Data

FusionFeed provides comprehensive coverage for foundational game data. Unless otherwise mentioned, all data is made available in realtime at low latency during the game.

FusionFeed data is ingested from the truest and lowest latency source possible, usually on-site. Data that pre-dates FusionFeed's operation within a given league is typically unavailable. NFL coverage for example begin in 2021, and data from earlier seasons is not available.

All non-static data (including video) is precisely stamped with wall-clock times so that it can be synchronized with any other data from FusionFeed or from third parties providing accurately timestamped data.

Audio/Video (AV)

FusionFeed provides AV streams from:

• TV network Cameras - We collect video by physically connecting to the raw camera SDI feeds on-site, before any quality is lost to encoding. The number of cameras we ingest varies by game. For a typical NFL game, we collect around 20 camera feeds, including:
  • Handhelds
  • Main 50
  • Left, main, and right 20
  • Endzones
  • Slashes
  • Beauty cams
  • Cable cams
  • Goal posts
  • All 22/29
• Team Cameras - We collect video from team cameras in a similar manner, by physically connecting on-site.
• Program Feeds - We provide "dirty" program feeds by physically integrating with the on-site OB truck.
• 8K Cameras - We set up fixed 8K cameras around the stadium to capture all the action that's not caught by TV cameras.
• FusionCams - These are virtual cameras that follow a specific subject, region, or group of subjects within the field of view of an 8K camera. Users can create FusionCams for use-cases such as isolating individual players or as an alternative to the traditional all 22 cameras.

Video is made available through FusionFeed via the HLS protocol. HLS video can be distributed at a global scale directly from FusionFeed with approximately 10 seconds of end-to-end latency.

On request, SRT feeds can be provided for video for viewing with sub-second latency from the camera glass to the client's screen.

Select AV streams are processed using state-of-the-art computer vision to provide:

• Bounding boxes
• Skeletal pose estimation
• Masks
• Camera calibration

See GraphQL Video for details on how to get started working with video using FusionFeed's GraphQL API.

Telemetry

FusionFeed provides X/Y coordinates for players, officials, balls, and other subjects.

This data may be based on hardware tracking devices or computer vision (CV) depending on the game. When there are hardware tracking devices worn, tracking data is typically collected in the stadium, as close to the source as possible. Otherwise, if players are not wearing hardware tracking devices or in cases where redundancy is desired, optical CV-based tracking based on our 8K video feeds is used.

Telemetry data is provided live, typically at under 3 seconds of latency and is commonly used for use-cases such as:

• Determining which players are on the field or bench for any given play, time, etc. I.e. determining "player participation".
• Analyzing player performance throughout the game.
• Creating 2D or 3D visualizations or game reconstructions.
• Deriving additional data points such as formations.

Scoreboard

FusionFeed provides data directly from the stadium scoreboard, typically at subsecond latency. Depending on the venue, this may come from a direct physical connection, e.g. to a Daktronics or OES serial port, or from a distribution server.

Statistics

FusionFeed provides detailed timelines and statistics for all covered games, live, in realtime. The source of these statistics varies by league.

For leagues with the notion of "official" statistics, FusionFeed collects official stats directly from the source and provides those by default, as soon as they're available. For the NFL for example, official play-by-play and cumulative statistics are typically updated immediately at the end of each play.

Alternatively or in addition to official statistics, FusionFeed can provide AI-based automated statistics. Our AI-based solution takes all available game data such as tracking and scoreboard data and builds a comprehensive understanding of the game live. Automated stats can provide superhuman speed, objectivity, and precision. They can be used to calculate statistics for plays as the plays are still unfolding or can be used to identify the precise time and location of any sub-play event.

Often the best solution for clients is to utilize a combination of official and automated stats.

Game Metadata

Game metadata, such as schedules and rosters are made available in FusionFeed as soon as they're available from the official source, and changes are propagated through FusionFeed as they're made in realtime.

---