The China Mail - AI only just beginning to revolutionize the NBA game

Using artificial intelligence, a top basketball team found the right defensive strategy that made the difference to win the NBA championship.

Data specialist Rajiv Maheswaran declines to name the outfit that leveraged AI analysis to victory, saying in a corporate video only that it happened several years ago.

That was "the moment that sealed it," added the co-founder of tech startup Second Spectrum, which provides the league with swathes of player positioning data gathered during crucial games.

Analytics have transformed the NBA over the past decade, with AI and other breakthroughs still ramping up.

Embryonic in the early 2000s, the revolution truly took hold with motion-capture cameras installed in every venue in 2013.

Ten years later, new tech upgraded renderings of the court from 2D to 3D, unlocking even more precious data.

Each player wears 29 markers "so you know not just where they are, but you know where their elbow is, and you know where their knee is," said Ben Alamar, a sports analytics writer and consultant.

"You're actually able to see, yes, that was a high quality (defensive) closeout," said Tom Ryan, head of Basketball Research and Development at the NBA, describing an often-used manouvre.

"It's adding more context to that metric."

"Now all 30 teams are doing significant analysis with varying levels of success," said Alamar.

Houston, Golden State and Oklahoma City were often cited among early adopters at the turn of the 2010s.

This season, Oklahoma City is on top of regular season standings, "and they play different," said ESPN Analytics Group founder Dean Oliver.

"They force turnovers, and they have very few turnovers themselves. So there are definitely advantages to be gained."

"It's not going to turn a 25-win team into a 70-win team during the season, but it can turn a 50-win team into a 55, 56-win team," according to Alamar.

AI allows for "strategic insights" like "understanding matchups, finding the situations where players perform well, what combinations of players," he added.

- 'Optimizing shots' -

None of the dozen teams contacted by AFP agreed to discuss their work on analytics.

"Teams are (understandably) secretive," Oliver confirmed.

Even before 3D, motion capture data was already shifting the game, taking basketball from a more controlled pace to something looser and faster, he added.

The data showed that faster play secures more open looks and a higher percentage of shots -- a development that some criticize.

On average, three-point shot attempts have doubled over the last 15 years.

"As a league now, we look deep into analytics," Milwaukee point guard Damian Lillard noted at February's All-Star Game.

While it perhaps "takes away the originality of the game... you've got to get in line with what's working to win."

The league is taking the issue seriously enough that Commissioner Adam Silver recently mentioned that "some adjustments" could be made to address it.

Even now, AI has "plenty of upside" yet to emerge, said Oliver.

"The data is massive, but converting that into information, into knowledge that can be conveyed to players, that they can absorb, all of those steps are yet to be done."

- 'Get calls right' -

The league itself is pursuing several analytics and AI projects, including for real-time refereeing.

"The ROI (return on investment) is very clear," said Ryan. "It's about getting more calls right, faster and in a transparent way to our fans."

"We would love a world where if a ball goes out of bounds and you're not sure who it went off of, rather than going to replay you look at high frame rate video in real time with 99.9 percent accuracy... That's really our North Star."

Spatial data can also extend the fan experience, shown off during the recent "Dunk the Halls" Christmas game between San Antonio and New York.

An alternative telecast rendered the game in video game-style real-time display, with avatars replacing live action images.

"We want to experiment with all different types of immersive media," says Ryan. "We just want to be able to sell our game and present it in compelling ways."

V.Liu--ThChM