Observed is not predicted
Keep measured positions separate from a coaching reference and from a modelled alternative.
Our evidence standard
Confidence
has to be earned.
Hockey is fast. Cameras miss things. Models make mistakes. Useful intelligence makes those limits inspectable.
Coverage is part of the result
A camera angle or unidentified player can limit the analysis. Missing evidence should not become a confident grade.
Source before summary
A claim should point back to the film, method and relevant coaching context.
The coach confirms the system
An extracted rule is a draft. The coach’s confirmation makes it an agreed reference.
Review changes the answer
Corrections are useful input. The workflow should preserve what changed and why.
People own the judgment
Analysis supports coaching and development conversations. It does not replace the people responsible for them.
Development status / September 2026
What’s real.
What’s next.
Real-film processing and reports
The engine has processed real broadcast and automated-camera games. Rink calibration, positional metrics and system-rule evaluation are being tested. Identity coverage varies substantially by camera and game.
Film review, reference views and Sitdown
Product workflows include demonstration data. A prototype screen is not evidence of production accuracy or general availability.
What we are building toward
Reliable cross-game identity, validated player comparisons, outcome-based positional value, wider deployment and real-time bench assistance require further development. Real-time coaching alerts are not a current service.
We agree pilot scope, footage suitability and data handling before work begins. Ask for a walkthrough of current results and limitations.
The next shift starts here
See more.
Develop better.
Bring your questions. We’ll bring the film.
Explore RynkIQ with the people building it.