Platform
Arrogant Wood is not a stats dashboard bolted onto a video player. The video is the primary record; everything else is indexed against it.
Up to eight synchronised feeds per game from arena fixed cameras, broadcast output or handheld uploads. Automatic sync by audio and scoreboard clock.
Game → period → shift → event → clip. Every object keeps a pointer back to the exact frames it came from, at the exact frame rate.
Player and official detection, jersey number recognition, pose estimation and puck tracking, reconciled across cameras into a single coordinate space.
Clips render from the master into the requested preset. Median render time is nine seconds for a twenty-second clip at 1080p.
Fourteen days at full quality, then tiered to object storage. Cold retrieval takes about forty seconds and costs a fraction of hot storage.
Per-team, per-season and per-folder permissions. SSO with SAML for league deployments; every download is logged.
After processing, one game is roughly 90,000 tracked frames, 1,400 events and 260 shifts — all addressable by time.
| Object | Volume per game | What it holds |
|---|---|---|
| Frames | ~90,000 | 25 fps tracking, 60 fps master video |
| Player tracks | ~38 | Coordinates, pose, jersey, team |
| Shifts | ~260 | Line changes reconciled with the clock |
| Events | ~1,400 | Entries, exits, shots, passes, hits |
| Clips | on demand | Rendered from master, never re-encoded twice |
Import tracking exports, push clips to your video coach, or pull everything through the API.
CSV and JSON event import.
Shift and event reconciliation.
Clip export and share links.
GPS and load data overlay.
Read how the tracking and transcoding layers are built, and what they cost to run.