Sidney Crosby and Connor McDavid are gliding to the face-off dot in the Penguins’ zone. Tie game. Final minute. Fans glued to their screens.
Up pops an accurate percentage of the odds of each star winning the critical draw.
Crosby gets tossed from the circle, Bryan Rust takes his place, and the odds adjust in the blink of an eye.
This is precisely the type of enhanced viewing — and, potentially, betting — experience hockey fans will witness down this season’s stretch run.
The NHL and Amazon Web Services, Inc. (AWS) partnered to develop the novel Face-Off Probability, a live, in-game stat that will be displayed as a graphic that instantly shows the odds of a player winning the ensuing draw.
Sportsnet will unveil the tool in Canada this week; viewers on ESPN and Turner can expect to see Face-Off Probability later this month.
Because hockey’s free-flowing nature allows so few in-period breaks to analyze and predict, the NHL and AWS worked in conjunction with the league’s broadcast partners to introduce the first predictive stat to the fans' experience.
“There's this little run-up to every faceoff where a broadcaster has a moment in time to provide the viewer with some meaningful insight,” explained Dave Lehanski, NHL executive vice president of development and innovation. “So, what you'll see is a graphic on-screen that will be updating and changing in real time leading up to the actual faceoff.
“More than anything, it just can be visually really compelling to see it prior to the play occurring, as the players are skating to the faceoff dot.”
To obtain accurate odds for every conceivable matchup, a small team of scientists and engineers on Amazon Machine Learning Solutions Lab began crunching data early last season. They sifted through all 200,000-plus faceoffs from the past 10 years, factoring more than 70 different data points.
Individual faceoff profiles were created for every NHLer, compiling factors such as weight, height, home and away success rates, head-to-head matchup history, handedness, game context, faceoff location, game score and time of game.
That profile is rounded out with puck and player tracking, then tweaked at subsecond latency with data gleaned from the very game taking place. For instance, if Crosby is on a heater that night, his odds will escalate.
The odds are fired into the cloud and onto the screen for commentators and mobile users to see before the linesman drops the puck.
“You’d be surprised,” said Priya Ponnapalli, senior manager of Amazon Machine Learning Solutions Lab. “Machine learning can find patterns in data that are not obvious at first glance. And by combining so many different variables, it gives us a chance to make predictions.”
While the NHL can now offer Face-off Probability for every faceoff in every game, broadcasters will use the tool at their discretion for story-telling purposes.
OK. So, with faceoffs occurring 50 to 70 times per game, when do we start making nickel bets on every single draw?
“I can envision a world where fans are betting on anything that happens within the context of any sporting events in real time — because the technology is going to allow it,” Lehanski said. “I do believe that will be possible, based on technology, infrastructure and the incredible capabilities that AWS brings to the table.”
First, however, the NHL wants its fans and broadcasters accustomed to seeing and using the tool.
And, make no mistake, Face-off Probability is just the tip of the iceberg here. Some other predictive tools are already being cooked up in the lab.
“We're looking at everything. We're looking at every aspect of the play with goaltending. We're looking at defensive zone coverage. We're looking at almost everything tied to offense — where offence is generated and how it's generated,” said Lehanski, excited by the possibilities.
“We now have the ability to, really, quote-unquote almost do anything. Because of the ability to deliver not only the analytics, but the graphic and the visualization in real time.”
