Future with AI | Champions of sports analytics want to change how cricket, boxing, table tennis, football are played, and viewed

At FC Goa, data scientist Amrit Murali runs AI models on match recordings for statistics such as expected possession, expected threats and expected goals.

As a table tennis player and coach, Deepak Malik always realised the need for analytics to excel at his craft. Back in the day, he would record a match, then go back and watch it all over again, before putting down pointers to share with his wards.

“The entire process was exhausting. With a group of 10-15 players, it’s almost impossible to work in this manner. And there was no way to present my feedback to the players where they could understand it better,” recalls Malik, a former national-level medallist.

At Hindu College Delhi, Malik met his future wife and business partner, Megha Gambhir. While she worked a techie job in the corporate world, Malik urged her to use technology to solve his quandary. Finally in 2018, Gambhir quit her position and the duo started exchanging notes on how they could utilise artificial intelligence (AI) to arrive at possible solutions.

Already, players and coaches can use AI to observe how minute changes, like the angle of the elbow when a player is hitting the ball, can change outcomes. (Photo by Josh Sorenson via Pexels)

A couple of years later, they co-founded Stupa Sports Analytics, hoping to change the way table tennis is approached. Once the match recording is uploaded to their app, the AI tracking software is able to extract data points from the game - say, the movement of the player or the angle of the elbow at which they are striking the ball. Or the number of points won on the forehand. It is then presented in the form of pie charts and bar graphs, for easier assessment by players and coaches.

Stupa’s tools are currently being utilised by some top-50 players of the world such as Adriana Diaz of Puerto Rico and Kanak Jha of the United States. Closer to home, it was instrumental in taking Malik’s ward, Manav Thakkar, to the top of the U-21 rankings of the world in 2020. They have also provided their services to the International Table Tennis Federation in the past, and today work with the national teams of Germany, France and Sweden.

“A player may not necessarily agree with a coach’s suggestions at all times. But now there’s data and videos to back it up, all available at a click. Besides being efficient, it has also changed the way the game is approached,” Malik says.

Stupa Sport Analytics cofounders Deepak Malik and Megha Gambhir.
***

Over the last few years, the physical prowess of athletes has been complemented by the use of AI to make training and coaching more scientific. At the Centre of Excellence in Sports Science and Analytics (CESSA) at the Indian Institute of Technology Madras, Babji Srinivasan has been collaborating with researchers to arrive at handy solutions for various sports. He is currently working with athletes at the Inspire Institute of Sports (IIS) in Vijayanagar to provide insights on performance.

Also read: IIT Madras enrols 30 candidates for first cricket analytics course

Over the last few months, they have been continuously monitoring boxers at IIS through the SmartBoxer system developed by CESSA. Two cameras installed at the ring capture footage of every bout over a period of time. This is then broken down by AI tools to provide insights such as the number and types of punches, and how they are utilising the ring. They work in sync with John Warburton, a British coach who works with IIS and the Boxing Federation of India.

“It’s essentially a SWOT analysis of every player that helps understand performance. The entire engagement is with the coach because he is the end user. And he now has data to backup what he’s observed. For instance, one of the boxers had a very wide gap between his feet and the data showed improvement when he modified his stance,” Srinivasan says.

CESSA has also developed insoles that are currently being used by the Royal Challengers Bangalore. It can return data such as the stride length, frequency, acceleration and the pressure exerted by the player to aid performance. This is also being extended to other sports such as weightlifting, wrestling and boxing, as well as for rehabilitation purposes at hospitals. And in the time ahead, they are looking to use the same at the MRF Pace Foundation.

“Bowlers like Wasim Akram were able to generate a lot of pace with a short run-up. For that, the stride frequency needs to be high and one needs to generate force when you hit the ground. These insoles can really help cricketers with important insights,” Srinivasan says.

For the batters, companies like StanceBeam have developed a cricket bat sensor that can improve their technique through data on timing, power and angle at which they hit the ball. Through a combination of sensors and video technology, they also provide video clips of the actual shot being played by the batter.

StanceBeam Striker. (Photo courtesy StanceBeam)

“AI was being used for other sports around the world and we realised it could also help cricket. Today, we have international players on board from Australia and the West Indies who use our sensor as part of training. In fact, a few have even picked their bats based on our technology to arrive at the best fit,” says Arminder Thind, CEO at StanceBeam.

Their early prototype was a small sticker sensor that could be pasted on the bat. But since batters didn’t like anything on the bat’s playing area, it has today evolved into an attachment that goes on top of the handle. They recently launched a watch that can also return similar advanced analytics.

“We are also looking at making Decision Review System technology cheaper and more accessible through AI solutions and hope to launch it in the time ahead.” Thind says.

Arminder Thind, CEO at StanceBeam,
***

At FC Goa, a football club that features in the Indian Super League (ISL), data scientist Amrit Murali has been working with the video analysts to provide a wide range of insights that can be utilised by the coaching staff.

“Most AI is built on a statistical machine learning algorithm. We use these models to crunch football data and it gives us new statistics that indicate something different about the game,” Murali says.

Every action from a match is extracted and after running AI models on this data, Murali is able to gather a set of statistics such as expected possession, expected threats and expected goals. These are valuable insights for upcoming fixtures.

“Parameters like expected pass completion that is calculated based on previous matches can tell you whether a player has over performed or under performed in a game. Or during set pieces, we can figure out the starting position of players and the zones you can target for scoring opportunities, based on how the opponent tends to defend. You can do this by watching videos as well, but the idea is to automate the entire process,” he says.

“In football, the major use of AI is in comparison, not predictions. So the idea is to set a baseline based on the other teams in the league and then see how the team is faring as the season progresses,” he adds.

Amrit Murali, data scientist, FC Goa.

Murali has worked with all kinds of coaches, some who have soaked in as much data as possible, and others, who prefer to bank on their intuition and understanding of the game. But he feels data and AI technology can provide the edge, especially when there are two equal teams facing off.

The most effective utilisation of this technology is for signing foreign players. FC Goa tracks thousands of matches across multiple leagues that are accessible through scouting platforms. They use the data to mark out players who fit their plans, before visiting to see them play in person.

“There was a player whom we identified purely through data around four years ago, who is today a prolific scorer in the ISL. Of course, there is a percentage of error involved, but it’s possible to do this using data,” Murali says.***

The scope of AI extends to the way we consume sports content as well. At FanCode, it is being used to quicken various processes and customising content for their users.

“AI is reshaping sports tech and gaming, transforming how we play, watch and engage with sports. We have been using it to offer personalised feeds, short-form content recommendations and real-time highlights generation to our users. It’s also helping us maximise the efficiency of all our communications and extract value from thousands of hours of sports content,” says Prasana Krishnan, co-founder at FanCode.

FanCode is currently using Google Cloud’s Vertex AI to analyse commentary during a live game, extracting data from it and sending out notifications to enhance the viewing experience.

“FanCode is also in the process of indexing every frame of video to create a custom highlights engine. Through this, for instance, a fan can watch all balls of Kagiso Rabada faced by Virat Kohli during a match or a tournament,” says Bikram Singh Bedi, vice-president and country managing director at Google Cloud India.

Stupa Sports have also developed SuperCast, through which they are trying to make table tennis more engaging for the audience by providing real-time data points during live coverage.

“If you look at cricket, there are so many things for the commentator to speak about based on data. Something similar for table tennis will certainly increase viewership,” Malik says.

AI can also be a handy tool for training, providing essential analysis and tips for coaches as well as budding players. Srinivasan believes these systems can make sports more accessible for aspiring athletes.

“There are so many kids interested and who want to reach the highest level. But the number of coaches are few and not everyone has access to one-on-one training at the grassroots. This can help novice coaches understand the game and in turn, guide their students better,” CESSA's Srinivasan says.

“Just imagine everyone having their own personal coach through AI. We see AI democratising sports for everyone,” StanceBeam's Thind adds.

But there are challenges that emerge time and again. Thind highlights the need for real-time data processing with limited lag, which can often be a task due to network issues. Funding remains a constraint for continuous monitoring and analysis, according to Malik.

Srinivasan, however, points out a bigger challenge that has little to do with technology.

“The traditional mindset of coaches can often get very difficult to cut across - either they see it as a threat or as junk. It takes effort to get the idea across to them that AI is never going to replace coaches. Things are gradually changing and people realise that AI is revolutionising everything. And if they don’t adapt, they will soon become obsolete,” Srinivasan says.

More in the Future with AI series: Artificial intelligence will be a weapon in the fight against climate change