Coaches and scouts have played a significant role in the success of teams worldwide. Their ability to identify and nurture sporting talents of the future and take real-time decisions during matches cannot be understated.
For a long time, they have relied on past experience, gut feel, and the knack of identifying ‘intangible’ player qualities to excel in their jobs. The cut throat competition in the sports industry and the ever-decreasing margins of error have made it imperative for coaches to modernise their approach towards training, as well as supplementing their current abilities with evidence-based decision making.
While technology is at the disposal of many professional coaches today, the use of artificial intelligence (AI) and machine learning hasn’t been democratised to more minor and local leagues. CoachFirst is one of the first organisations to address this gap. Founded by technology expert Trevor White, National Football League (NFL) kicker Pat O’Donnell, and former all-American rower Megan O’Donnell, CoachFirst is an all-inclusive platform to connect players and coaches. Currently, in collaboration with Quantiphi, the company is exploring player performance analytics through deep learning and computer vision algorithms.
Unlike traditional coaching, where most of the time trainers rely on their gut to improve player performance, AI can influence in-game strategy, optimising the team’s starting lineup, and studying opponents’ moves with granularity. Sports like football involve fiercely fast motion and close contact that can sometimes become challenging for coaches to obtain valuable analysis. This is something that fits right into the alley of AI.
Understanding a player’s performance involves evaluating a decided set of game parameters during a live match or a training session. This is done by leveraging deep learning use cases like player tracking, action recognition, and real-time ball tracking.
The ability to objectively analyse what players are doing could have a lasting impact on strategic decisions coaches make, both before and during a game. Video intelligence models combined with visual dashboards make it easier for coaches to interpret and prepare players well, without biases. This creates transparency for players as they gain meaningful insights on how to improve and measure progress over time.
Let’s understand this better with an example:
Quarterbacks are the leaders of the offence and are responsible for calling the play in the huddle. Almost always, they are the player with the responsibility of throwing forwarding passes. So the metrics which can be used to evaluate quarterbacks include:
● Time to throw (the time elapsed from the time of snap to the time of the throw)
● Average completed air yards and average intended air yards
● Longest completed air distance (the number of yards the ball has traveled on a pass)
Complex metrics like ‘aggressiveness’ can be quantified too. For instance, the number of passing attempts when the defenders are very close (a yard or less) from the receiver. For punters, the metrics could vary from ‘hang time’ and the time elapsed between ‘snap to kick’. It is crucial here that coaches could think of parameters necessary to measure for a particular game and examine if AI could measure it for them.
“At CoachFirst, we are really excited about our partnership with Quantiphi,” said Trevor White, CoachFirst chief executive. “We see a significant opportunity to help coaches and athletes at all levels make more informed decisions leveraging artificial intelligence. Our solution derives analytics that are typically overlooked by coaches at fundamental camps, one-on-one training sessions, and competitive competitions.”
While metrics for analysing the player performance of different sports differ, all models have a common ground. In the case of football, a model must require a copious amount of critical metrics such as the apex of the ball, hang time, time to throw, aggressiveness, and snap to kick rate to train itself. The model’s ability to track players and learn player actions (like kicking or throwing) and events (such as a touchdown or a tackle) are one of the starting steps. Other key elements of the solution include, but are not limited to:
Easier said than done, one could run into limitations when building the model. For example – helmets blocking players’ faces or too many athletes in one frame. This could impend the accuracy of such models. Thus they need to be circumvented smartly by leveraging parameters like jersey numbers.
Generated insights from action video footage will elevate the game, enhance fan experiences and bring transparency to the sports ecosystem.
Patrick Murphy, vice president, Quantiphi
The answer is a resounding yes. Every sport is different, requiring its metrics, but the approach lies in understanding the utility of a metric and overcoming its limitations. For instance, drawing real-time data revolving around tactical assessment, movement analysis, video, and statistical databasing can help construct predictive models to anticipate team and player performance for soccer, baseball, and basketball leagues. Correspondingly, major league scouts leverage computer vision, machine learning, and other forms of AI algorithms to analyse player performance analytics, game videos, and data from various sensors to identify talent and make sure no talented athletes fall through the cracks.
Partnerships similar to Quantiphi and CoachFirst are the next decisive steps in the sports ecosystem. AI enables football coaches to blend good old human intuition with new-age technologies to improve players’ techniques, maximise training results, and devise personalised programmes. The technology will help coaches understand players’ strengths and weaknesses, enhance their development and coaching practice and get in-depth reviews of players in real time.
Visit the link to learn more about Quantiphi’s capabilities for sports.