Soar | Signal

01 - The Gap

Thousands of Actions. None Captured.

A single MMA bout generates 300+ strikes. A tennis match produces 600+ shots. A basketball game has 2,500+ trackable events. Capturing this data today requires human annotators watching footage frame by frame. It doesn't scale.

LIVE EVENT STREAM

0 EVENTS/SEC

MULTI-SPORT | REAL-TIME CLASSIFICATION

TOTAL: 0

300+

Strikes Per MMA Bout

600+

Shots Per Tennis Match

2,500+

Events Per Basketball Game

0%

Automated at Venue Level

A trained analyst can tag 10-15 events per minute from video playback. A single basketball game takes 3+ hours of annotation. Multiply by every game, every practice, every venue. The economics don't work. The data stays trapped in video files.

02 - Capture

Venue-Side Camera Arrays

Calibrated cameras mounted around the venue with perfectly synchronized capture. Athletes don't wear anything. The system sees everything from multiple angles and pinpoints real 3D positions.

VENUE: TRACKING ARENA

TRACKING

8x CAMERAS | MULTI-VIEW TRIANGULATION

ATHLETES: 0

0

Wearables Required

8+

Cameras Per Venue

51

Frames Per Second

±2cm

3D Position Accuracy

03 - Understand

From Pixels to Structured Events

AI identifies every athlete and their movements. Sport-specific models classify every action. A cross becomes a structured record with timing, speed, target zone, and 3D position. Real-time extraction, not video annotation.

EXTRACTION: 3D SKELETON

ACTION: CLASSIFYING

17 KEYPOINTS | MULTI-VIEW TRIANGULATED

VELOCITY: 0.0 m/s

17

3D Keypoints Per Athlete

<50ms

Classification Latency

20+

Action Types Per Sport

0.90+

Confidence Threshold

BODY TRACKING

Real-time AI processing on venue hardware. The system automatically detects every athlete and maps their body position. Multiple camera views work together to pinpoint exact 3D location with centimeter accuracy.

ACTION CLASSIFICATION

Body movement patterns feed sport-specific classifiers. A jab is recognized by arm and shoulder extension at speed. A forehand is identified by hip rotation driving racquet-arm acceleration. Raw motion data, not video, powers the analysis.

04 - Every Sport

One Platform. Any Sport.

The same camera hardware and pose estimation pipeline. Sport-specific action models plug in on top. Each sport generates its own event ontology - the actions, metrics, and spatial data unique to that game.

COMBAT SPORTS

MMA / Boxing / Muay Thai

Strike events: type, limb, target zone, velocity

Landed / blocked / evaded classification

Clinch entries, takedowns, knockdowns

Fighter distance, stance, ring position

Round-by-round breakdown

300+ classified actions per bout

RACQUET SPORTS

Tennis / Pickleball / Padel

Shot events: type, speed, spin, placement

Rally length, shot selection patterns

Court coverage, player positioning

Serve analytics: speed, placement, ace rate

Winner / error classification

600+ tracked shots per match

BASEBALL / SOFTBALL

Hitting, Pitching, Fielding

Swing mechanics: bat speed, attack angle, contact

Exit velocity, launch angle, spray chart

Pitch tracking: speed, movement, location

Fielding routes, throw velocity

At-bat outcomes with full context

280+ pitches analyzed per game

BASKETBALL

Full-Court Player Tracking

Shot events: location, type, make/miss

Player speed, distance, court coverage

Pick-and-roll detection, defensive rotations

Assists, rebounds with spatial context

Possession analysis, transition scoring

2,500+ trackable events per game

GOLF

Swing Analysis / Practice

Club speed, face angle, swing path, attack angle

Ball flight: launch angle, spin rate, carry

Putting: read, speed, line, break

Full kinematic chain decomposition

Practice session aggregation + trends

40+ swings analyzed per round

SOCCER / FOOTBALL

Full-Pitch Player Tracking

Pass events: type, distance, accuracy, target

Shot on goal: speed, placement, xG

Sprint speed, distance covered, heat maps

Tackle, interception, aerial duel outcomes

Formation shape, pressing triggers, offsides

3,000+ trackable events per match

05 - The API

One Endpoint. Every Sport. Real-Time.

REST endpoints for events, athletes, sessions, and spatial data. Real-time streaming via WebSocket or batch export. Your platform consumes the data - we handle the capture and classification.

GET /v1/events?session=s_7f2a&type=strike

[
  {
    "id": "evt_c91a",
    "timestamp": 1711036800.342,
    "type": "strike",
    "subtype": "cross",
    "actor": "ath_01",
    "target_zone": "head",
    "velocity_ms": 12.4,
    "position_3d": [1.23, 1.65, 0.82],
    "landed": true,
    "confidence": 0.94
  }
]

GET /v1/sessions/s_7f2a/summary

{
  "session_id": "s_7f2a",
  "sport": "mma",
  "duration_s": 900,
  "athletes": 2,
  "total_events": 342,
  "breakdown": {
    "strike": 287,
    "takedown": 4,
    "knockdown": 1,
    "clinch": 12
  }
}

REST

+ WebSocket Streaming

<100ms

API Response Latency

JSON

+ Protobuf Support

OAuth 2.0

Authentication

UNIFIED DATA FORMAT

Every event captures the same core information: when it happened, who did it, what action, where it occurred, and confidence level. Sport-specific details extend it — speed for strikes, spin for racquet shots, launch angle for baseball. One integration works everywhere.

06 - Intelligence

From Events to Insights. Automatically.

Every consumer could compute their own stats from raw events. They won't. The analytics engine transforms event streams into live dashboards, session breakdowns, athlete progression, and predictive metrics. No analyst in the loop.

ANALYTICS: LIVE SESSION

ROUND 2 | 2:47

COMPUTING: 342 EVENTS → 47 METRICS

LATENCY: 12ms

GET /v1/analytics/sessions/s_7f2a/live

{
  "round": 2,
  "clock": "2:47",
  "fighters": {
    "ath_01": {
      "total_strikes": 87,
      "accuracy": 0.43,
      "power_shots": 12,
      "fatigue_index": 0.72,
      "zones": {"head": 24, "body": 31, "legs": 32}
    }
  },
  "alerts": [{"type": "fatigue", "delta": -0.34}]
}

GET /v1/analytics/athletes/ath_01/progression

{
  "sessions": 47,
  "trends": {
    "accuracy": {
      "current": 0.43,
      "30d_avg": 0.41,
      "delta": "+4.8%"
    },
    "power": {
      "current": 8.7,
      "30d_avg": 8.2,
      "delta": "+6.1%"
    },
    "volume": {
      "current": 312,
      "30d_avg": 287,
      "delta": "+8.7%"
    }
  }
}

REAL-TIME

Live Analytics Updates

47+

Metrics Per Session

ZERO

Consumer-Side Computation

$0

Additional Hardware Cost

LIVE INTELLIGENCE

During a session, metrics update per-event. Strike rates, accuracy, zone breakdowns, fatigue curves - computed automatically, streamed over WebSocket. Broadcasters embed directly. Sportsbooks feed prop markets. No analyst required.

ATHLETE PROGRESSION

Cross-session trend analysis. Accuracy improving? Power declining? Which techniques are developing? Coaches see this for every athlete at every session. The value compounds with the data - 50 sessions of history is worth more than 5.

07 - Architecture

Edge to Cloud

Processing happens at the venue on local hardware. Structured events flow to the cloud. Consumers query what they need. No video leaves the building — only data.

CAPTURE

Venue Cameras

8 Cameras (Multi-View)

PROCESS

Edge GPU

Pose + Track + Classify

SIGNAL API

Cloud

REST + WebSocket + Storage

CONSUMERS

BROADCASTERSReal-time stats overlay + graphics

COACHING PLATFORMSSession analysis + athlete progression

SPORTSBOOKSLive event feeds + historical datasets

VENUE DASHBOARDSMember analytics + engagement

NO VIDEO LEAVES THE VENUE

Raw camera feeds stay on the edge PC. Only structured event data - JSON payloads averaging 200 bytes per event - flows to the cloud. A full MMA bout generates under 100KB of event data. No video storage costs, no bandwidth constraints, no privacy concerns around raw footage.

OFFLINE CAPABLE

Edge processing runs independently of cloud connectivity. Events are queued locally and batch-uploaded when the connection is available. A venue with intermittent internet still captures every action - the data syncs when it can.

08 - Markets

Four Markets. One Data Source.

The same structured data, consumed differently by each buyer. The hardware install is the cost. The API subscription is the revenue.

VENUES & GYMS

Member Analytics

Session-by-session progression

Technique analysis dashboards

Equipment utilization metrics

Member engagement + retention

The hardware install at the venue. Members get data on every session - punch counts, shot placement, swing mechanics. Venue gets utilization analytics. Monthly subscription per location.

BROADCASTERS & MEDIA

Real-Time Stats

Live stats overlay during broadcast

Post-match graphics + breakdowns

Automated highlight generation

Second-screen data feeds

Per-event licensing. Broadcast gets real-time strike counts, shot speeds, player positioning - formatted for on-screen graphics. Premium tier includes WebSocket streaming for sub-second latency.

COACHING & TRAINING

Athlete Development

Automated session reports

Comparative benchmarking

Technique trend analysis

Platform integration (API)

Per-athlete monthly subscription. Coaches see every session quantified - what changed, what improved, what regressed. Integrates with existing training platforms via API.

SPORTSBOOKS & FANTASY

Verified Live Data

Live event streams, sub-second latency

Prop bet data (strike count, speed)

Historical datasets for modeling

Official data partnerships

Premium tier. Highest margin. Sportsbooks pay for verified, low-latency, structured event data they can't get anywhere else. Prop bets on strike counts, shot speeds, player stats - all backed by automated capture.

09 - The Flywheel

The Data Compounds

More venues generate more data. More data trains better models. Better models increase accuracy. Higher accuracy attracts more consumers. The installed base is the moat.

VENUE NETWORK

VENUES: 0

DATA FLOW

EVENTS: 0

PHASE 1

First Venue

Single sport, single location. 50K events/month. Prove the data quality and API reliability. First paying consumers validate the model.

PHASE 2

10 Venues

Multi-sport coverage. 500K events/month. Cross-venue models improve accuracy for all locations. Coaching and broadcasting customers come online.

PHASE 3

100 Venues

5M events/month. Sport-specific models reach production accuracy. Sportsbook integrations. Historical dataset becomes independently valuable.

PHASE 4

Industry Standard

The default data source. Every venue wants the install because consumers demand the data. Network effects make the dataset unreplicable.

THE MOAT ISN'T THE MODEL

Pose estimation models commoditize. Action classifiers can be replicated. What can't be replicated is cameras installed in hundreds of venues, generating millions of events per month, feeding sport-specific ontologies that no competitor has the training data to match.

PRICE BY DATA, NOT HARDWARE

The camera install is a sunk cost. Revenue comes from the API - per-session for venues, per-event for broadcasters, monthly for coaching platforms, premium for sportsbooks. More usage means more revenue. Incentives align.