How the NHL Is Using AI Right Now — And What It's Actually Changing

Hockey has always been hard to quantify. The puck moves fast, players cycle through lines every 45 seconds, and the game rewards instinct in ways that resist tidy measurement. That made it one of the last major sports to go all-in on data. The last two seasons changed that.

Here's what's actually been deployed — and what it's doing on the ice.

The Infrastructure First: Sensors in Every Venue

Before any AI model can run, you need data. The NHL's Puck and Player Tracking (PPT) system laid that groundwork. Infrared emitters and cameras installed in all 32 NHL venues track sensors embedded in the puck and in every player's sweater. The system transmits positional data at 100Hz — 100 data points per second, per player, per game.

That data infrastructure is what made the NHL's AI partnership with Amazon Web Services (AWS) possible. Starting in 2021, the NHL and AWS began collaborating to make the most of these data sources.

Face-Off Probability: The First Real ML Deployment

In 2022, Face-Off Probability became the first AI/ML-driven NHL analytic to launch within the NHL EDGE IQ platform, helping determine who is most likely to win a specific face-off based on multiple historic and in-game data points.

Face-offs happen roughly 60 times a game. Winning them at a higher-than-average rate is one of the cleaner predictors of puck possession, and possession correlates directly with scoring. Coaches have always tracked face-off win percentages — but a model that accounts for matchup-specific variables (this center, against that center, in this zone, in this game state) is a different tool entirely.

Opportunity Analysis: Reframing Every Shot

In April 2023, the NHL and AWS launched Opportunity Analysis — a metric that assesses the difficulty of a shot at the exact moment of release. It processes how fast the puck was traveling, the shooter's change in angle, the goalie's positioning and height, and dozens of other variables drawn from the live PPT data stream. The output is a simple rating: high, medium, or low chance of a goal.

According to Brant Berglund, NHL Senior Director of Coaching and GM Applications: "With this product, we're going to be able to output massive amounts of data on the play leading up to every shot, curated in very close to real time."

For player development, this matters more than raw shot totals. A forward generating high-opportunity shots from the slot is a different asset than one generating the same volume from bad angles. Opportunity Analysis gives coaches and GMs that resolution.

Ice Tilt: Measuring Momentum Objectively

The 2023–24 season brought Ice Tilt, the fourth metric built on the NHL EDGE IQ + AWS stack. It measures team momentum — not based on score, but based on the positional flow of players and the puck. The data loop completes in under three seconds, running across all 16 active NHL arenas simultaneously. NHL fans could see the metric in action during 2023–2024 season broadcasts, where the data drove visual representations of real-time momentum shifts.

The coaching application is the more interesting one. A team that consistently loses Ice Tilt in the third period, even in games they win, has a late-game structural problem. Identifying that with data rather than gut feel is a material difference for a coaching staff.

Sportlogiq: The Platform Running Across All 30 Teams

The AWS partnership covers league-wide broadcast analytics. The tool most embedded in day-to-day team operations is Sportlogiq's iCE platform. Through the use of advanced AI and computer vision technology, Sportlogiq provides 30 NHL teams and another 100 professional organizations with analytics, player development, and scouting tools.

Their system runs on a multi-camera setup inside arenas, achieving 99.9% tracking accuracy with less than 10cm localization error. Pre-game, in-game, and post-game reports go directly to coaching staffs, with all metrics linked to video clips for immediate context.

Coaches can identify optimal line combinations and defensive pairings using expected goals metrics weighted against strength of opposition. That's not a spreadsheet exercise — the platform surfaces it automatically.

The scouting angle has real precedent. Sportlogiq's analytics system ranked Sean Durzi, a defenseman in the Ontario Hockey League, in the top 40 ahead of his draft year. Top teams passed on him. He went undrafted, then broke out the following season. "Teams could have got him as a late-round pick last year if they'd followed what our system was telling them," said Christopher Boucher, manager of hockey analytics at Sportlogiq.

Dallas Stars: A Team-Level Benchmark

The Dallas Stars received the 2024 Stanley Award in Strategy, Analytics, and Innovation, awarded during the annual NHL business meetings. Their analytics team built internal AI tools that, according to their analytics lead, made implementation "five times faster on the implementation side of things."

The business analytics group expanded beyond on-ice performance to cover TV ratings, merchandise, and food and beverage — treating the franchise as a data operation front to back. As Bowman put it: "At the beginning of my career no team had anything like that, and now I can't imagine operating without them. They are our insight center."

SAP on the Bench

The SAP-NHL Coaching Insights app, built for iPad, gives coaches real-time access to tracking and statistical data during games. Enhancements to the platform were shown at the NHL's 2023 Tech Showcase in Seattle, alongside SAP Venue Metrics for operational analytics.

Getting data into a coach's hands during a game used to mean paper printouts between periods. Now it's a tablet refreshing on the bench. The shift sounds incremental. In practice, it means in-game adjustments driven by live data rather than memory of the last two periods.

What's Still Missing

AI has made the NHL faster at measuring what's already happening. What it hasn't fully solved: predicting individual player development curves with accuracy, modeling player chemistry before putting two players on a line together, and injury prevention beyond broad workload monitoring.

Those are the next problems. The infrastructure — sensors in every rink, a real-time data pipeline, ML models running at game speed — is already in place. The question for the next two seasons is which organizations build proprietary layers on top of the league's shared baseline. The ones that do will have a real edge on teams still treating AI as a broadcast storytelling tool.

The data is already on the ice. The gap now is who uses it.