December 12, 2016

    Understanding Predictive Models: There Are Never Any Guarantees

    When people start discussing predictive analytics, conversations can quickly get vague and cryptic. Anyone can make outrageous claims out of context based on limited information, and back it up with a series of references and big words. Understanding what predictive actually means will allow you to put these claims into context and understand how or if they apply to your athletes.

    Insurance anyone? 

    Predictive analytics and predictive models suggest an increase odds of an outcome based on a change. Instead of diving into statistical “jargon,” we will use an example from a game of chance we are all more familiar with, Blackjack. When playing the game of Blackjack, the probability of getting “Blackjack” or 21 is 4.78% or about 20:1.  But, if the first card dealt is an Ace, the probability increases to 31.07% or closer to 2:1. To a coach, understanding the odds of hitting 21 increase significantly depending on the first card drawn is relatively easy to understand.

    Having an Ace does not guarantee hitting a Blackjack, nor does not having an Ace guarantee not hitting a Blackjack. However having an Ace does significantly increase these odds.

    “The #1 Predictor of Future Injury is Previous Injury” (1,2,3,4,5,6)

    What does that mean? If you’ve been hurt before it is guaranteed you are going to get hurt again? No. If you have never been hurt before does that mean you are never going to get hurt? No. This is a simple concept that has been misrepresented and misconstrued. Does a high Training Stress Balance guarantee injury? No. Does a 21 on the FMS guarantee an injury free athlete?

    Sparta’s Predictive Data

    We have seen an increase in an athlete’s DRIVE (Average Relative Concentric Impulse) as a predictor of elbow injuries in baseball athletes. If we look at the statistical analysis specifically “The odds of having a Elbow injury vs no injury will increase by 3.02 x for a one-unit increase in Drive. In other words, every time the Drive variable increases by one T-score, the odds of having an elbow injury vs. not having an injury goes up by 3x.

    We have also found a positive correlation between increased LOAD and DRIVE and an increase in offensive baseball statistics such as batting average and home runs. As well as positive correlation between a high EXPLODE and low LOAD variable, and increased minutes played in Division I Men’s Basketball athletes.

    These are just a couple of examples we have found looking at how our Sparta Signature applies to injuries, performance, and athlete resilience. Does this mean we can guarantee an elbow injury or more home runs? Unfortunately for all of us, it does not.

    Be a skeptic, but search for meaning

    There is a saying in statistics, “All statistical models are wrong, but some are meaningful.” We will never be able to make guarantees, no one will.  And we should all be wary of anyone who claims to be able to. If it sounds too good to be true, it usually is.


    Just because we cannot make guarantees does not mean these predictive models do not have value. Knowing that previous injury is the number one predictor of future injury should help you and your organization identify at risk athletes, and hopefully take some sort of action.  We don’t use our predictive data to make guarantees, we use this data to make decisions and drive interventions.


    1. McKay, Graylene Dawn, et al. “Ankle injuries in basketball: injury rate and risk factors.” British Journal of Sports Medicine 35.2 (2001):> 2. Hägglund, Martin, Markus Waldén, and Jan Ekstrand. “Previous injury as a risk factor for injury in elite football: a prospective study over two consecutive seasons.” British journal of sports medicine 40.9 (2006): 767-772.

    3. Dvorak, Jiri, et al. “Risk factor analysis for injuries in football players possibilities for a prevention program.” The American Journal of Sports Medicine 28.suppl 5 (2000): S-69.

    4. Arnason, Arni, et al. “Risk factors for injuries in football.” The American journal of sports medicine 32.1 suppl (2004): 5S-16S. 5. Engebretsen, Anders Hauge, et al. “Intrinsic Risk Factors for Groin Injuries Among Male Soccer Players A Prospective Cohort Study.” The American Journal of Sports Medicine 38.10 (2010): 2051-2057.

    6. Chomiak, Jiri, et al. “Severe injuries in football players influencing factors.” The American journal of sports medicine 28.suppl 5 (2000): S-58.

    Tag(s): Data , Education

    Other posts you might be interested in:

    View All Posts