SPATIAL INTELLIGENCE FROM SPORTS VIDEO

Make the arena measurable.

ArenaMetric turns broadcast video and positional feeds into continuous player and ball tracking, spatial features, and machine-ready streams for pricing, prediction, performance, and media products.

See it in motion Explore the pipeline
Video-nativeUse standard broadcast footage
ContinuousTrack state between events
ReusableOne core, many products
LIVE / SOCCER STREAMING
PRESSURE0.78
SEPARATION8.4m
STATEOBSERVED
VIDEO → POSITION → METRICFRAME 01842
THE MISSING LAYER

Event feeds tell you what happened.
ArenaMetric shows you what is happening.

Score, possession, and play-by-play describe events after they occur. Spatial data captures the continuous context around those events: where every player is, where the ball is, how fast they are moving, and how those relationships are changing in real time.

01

Event data

A pass completed. A shot taken. A goal scored.

10:41SHOTRecorded
10:39PASSRecorded
10:34POSSESSIONChanged
02

Spatial intelligence

Pressure rising. Space opening. Matchups changing.

PLAYERx / y / v BALLx / y / v STATEconfidence FEATURESderived
WHO IT IS FOR

Built for operators, teams, and media groups that need more than a box score.

The underlying value is the same across buyers: turn raw video or positional feeds into a reliable, reusable spatial layer that can feed live models and long-horizon analysis alike.

01

Sportsbooks & exchanges

Improve live repricing, richer props, higher-confidence micro-markets, and downstream risk controls.

02

Teams & analysts

Generate searchable spatial datasets for tactical review, matchup study, and historical player analysis.

03

Media & fan products

Power overlays, minimaps, second-screen experiences, automated stats, and dynamic content.

SEE THE SPATIAL LAYER

One game. Two views.

The same live sequence becomes a normalized field model: detected players, persistent identities, team assignments, ball position, and movement over time.

REAL-TIME VISUALIZER
Broadcast frame → player detection → persistent tracking → field-space coordinates
FROM PIXELS TO METRICS

A reusable spatial intelligence pipeline.

ArenaMetric separates the input source from the analysis layer. Video and positional feeds converge on the same normalized representation so downstream systems consume one consistent shape.

Inputs
▶
INPUT ALive or archived videoBroadcast, OTT, replay, venue capture
⌁
INPUT BPositional feedLicensed coordinates or tracking stream
ArenaMetric core
PIPELINEDetect → Track → Map → Filter → AnalyzeOne coordinate shape. One feature layer. Confidence-aware state.
Outputs
STREAMPositions + confidenceCoordinates, identities, velocities
FEATURESDerived metrics + eventsPressure, assignments, zones, game state
DATASETReplay + historical analysisBacktests, training sets, audit history
01

Detect

Identify players, officials, the ball, and field reference points frame by frame.

02

Track

Maintain identity through motion, occlusion, zoom changes, and camera cuts.

03

Map

Project pixel-space detections onto standardized field coordinates as the camera moves.

04

Filter

Smooth trajectories, bridge gaps, and carry confidence-aware projected state when needed.

05

Analyze

Derive sport-specific features: pressure, separation, assignments, routes, zones, and game state.

06

Stream

Publish continuous positions, velocities, identities, confidence, and features to downstream systems.

MACHINE-READY OUTPUT

Not another dashboard.
A data layer.

ArenaMetric is designed to sit upstream of decision systems. It outputs structured spatial state and higher-order features that can be streamed live, stored historically, replayed, and backtested.

  • Field-space and camera-space coordinates
  • Velocity and team assignment
  • Ball position and field keypoints
  • Per-detection confidence
  • Observed, projected, and recovered tracking state
  • Sport-specific derived events and spatial metrics
spatial-state.json
{
  "timestamp": "2026-09-13T19:22:14.284Z",
  "frame": 1842,
  "players": [
    {
      "id": "A_17",
      "team": "A",
      "position": { "x": 72.4, "y": 31.8 },
      "velocity_mps": 6.1,
      "confidence": 0.97,
      "state": "observed"
    }
  ],
  "features": {
    "pressure_index": 0.78,
    "nearest_defender_m": 8.4
  }
}

Illustrative schema

ONE LAYER, MANY PRODUCTS

Spatial context changes what you can build.

Once player and ball movement becomes a live data stream, the same intelligence can support multiple products without rebuilding the acquisition layer.

01

Live pricing & risk

Use formations, matchups, separation, pressure, pace, and field state as model inputs. Reprice with context instead of waiting for the next event.

LIVE MODELS → PRICING → MARKETS
02

Prediction markets

Support tighter two-way markets with continuous probability inputs, timestamped state, confidence, and replayable history.

03

Team & scouting analytics

Turn game video into searchable spatial datasets for tactical review, matchup analysis, and longitudinal player evaluation.

04

Broadcast & media

Power live minimaps, overlays, second-screen experiences, automated stats, and context-aware highlights.

05

Data licensing

Create a structured feed for competitions where official tracking is unavailable, thin, or uneconomical.

06

Historical research

Replay archived footage through the same pipeline to generate training sets, backtests, and comparable historical features.

SPORT-SPECIFIC FEATURES. SHARED CORE.

One infrastructure layer. Different vocabularies.

Tracking, streaming, trajectory filtering, replay, and evaluation stay reusable. Sport-specific models and features plug into the same core.

DEMONSTRATED END TO END

Soccer

Continuous, low-event play makes soccer a demanding spatial problem. ArenaMetric tracks the state between events—the part a conventional event feed does not describe.

Possession progressionPressureZone occupancyPlayer distance & speed
Broadcast soccer footage paired with a 2D player and ball tracking visualization
SPATIAL MODELFrame-aligned player + ball state
BUILT FOR REAL VIDEO

Production conditions are messy.
The data should say so.

Broadcast feeds drop. Cameras cut. Players overlap. Weather and lighting change. ArenaMetric keeps state explicit so downstream systems know the difference between a direct observation and a projection.

◎

Confidence-aware

Every detection can carry confidence and tracking state for downstream guardrails.

↻

Graceful recovery

Trajectory projection bridges short gaps and identity can recover after occlusion or cuts.

▦

Replayable

Process live streams or archived games with the same analysis path for testing and research.

⇶

Horizontally scalable

Scale concurrency by adding compute rather than adding operators game by game.

DEPLOYMENT

Meet the data where it lives.

Containerized processing can run on cloud GPU infrastructure or inside an existing media environment. Live and archived streams feed the same path; structured output can integrate with distributed event streaming, APIs, model pipelines, and data lakes.

VIDEORTSP-equivalent
COMPUTEGPU-accelerated
MODELive + historical
OUTPUTStreams + datasets
THE ARENA IS ALREADY ON VIDEO.

Make it measurable.

Tell us what you are building, the sport or data environment you are working with, and where spatial intelligence could fit.