Professional Sports & Big Data – Gaining the Competitive Edge via Analytics and Understanding

With the “Big Game” officially under wraps and a congrats due to the Baltimore Ravens for pulling out what was a hard fought NFL championship, we turn our attention to topics we sincerely enjoy discussing on the TopCoder Blog – Big Data, algorithms, and analytics.

Today’s post features a candid interview with Jose Fernandez, a specialized strength and conditioning coach and performance consultant who has worked with pro basketball teams, English Premiere League (soccer) clubs, and more. When you see headlines such as Buffalo Bills to Start Advanced Analytics Department and you pair it with the knowledge that sensor technologies are rapidly finding their way into the apparel these high-performance athletes are donning, you realize that pro-sports is currently undergoing a reformation of sorts. This road has the potential to lead to incredible strides in injury prevention, performance optimization, advanced (data-driven) scouting and beyond. It all starts with understanding, and with that said, we’d like to thank Jose Fernandez for sharing the below insights with us on the TopCoder Blog. 

Jose, many teams are seeking out professionals to help teams win with data. Daryl Morey from the Houston Rockets stated that the competitive advantage comes from having the right forms of data that are insightful. What are your thoughts on what teams are missing with data in general? 

Generalizing on what teams are missing is difficult but when I see organizations embracing innovation either by using the latest technology or applying new forms of data, I question whether the basics are being done first. In the NBA, it is not uncommon for players to go on the court after having a burger with chips and that does not show on the SportVU Stats. Every team in the Premier League knows everything about distance, speeds and acceleration metrics of each player for any given training session but how many can provide accurate details of the last 20 gym workouts?

I am a fan of innovation and love working with data. We are currently immersed in very exciting times but before rushing to add the new trendy device to our medical department the focus should be on the current dynamics and working protocols, ensuring they are consistent and adaptable. Only in this way one can determine what information is insightful and what is not or what new data set will provide a performance advantage.

Mobile technologies are very disruptive with sports science as they allow research to be pushed into the training room and sport fields. What are your thoughts about mobile tools for coaches and medical staff that can get better information more quickly? How are the cloud and platform use changing how teams collaborate?

The adoption of mobile technologies has been one of the most important changes in sport science over the last two years. Every athlete has now a smartphone or tablet that opens the door to a new dimension of data as well as new possibilities in the way it is being collected. This is undoubtedly impacting how teams are setting up monitoring strategies. Having more information available at a glance also increases the chances of measuring unnecessary data but regardless, the concept of mobile monitoring is perfect, at least a priority.

Realistically, the truth is that we are still at a very early stage and although mobile technology is slowly evolving it still has limitations and some teams are turning to their own in-house mobile solutions to ensure efficacy and compliance of their athlete monitoring strategy. The most straightforward example that comes to my mind is in the field of mHRV (mobile Heart Rate Variability). Other examples can be simple readiness questionnaires, etc.

The second part of the equation after data collection is data management. While some organizations have no problems in building their own enterprise solution others are using third party platforms like Dropbox to speed up data sharing. This is a great way to temporarily get around the problem but in order to continue making progress, API integration should be the next area of evolution for medical and monitoring technology companies.

Dashboard design and data visualization is paramount in helping management and team coaches understand and utilize data to help accelerate performance, however it is still early in this journey. What are mistakes you have seen and what solutions do you use to solve this problem so the proper data is collected and visualized?

I shake my head every time I see teams investing a big part of the medical and performance budget on perfect AMS (Athlete Management Systems) that are far from perfection and do not provide a real performance advantage. Data do not always need to be in one place and sometimes breaking things down helps to detect bugs in different areas like current working dynamics within and across different departments, communication strategies or staff’s education on data acquisition and understanding of each metric. The goal is to continue refining each part of the process to ensure progress is constant from one season to another. Coaches are not engineers nor they should be and Excel is in most of the cases the best and more flexible solution and something they are familiar with.

Using Excel as a basis can eventually lead to more specific software like Klipfolio which is a simple KPI panel that requires no coding knowledge and coaches can use to customize their own dashboards. It also integrates with other products like iFormbuilder through their XML API, which makes collecting data from surveys faster and enlightening for both, coaches and athletes. Readiness surveys are commonly used by almost every coach, whether they are short or long questionnaires, and better strategies can be implemented to avoid boredom and ensure reliability of the data.

The process is well explained in this video below. 

Roambi is useful for more extended reports (end of the month, seasonal reports) as it provides dynamic graphs where information is layered in different levels and coaches can drill down deeper into their data with only a few taps.

Performance and medical dashboards should be kept simple. Stephen Few’s bullet graphs are useful to compare current vs. target (benchmark) values. Sparklines enable analysis of change over time and reveal patterns behind the data. A simple traffic light system can highlight abnormal scores and heat maps of the human body help to visualize neuromuscular imbalances, kinetic dysfunctions or local fatigue. More than that can turn into something confusing and unlikely educational.

Try searching “Data Visualization Software” on Google and you’ll get a bunch of different products all of them with their pros and cons. Swapping from one solution to another is relatively easy but the real advantage is in polishing the process behind all these data.

Manchester City collaborated with OPTA to release their data to the community to Crowdsource the football minds. How can the sports performance and medical community do something similar without losing the competitive advantage?

First of all, full credit to Manchester City for being pioneers in embracing crowdsourcing in an attempt to evolve and improve. I think it was a great idea for both the team and the community of 5.000+ performance analysts who registered for the project and had the chance to being exposed to such levels of data.

Performance and medical data is slightly more delicate than tactical data for a number of reasons and I will try to state a few:

• - There is no consensus on what metrics should be measured in every different sport.

• - Even when the same metrics are measured, there is still no consensus on data acquisition protocols, for example a team may prefer to measure HRV  (Heart Rate Variability) levels immediately after players arrive at the training ground and others after a 5 minutes standard warm up.

• - Even if consensus is established on what and how to monitor, many teams would not be willing to share medical and performance data as it tells which teams are training and which ones are going to the training ground to read the newspaper and get a free meal.

• - Performance and medical data is not managed in the same way across different teams competing in the same league, unlike tactical data which relies on Opta, Prozone, etc…

Unfortunately, the sports science community is behind in this area and until basic aspects like the ones above are solved Crowdsourcing is unlikely to happen. Although I don’t have much details about the project, an interesting approach is the one being implemented in the MLS league with the Adidas miCoach system but that is only one part of the whole. 75% off all the other medical data like treatments, workouts, or training compliance won’t be shared. Yet it may be a good way to get things kick started.

Last year you spoke at the Boston Sports Medicine and Performance Group about Tensiomyography (TMG); a tool to diagnose the muscle status of athletes. With so many tests you have done over the last five years what type of algorithm is now possible to help predict injury? Does the algorithm evolve over time?

The DYNAMO algorithm is something that I provide for teams that are currently using TMG. I decided to release it after doing close to a thousand tests from 500+ athletes competing in a variety of sports with the only purpose to quantify peripheral stress levels in a simplified manner and provide an idea of neuromuscular readiness. It was created by piecing together data regarding muscle fiber composition, muscle contraction times, electromechanical delay, current muscle stiffness, as well as symmetries and synchronization levels across different muscle structures. It is sensitive to the requirements of each sport as well as the individual characteristics of every athlete.

With DYNAMO, coaches can immediately get an idea of whether their athletes are fully, partially or poorly recovered from a neuromuscular perspective as well as an indication of which areas are more affected in order to implement more informed interventions. Mechanical impairments due to fatigue and kinetic dysfunctions are proven to play an important role in any injury mechanism. This doesn’t show on HRV trends or GPS statistics. It is key to be more specific on what to monitor and how.

Evolution comes from trying something in a robust environment and see how it works. Feedback, more data, and relationships with other KPI’s will undoubtedly help to sharpen the algorithm and some slight updates for the 0.2 version are currently being written down as I answer this question.

With sincere gratitude for both the time spent and thoughtfulness displayed by Jose in answering these detailed questions, we thank him for his contribution to the TopCoder Blog and wish him nothing but greatness as he continues to push his field forward via better data and enhanced understanding. Thank you Jose. 

Technology, Open Innovation, the Future: Subscribe to the TopCoder Blog (top right)

You Tweet? So do we. Join us on Twitter. 

 G+’er? Add TopCoder to a Circle today!

 Crave great videos on innovation and technology? Subscribe right here!