Predictive modeling is the rage these days. Daily fantasy sports have helped to spur a wave of data-driven models and curated algorithms.
Fantasy investors are more informed than ever, with rich resources available for every sport. One element that remains unpredictable, though, is injury. We want to know which players are at increased risk before we invest, but this has proved to be a fickle pursuit over time.
Enter Sports Injury Predictor, a platform fueled by a patent-pending algorithm that seeks to formulate predictive injury assessment based on myriad factors specific to a given player’s history and usage pattern.
In 2012, just a few years into his fantasy football career, Sports Injury Predictor’s Jake Davidow marveled at the market’s pricing of Darren McFadden. DMC was consistently going higher than Adrian Peterson, who was returning from an ACL injury.
In many ways, the market deemed Peterson a prohibitive risk and McFadden a potential difference-maker. We all know how Peterson’s 2012 went. And, of course, we know how disappointing shares of McFadden proved to be.
Davidow was flummoxed by the lack of available injury resources and the absence of a central database for what is a highly influential portion of fantasy investing. So, he worked with a friend with extensive risk analysis experience thanks to a background in insurance to assess which factors influence injury outcomes. Below you’ll find a brief breakdown of the model’s foundation:
Injury history: An account of every injury that has taken place to skill position players in the NFL and college for the last 10 years. Includes type of injury, games missed, surgery required and more. Our ever growing injury database goes back all the way to college for all the skill position players. This database goes back 14 years and has nearly 500 players injury history all stored in great detail.
Injury Correlation: Next we have our correlation matrix that weights the different injuries by investigating the relationship between them. By keeping a running count of which injuries lead to other injuries we are able to dynamically weight the chances of an injury reoccurring or causing another type of injury.
Biometrics: We also take into account biometric data such as age, weight and height as some physical profiles are more resistant to injury than others. For example, short, thick running backs tend to get hurt less than tall skinny ones.
Game data: Finally we use game data such as position and projected touches to act as a lever that defines the player’s exposure to risk. Players who touch the ball more often are more likely to get injured.
Using Sports Injury Predictor as a potential resource heading into the 2015 campaign, let’s look at some of the “riskier” commodities at each position as revealed by Sports Injury Predictor’s algorithm.