons from 2008 to 2015 where the player:
- Played at least one game during a week in which he appeared on the injury report.
- Played at least three games not on the injury report.
Player seasons also had to satisfy at least one of the following criteria:
- Player had at least 240 pass attempts.
- Player had at least 100 rushing attempts.
- Player had at least 30 receptions.
For players who met those criteria, I used each game where a player appeared on the injury report as an individual data point, calculating the percent difference between the player's fantasy points (PPR scoring) in that game and the player's average of fantasy points in all healthy games during that season.1
First, A Visualization
Before we dig into the data, I'm going to provide a visualization to more clearly show how I arrived at the numbers.
Let's take C.J. Anderson in 2015. On one hand, Anderson didn't appear on the injury report eight times during the 2015 season: Weeks 1, 4, 5, 6, 7, 11, 12, and 13. In those eight games, he averaged 12.24 PPR points. On the other hand, Anderson made seven appearances on the injury report: Weeks 2, 3, 9, 10, 15, 16, and 17. For each of the games he played on the injury report, I'd calculate the difference between how many PPR points he scored in that game and 12.24, his average in healthy games, as seen below:
|Week||Injury Tag||Body Part||PPR Pts||Avg. When Healthy||+/- When Healthy||% +/-|
Anderson was actually an extreme case. He essentially produced only half as many fantasy points per game when playing hurt as he did when he was healthy.
How Each Position Performs When Appearing on the Injury Report
Now that you have a better understanding of the methodology, let's begin to dig into the data.
Note that I treated questionable and doubtful designations the same due to the relative rarity of instances where a player plays with a doubtful tag. My reasoning is that if a player is listed as doubtful and played, he should have really been listed questionable in the first place. If the doubtful player doesn't play, he is of no concern to us in this study.
A summary of the data is below. It shows the percent difference in PPR points between games played healthy and games played in weeks when a player was on the injury report, broken down by position. I left the sample size out of this table to remove clutter and so that it is easy to refer to and share. However, the sample sizes are included in the 'total' section of the positional breakdowns that follow. Remember, now that the NFL has eliminated probable tags, the first column is the most relevant column since it takes an average of playing through all injury tags regardless of designation.
- Quarterback is the only position that doesn't see a drop-off due to injury.
- The overall effects of playing with an injury tag on running backs, wide receivers, and tight ends are remarkably consistent, hovering between 8.5 and 9.9 percent. In other words, you can expect an 8-10 percent drop-off in production on average when a running back, wide receiver or tight end appears on the injury report.
- As I mentioned, questionable designations result in a more severe drop-off in production than probable tags for running backs and wide receivers, but not for tight ends. This is where the removal of probable tags hurts most -- it will likely result in us either overestimating or underestimating the decrease in production of running backs, and especially wide receivers, because we won't have a clear picture on the severity of their injury.
The Effect of Injuries to Each Body Part
Now let's dig into the injuries broken down by each body part for each position. As you peruse the data, keep in mind that the sample sizes for certain injuries are very small. I'll focus takeaways only on injuries where there is some semblance of a significant sample size.
|QB Injury||# of Games||+/- Fan Pts/Game||# Games Probable||+/- Fan Pts/Game||# Games Questionable/Doubtful||% +/- Fan Pts/Game|
- Even when a quarterback plays through an injury that may sound like a cause for alarm because it is directly involved with the throwing motion, such as an injury to the shoulder, arm, or hand, there shouldn't be a cause for concern. Why is this? It's likely that if a quarterback is too hurt to throw properly, he simply isn't allowed to play.
- Patriots coach Bill Belichick is arguably one of the greatest trolls of our time when it comes to the injury report, routinely listing Tom Brady on the report even though there's no chance Brady will sit out. So what happens when you remove Brady, who accounted for 5.6 percent of the total sample? Instead of a modest increase, there is essentially no change in production between games played on or off the injury report. It doesn't change the overall takeaway, but it does explain why the results show a collective increase -- Brady was playing 32.5 percent better when listed on the injury report! Belichick is an evil genius.
|RB Injury||# Games||+/- PPR Pts/Game||# Games Probable||+/- PPR Pts/Game||# Questionable/Doubtful||% +/- PPR Pts/Game|
- As we saw in the Anderson example above, ankle injuries are very detrimental to a running back's fantasy production, causing an 11-percent drop-off, on average. Running backs playing through ankle injuries (regardless of designation) saw a 9.3-percent decrease in rushing attempts and a 4.0-percent decrease in yards per carry.
- Finger and hand injuries both cause a double-digit percentage dip in production. This may be because coaches show less trust in ball carriers with hand injuries: running backs with finger injuries received 10.1 percent fewer carries than when healthy, and those with hand injuries received 6.3 percent fewer carries than when healthy (both regardless of designation).
- As you'll see in the case of wide receivers, hamstring injuries seem to only be a cause for concern when a player is listed questionable, not probable. Running backs with hamstring injuries saw no decline when playing with a probable tag, but 12.9-percent dip when playing with a questionable tag. Decreased volume was the culprit; yards per carry were actually a bit higher when playing hamstrung with a questionable/doubtful tag, but rushing attempt volume decreased by 13.8 percent. This is one situation when the aforementioned ideal scenarios are reversed: we'd rather running backs with "probable (hamstring)" designations be removed entirely from the injury report. Since we have no way of knowing if that will be the case, it is probably best to err on the side of caution rather than assume the hamstring injury is of no concern. This could take running backs with hamstring off the table in DFS cash games unless they are good enough values to withstand a near 13-percent dip in expected production.
- Knee injuries predictably lead to an 11.9-percent dip in running back production overall, and a more significant when the tag is questionable/doubtful (-16.6%) than probable (-11.9%), although the removal of probable tags doesn't hurt as much in this case because the decrease in production is significant either way. Running backs with knee injuries suffered an 8.2 percent dip in rushing attempt volume and a 4.1-percent dip in yards per carry average.
- Foot and toe injuries are also not surprisingly problematic for running backs, leading to 13.2-percent and 19.3-percent drops in production, respectively. There was no drop-off in workload when running backs played through a foot injury (regardless of designation), but the 17.3-percent drop-off in yards per carry was one of the most severe among all injuries. Toe injuries (regardless of designation) saw a drop-off in both attempt volume (-15.6%) and yards per carry (-13.5%).
|WR Injury||# of Games||+/- PPR Pts/Game||# Games Probable||+/- PPR Pts/Game||# Games Questionable/Doubtful||% +/- PPR Pts/Game|
- Ankle injuries cause an 11.8-percent drop in production for wide receivers, affecting both volume and efficiency. Regardless of designation, there was an 8.4-percent drop in targets, and 9.2-percent drop in catch rate, and a 6.2-percent drop in yards per reception.
- Foot injuries are known to give wide receivers trouble, and the data backs that up. Wide receivers saw a 19.3-percent dip in production when playing through a foot injury, including an astounding 25.3-percent dip in 44 instances when questionable. Wide receivers playing through foot injuries (regardless of designation) suffered a 9.3-percent dip in targets, a 9.1-percent dip in catch rate, and a 4.7-percent dip in yards per reception.
- Hamstring injuries are very common, and it appears the severity of the injury matters, so this is another area where the removal of probable tags hurts us from a forecasting perspective. Wide receivers playing through a hamstring injury with a probable tag saw a modest 3.0-percent dip in production, but a more significant 13.8-percent dip with a questionable/doubtful tag. With a questionable/doubtful hamstring injury tag, wide receivers saw a 10.6-percent dip in targets and a 10.9-percent drop in catch rate. This is another situation where it would probably be best to err on the side of caution in DFS cash games.
- Knee injuries were another injury where severity came into play. Wide receivers saw no drop-off when probable, but an 11.5-percent drop-off when questionable/doubtful. Knee injuries with a questionable/doubtful tag seemed to affect a receiver's ability to get open more than anything else: wide receivers suffered a 12.5-percent drop in targets, but only a 3.1-percent drop in catch rate and no change in yards per reception. Once more, I would err on the side of caution in DFS cash games.
- Concussions aren't something you'd think would have an effect on production, but receivers have struggled coming off a concussion with a questionable/doubtful tag, with targets dropping by a massive 23.6 percent. There were only 18 instances, so the sample is small, but it's possible that quarterbacks are electing not to throw the ball to wide receivers in situations where they'd take a big hit when they're coming off a more severe concussion. Concussion data for running backs and tight ends doesn't show a similar drop-off, so this could just be noise.
- Illness is another injury you wouldn't think would have a major impact on production, but in 31 instances when a wide receiver was questionable/doubtful with an illness, a 16.2-percent average drop-off in production ensued. This drop-off was all in the target department -- targets decreased by 11.1 percent, while there was no decline in catch rate or yards per reception. The sample was split nearly 50/50 in terms of probable and questionable, so this will be yet another frustrating situation to forecast now that probable tags have been removed.
|TE Injury||# Games||+/- PPR Pts/Game||# Games Probable||+/- PPR Pts/Game||# Games Questionable/Doubtful||% +/- PPR Pts/Game|
- As with the other positions that have to consistently run to accrue fantasy production, ankle injuries hinder tight ends, causing a 11.4-percent drop-off in production. The sample is relatively small (23 instances), but the 27.3-percent drop-off with a questionable/doubtful tag is noteworthy nonetheless, especially since there is essentially no drop-off when probable. Since the split between probable and questionable is so huge and we'll now have no way of telling if a questionable player is truly questionable, I'd exercise extreme caution with tight ends with ankle injuries in DFS cash games, probably avoiding them altogether. Most tight ends are merely complementary parts of their team's passing game, and when a tight end has an ankle injury (regardless of designation), his targets dip by 14.3 percent.
- At -15.4 percent, knee injuries lead to an even larger dip in tight end production than ankle injuries. Again, a tight end playing with a knee injury (regardless of designation) saw his targets drop, in this case by 11.4 percent on average.
How to Use This Information
The bottom line is, for non-quarterback skill positions, players who are playing through an injury see a tangible decrease in production from when they're healthy. I'll leave it up to you to draw a line in the sand as far as what sample size is sufficient for each injured body part, but with a sample size of 997 games for running backs, 1,639 for wide receivers, and 478 for tight ends, it's clear that the overall percentage drops in production for each position are applicable.
I'll end by briefly outlining some ways to put this information into action.
How to treat players who will play through an injury in weekly projections has long been unclear, but now we have percentages to apply to players at each position, which can also be altered depending on injury type if you feel the sample is adequate. Keep in mind this works both ways: if a player has been healthy all year and is now playing injured, you'd apply the appropriate percentage drop-off. But if a player has been playing hurt all year and is now healthy, you'd actually bump him up.
Referring once more to the C.J. Anderson example I gave near the beginning of the article, it's also clear that the data from this study has implications as far as preseason projections. A player who played hurt for a large portion of the season and suffered a huge drop-off like Anderson did in 2015 will likely see some positive regression (otherwise known as injury luck), so it would be unwise to simply use the previous season as a baseline without taking injuries and performance while injured into account, and vice-versa. Of course, we'd still need more information to do this properly -- namely how many times on average players at each position appear on the injury report over the course of a given season
Because we know injured players do suffer a tangible drop-off in production when injured, you can use that information to tilt the scales in a start-sit decision. What's more, if you already know that a player will suffer a drop-off in production even if he does play, you may in some cases be able to save yourself the agonizing wait to see if that player is active, instead going in another direction from the jump.
This information has huge DFS implications because you're trying to get the maximum amount of points out of every salary cap dollar, and in many instances, you're trying to maximize projected points. Even a drop-off in production of a few percentage points could alter a player's value, and thus alter the entire optimal lineup for a given week. Obviously, this is more of an issue in cash games. In tournaments, you could leverage the information the opposite way, making contrarian plays where other may shy away from a player due to an injury, especially one that isn't very severe.
- Running backs, wide receivers, and tight ends experience an 8-10 percent drop in production when playing in games when they are on the injury report (regardless of designation).
- Quarterbacks' production is not affected when they appear on the injury report.
- Running backs and wide receivers have historically seen a larger drop-off (12-15 percent) when questionable than probable (6-7 percent).
- Probable tags are now removed, which could lead to either 1) players that would have formerly been listed as probable no longer being included on the injury report, or 2) players that would have formerly listed as probable being listed as questionable.
- No. 2 is the ideal scenerio or fantasy, because No. 1 would result in certain players who are playing slightly hurt being unidentifiable. At least if probable players are being listed as questionable, we can forecast some drop in production, even if we don't know the exact scale.
1. To be clear, I'm taking an average of all healthy games throughout the course of a season and comparing them to games played when on the injury report, whether or not those injured games came before or after those games included in the "healthy" baseline. I'm essentially attempting to maximize the sample size included in the "healthy" baseline. I thought this method fit with the purpose of the study more than simply comparing a player's performance when injured to healthy games only up to that point, which would drastically reduce the sample size and skew the "healthy", or "true" performance baseline. There's obviously some give and take involved with regards to sample size, as I set the minimum for healthy games played to three, which may seem low. My reasoning was that I didn't want to not include the rare but important instances when a player played a large chunk of the season hurt. It should be noted, however, that there were rarely any instances where players played less than five "healthy" games.
Photo by John Grieshop/Getty Images.