Sometimes, love is accurately identifying just how wrong things are going in the spouse’s latest project. Megha Gambhir put the figure at “95 percent”.
Her husband, table tennis coach Deepak Malik, was boggled over why his young ward Wesley do Rosario’s southpaw advantage wasn’t translating into more victories with wicked playing angles, though he intuitively knew what was messed up.
It was because – Megha would tell him calmly after a loss – the boy was not targeting the extreme forehand corner and playing across to the opponent’s comfort zone on the backhand. And losing speed on the ball along the way.
“She said 90-95 per cent of his returns were sent wrong even as the ball quality reduced. It was an eye-opener to be pinpointed to the exact problem!” Deepak exclaims.
Having nailed the mistake in a video analysis, the Goan junior would bound out of the also-ran grade and make the country’s Top-4 in the next 5-6 months, having repaired a bunch of other tactical mistakes. Megha, a hobby shuttler when young but an intrepid techie with Artificial Intelligence pro Anand Hariharan, had helped with another piece of the jigsaw to send yet another aspiring paddler on his way.
This last week, Stupa Analytics, of which Megha is the CEO, tied up with the International Table Tennis Federation (ITTF) for performance-analytic tools making them available to 226 member countries. ITTF reckons 30 million play the sport professionally across the world, making it a 1 billion market. And a simple practice video shot on a telephone can throw up a plethora of data – from nature of points won on service, long & short length conversions, types of errors & winners, clusters using ball heat maps, trend of ball speed, video highlights backing data, stroke and hand-type analysis.
Table tennis, a sport that is leaving its bashful presence amongst racquet sports behind with bold sponsorship partnerships with Coca Cola last year, is leaping into digitalisation of its player data – first with online coaching via Pingprofy and now AI analytics with Stupa.
Circa 2018, Megha had watched her husband and former cadets national coach Deepak agonise over a practice video, sitting with pen and paper and manually dissecting a player’s game. “I would sit with Excel sheets and jot down the percentage of strokes and winners, strengths and patterns in tactics,” he recalls.
Megha, with fruitful tech consultancy stints at HCL, Cognizant, PepsiCo and Ernst & Young, was itching to start something of her own, when Deepak would request her help to design a tool to ease his workload.
Starting out with incremental tables, albeit with manual data extraction where a match would still take 4-6 hours, she would rope in AI pros to successfully scale up a model to analyse.
The Rohtak-born coach, who had picked up the sport in Class 8 over his parents’ friend’s dining table, was now helping more than his dozen wards to read their games better, and plan practice sessions in Delhi. The sport was also the reason the couple met. “I got admission into Hindu College in Delhi only on the basis of my sports quota. Without TT, it was impossible. We met there,” Deepak recalls.
A bit geeky, he would shun all the emotional rollercoasters of sports biopics to focus on how the coaches worked in movies. “I watch every sports movie to observe the coach. But the one that sparked this idea was Miracle (2004), about USA’s Winter Olympics ice hockey win over Russia. I wondered if they could use data back in 1980, why aren’t we doing it in TT now?”
Megha would bring alive his vague idea by assembling a team of (now) 20 IT and game analysts, as the product helped them crack deals with individual federations from US, Portugal, Hungary and Sweden.
Ball tracking along with ball speed posed the most difficult problems for AI to resolve through a mobile camera. “Evolution of this feature was simply a roller coaster… as at many steps we felt it is difficult to achieve but the persistent effort of Anand Hariharan, our chief AI Architect, along with the passionate team led to the big breakthrough…. it is simply amazing to see ball speed and tracking just on the mobile. The second insight which we had fun developing is something that we are doing now – auto suggestion and predictions,” Megha says.
Back home, 21-year-old Surat lad Manav Thakkar would benefit early. At the 2019 Senior Nationals semifinal against G Sathiyan, Deepak would take Thakkar to the venue at 6.30 am, and over the next four hours use the freshly-culled data to arm him with a surprise weapon. Playing closer to the table, when trailing 5-9 in the decider, Thakkar would blitz the next six points on a bounce-changing strategy to win the match. “He’d never taken even a game off him (Sathiyan) before. But data helped us devise the strategy,” Malik says.
Plans for the future
Megha is aiming higher, well aware that the East Asian powerhouses are masters of the game. “We know China, Japan, Korea have their sophisticated in-house analytics and they have been recording opponents extensively. But Stupa will open this up not just to Indians, but also North America and Europe. If more players become self-aware and get data-driven insights, it can optimise training and raise levels,” she muses.
Personally, she’s keen on forays into badminton, squash, tennis as well as billiards, snooker and pool at a later date. “I didn’t start out as a TT player, but I was keenly interested and travelled to tournaments to understand the sport’s working,” she says. Two analysts are now completely into AI-based data – a plug-in camera records the match, video goes on Cloud, and results are delivered within minutes.
Work on an angle -agnostic software is underway. And videos could make way for sensor-based analysis. “We’ll go deeper. So, make it so nuanced that it can identify a forehand topspin push chop from something that’s very close to it in action, but lands differently,” Megha says.
Tournament sites, broadcasters as well as betting sites will be able to access the data for live streams and graphs, from ITTF’s multiple engagements. What the husband-wife duo though knows minutely is that all of it hinges on how the backroom work looks in the foreground. “Execution is all-important,” she says.