If you used us last year there is a very good chance we were one of the keys to you winning your league. We have more algorithm updates lined up for this year to give you an even greater edge but more on that later.
Here is a rundown of how we did in 2014.
How Injury Prediction works
Before we start here is some context for those of you who are not familiar with how we work. We track the injuries of every skill position player in the NFL. We have a database of roughly 500 players that grows every year with all the players who enter the NFL via the draft and free agency. We have injury data on all of these players going back to college. We layer other factors on top of a player’s injury history such as biometric data, what position they play, projected touches to name a few.
That information gives us a probability of a player being injured in the coming season as a percentage.
Now everyone who is going to play fantasy football semi-seriously takes a moment to wonder whether certain players they are interested in this coming season are likely to be injured. The problem with doing it that way is that if your assumptions of what causes injury are wrong, then your outcomes will be wrong too. We have tested hundreds of inputs using regression models and have worked out the main signals from the noise giving way more accurate outcomes allowing for solid decision making at draft time and when evaluating trades.
Being able to quantify injury probability is so important is because it gives you the ability to capitalize on inefficiencies in the pricing of players in the draft and subsequent trades throughout a season. These inefficiencies are created by the masses guessing what the likelihood of a player getting injured are using assumptions such as age, previous workload and how many games the player has missed in previous seasons. If any of those three factors are your departure point in trying to determine risk you will be doing a Ricky Bobby prayer every Sunday before kickoff:
We have proven without a doubt that the factors mentioned above (age, workload and previous games missed) do NOT have any statistical significance in working out which players are going to get injured (there will be more posts on this to come soon).
In 2014 you won at Fantasy Football when you avoided over paying for risky players (Calvin Johnson, Andre Ellington) and found value in players the market faded due to concerns with their injury risk (Marshawn Lynch, Jeremy Maclin) based on incorrect injury probability assumptions.
We were very accurate in 2014
We took all the players in our database and group them into percentile buckets the breakdown of which is in the table below. In order for our math to be viewed as a success it should predict the correct outcome on both ends of the spectrum. Those who are less likely to get injured need to not be injured and those more likely to be injured should have got injured.
The outcome of last season’s predictions was that we were correct on both ends of the spectrum. For players who were in the top percentile (70 – 99.9%) 75% of them got injured and missed more than 3 games on average. For those in the bottom percentile (0 – 30%) only 31% were injured and those who were injured missed roughly 1 game on average.
The correlation between high probability of being injured and more games missed makes sense. If a player has enough signals to flag as a high risk the chances are the injury is likely to be more severe.
As you can see using the injury Predictor model is way better than guessing or resorting to Ricky Bobby Sunday morning prayer sessions.
Some shameless name-dropping
We had some high profile wins last year. A quick snapshot looks like this:
Ellington was our player for 2014 most likely to get injured. He hurt a muscle in his ankle in the preseason and was unable to recover. Unfortunately he was still drafted as RB11 (Yahoo ADP) but finished outside the top 20 (standard leagues).
Money Lynch was drafted as RB7 due to injury concerns based on his age but produced as an RB3. Romo’s back injury (herniated disc) while very painful is not something that keeps players out of games. We have many instances of this injury to model off of and so we were really confident that he would not be at risk of injury and missed games in 2014.
The only one in this snapshot who is a tease is Arian Foster as regardless of where you drafted him he worked out to be a value pick. He did miss 3 games through injury so in a zero sum game we got it right – he just turned out to be insanely good enough for that not to matter.
Julio Jones tried to make us look bad
Another name worth mentioning is Julio Jones. We had Jones as one of the most likely players to get injured in 2014. This was as a result of the re-fracture on his foot the year before along with the hamstring and ankle injuries he has suffered in the past (his injury profile can be found here). Jones then went on to make us look bad by dropping 1,593 yards on us for the season.
However – on closer inspection we see that 38% of his production came in a 3 week period of which one of those weeks was in the first fantasy playoff week. If you owned Julio last year he most likely did not lead you to a title. The reason for his drop in production over a 6 week period was a foot injury.
This is exactly the kind of volatility you expect to see from a top tier player but one who battles with injury. If you drafted Jones you had just as explosive but safer options available at similar ADPs such as Dez Bryant, Antonio Brown, Demaryius Thomas and Jordy Nelson.
Injury Prediction is here to stay as the dominant method of determining risk in the coming season. You win at fantasy football when you can avoid over valued players and find the under valued players. People trying to guess which players are more likely to get injured create inefficiencies in market price that can be exploited by those in the know (if you weren’t before you are now one of those people).
We will be making some adjustments to the algorithm this year that will help improve our results even further. Unfortunately we can’t give too much away but you should notice an even higher increase in accurate outcomes on both ends of the injury spectrum.
To help you crush it this year we have some fantastic new tools lined up that we’d love you to try out as you begin your research for 2015.
Player Search – Use this tool to find in depth injury history and probability for each player you are researching
Compare players – This is a great tool if you’re looking to compare players that interest you to assess their risk
Search by position – This is a quick reference tool that will allow you to discover which of the top 10 players have the most and least risk associated with them
The one change this year is that there will be a fee of $4.95 for the season to make use of these tools and our data. Basically we’re going to help you win your league for less than it costs you to get your favorite 6 inch from Subway. Value!
We’ll be publishing content on the site but you can also find us on Rotoworld.com and Draftsharks.com where we will be providing injury content leading up to and throughout the season.
If you haven’t please sign up for our emails and as always follow us on Twitter @injurypredictor