How The Next Generation Of Sports Will Be Driven By Machine Learning
Future Of Recruitment and Scouting Talent
Why should data only tell the past in sports? Currently, players are primarily recruited and drafted based on past accomplishments. However, previous performance is a lagging indicator of a players future success. There’s a considerable amount of human bias involved in assessing talent. Instead, athletes should be purely examined by data-driven analytics for predictive outcomes. More efficient recruiting and scouting talent will be accomplished with the use of machine learning.
Machine learning, a sub-field of artificial intelligence (AI), explores the construction of algorithms that can learn from and make predictions on data. Machine learning is everywhere. Industries such as shipping, automobiles and venture capital are using algorithms to predict outcomes. “A few days of advantage can mean the difference between winning and losing a deal,” early stage firm SignalFire Chief Executive Chris Farmer tells Wall Street Journal. Farmer continues that operating the traditional way “is playing a game of telephone — with tin cans and a string.”
Yes, sabermetrics, made popular by Moneyball, applies analytical evidence to assemble baseball teams, but what about an algorithm that learns throughout its lifetime free of human bias? The value is seeing the future and that’s what new disruptive startups are tackling today. One sportstech startup at the center of it is Brooklyn Dynamics.
Brooklyn Dynamics is an AI platform focused on talent ID, scouting and team cohesion. The proprietary algorithms are based on a decade of elite athlete data and successes. Clients include 5 MLB World Series champions, NFL teams and world-class events, such as The Tour de France and F1 racing. Brooklyn Dynamics masters predictive analytics and maintains a 97% accuracy rating in MLB and NFL drafts, forecasting player vs player comparisons, salary cap modeling, and live in-game scenarios. Recently, I spoke with Mathew Cole from Brooklyn Dynamics about the future of sports recruitment:
What Is The Current Recruitment Model?
Combines, grand finals/short term “camps” are all antiquated. Why do we scout players based on single events when we recruit them to play long seasons? The issue is, under the current environment you can not monitor all players, across all leagues as it's not physically or financially possible.
Platforms that collect data, standardize and create uniformity in the collection allow players to be evaluated 24/7. The premise we work under is that we are creating the data CV of the athletes for the entire life cycle. Take away any human bias…you evaluate a player and their data, you don't see race, nationality or other factors that often cloud judgment.
In speaking with NFL presidents, NBA owners, and GM, they say their biggest reasons for attending the combine is not always about the players physically…They often attend as it’s one of the few annual events where everyone is in one place so it’s the perfect networking. They may learn how the athlete deals with the media schedule/pressure, however, the real value to attending is the fact all contacts league-wide (scouts and agents) attend.
How Does Data Solve Declining Participation?
One of the key drivers for me personally is establishing a digital toolset that allows athletes to be recruited regardless of where they are. If you look at high talent clusters — football in West Africa, East Europe, South America — they are generally under-resourced areas that additionally suffer under corrupt systems and political fighting. Establishing uniform data platforms that allow no human bias, and only data analytics to evaluate talent levels ensure that players can be recruited on true information.
If players feel there is a genuine path to success, then they stick with the sport or take the sport more seriously. Allowing the data to be the deciding factors, show players there is no incumbent system that is against them.
Are There Misaligned Objectives?
Generally, there is a feeling amongst teams that better transparency on data, will benefit everyone.
As per the NFL Collective Bargaining Agreement (CBA), there are strict rules around what medical information the teams can know about the players especially around contract time. As in the GM or the organization who is negotiating the player contract will often not have full viability into his/her own teams medical status. The CBA and in particular the players/agents, say this is personal information that if obtained will negatively affect the players to negotiate a fair value contact.
The teams obviously wanting better access to data (medical, physical, performances) want to optimize for wins! If that means cutting or trading a player for the benefit of the organization, so be it. Having better access and transparency to the data, and engaging AI platform like Brooklyn Dynamics that can accurately model this forward creates the strongest environment for wins.
The players and agents won’t like this as it is likely to mean shorter contracts, less security and less guaranteed income for agents.
What About Privacy, Ownership, And Protection?
This is similar to what I mentioned around pushing the envelope and establishing governance when the governance will always be behind innovation.
As a machine learning company, we have the ability to model out the future across a range of topics. The more data we have, the more questions we solve the exponentially smarter our engine becomes. As mentioned, we anonymize our data to avoid privacy conflicts. As this industry scales, you will start to get companies, organizations, and players to set clear ownership guidelines.
Example:
Player (A) in college produces 3 years of performance data. Who owns the data:
- The College?
- The player?
- The company that tracked the data — Catapult/Zebra/medical companies?
If each has a reasonable claim to the data, then how does that athlete use that data for their desired outcomes? Recruitment, training, playing improvements? Can they transfer from college to the professional team? If so, does a company like Catapult have to share with a competitor and if so, how do you establish a set of “best practice” so the data can be clean? Or, is the data simply “lost”? If the player can’t use it for their benefit, then why should the college be able to use it to benchmark performance or recruitment purposes?
Why Do Teams Need Your Machine Learning Engine?
We collect historical data from every team we’ve ever worked with as opposed to a team using their own in-house data, human characteristics, and patterns of behavior.
What’s Next For Brooklyn Dynamics?
Soon, we will be launching our first consumer-facing mobile app, Insyte, providing scouts, agents, and GMs access to high-level machine learning and predictive outcome modeling in the palm of their hand. The Insyte app will be free to all athletes, of all ages, allowing athletes to see their global ranking and to be showcased in an ever-evolving global market to scouts, high schools, colleges and professional teams around the globe. For the scouts and GMs, this open market of athletes ensures Insyte is pulling in an “evergreen” set of data points, increasing the value, accuracy, and range of predictive outcomes possible.
The value proposition is different for every team because teams are looking to derive different information. By negating all human bias, machine learning engines like Brooklyn Dynamics take away the grey area of what can or can’t be done with players’ data. There are positive signs that we are heading in this direction. However, there won’t be widespread acceptance until more teams and players continue to push the envelope. The accumulation of data, not algorithms, is the key to machine learning success. There has to be innovation first before you build guidelines around it. Sports have come a long way, so why hasn’t data-driven analytics?
