What This Is
The Field Intelligence System is a self-contained, field-deployable platform designed to mount on a 15-foot pole at the midfield touchline. It runs off a 12V battery, brings its own WiFi network, and manages its own 5G uploads. The operator sets it up, runs a short calibration sequence, and starts the match. Everything from live streaming to post-game delivery is automated from that point.
FIS operates in two modes. On game day it streams live, tracks players and ball, captures broadcast-quality audio, and delivers stats and highlights automatically. At practice it detects drills, tracks load on both teams, surfaces real-time alerts to the coaching tablet, and produces session reports for coaches and players. Across sessions it builds a longitudinal development record for every player - weekly and monthly reports, seasonal summaries, and a program-level view for directors with placement and development decisions across all sites and skill bands.
Live
📡 Sideline Streaming
Parents on-site connect to the field WiFi and watch live video in a browser - no app required. Remote family gets a 5G stream link before kickoff.
Analytics
📊 Real-Time Tracking
Two 4K cameras - one per half - and two Hailo-8 accelerators track players and ball across the full one half of the pitch. ByteTrack and homography map positions to real-world field coordinates. Players are identified by name from kickoff - no post-game ID work needed.
Sensors
🦵 Bilateral Foot Sensors
FIS-native IMU sensors strap to each cleat - one per foot. They stream live kick, sprint, turn, and load data to the GEM12 MAX during the game, and log 1000 Hz raw data for deep post-game analysis. NFC check-in binds each sensor pair to a player by name before kickoff.
Audio
🎙️ Spatial Audio
Two RØDE mics deliver broadcast-quality stereo: field sounds left, crowd right. A Movo DoubleMic-UC adds per-team crowd energy as a goal confidence signal.
Tactics
🧩 Tactical Intelligence
FIS recognises named multi-player sequences as they develop across a game - wide carry and cross, third man runs, counter attacks, switches of play. After the game it tells you which plays produced goals, which ones produced turnovers, and which patterns opponents ran against you. Plays spread through competitive ecosystems fast. Having a name for what you saw is the first step to preparing for it.
Practice
🏃 Practice Mode
Same hardware, different session type. Drill detection, rep tracking, bilateral load alerts, and role assignment run automatically. A coaching assistant tablet confirms what the system sees. Practice reports for coaches and players are delivered before the drive home.
Development
📈 Longitudinal Reporting
Weekly and monthly player development reports track each individual's arc - kick velocity, decision speed, foot balance, physical load. Coaching insights digests surface which drill designs are producing results. A program director view spans all sites and skill bands, bringing consistent longitudinal data to placement decisions across every team.
AI
🧠 Edge AI Processing
All inference runs on dedicated AI silicon at the field — no footage sent to the cloud, no processing lag, no per-game compute fees. The same hardware that tracks every player and detects every kick also encodes and streams live video, all from a 12V battery.
Training
🔬 Enhanced Training Instrumentation
At practice, additional sensors beyond the standard shoe pair can be deployed — vest-mounted IMUs, positioning anchors, and an instrumented ball. Three sensors per player capture full-body movement coordination: not just what the feet do, but whether the body and feet are working together correctly on every touch, turn, and dribble. More instrumentation means more detailed session metrics and richer player development reports. See Section 07 for the full instrumentation framework.
Why Sensors Matter
Cameras see where players are. Sensors see what their feet are doing. The two data sources answer different questions, and some of the most useful coaching signals only come from the sensor layer.
What cameras cannot see
Vision tracks position and movement across the pitch. It cannot confirm which foot touched the ball, measure the force of a kick, detect a gait imbalance building through the second half, or flag the moment a player starts favouring one side. Those signals require contact — a sensor strapped to each cleat.
Live vs. post-game
Third-party sensors sync post-game over Bluetooth, which means their data only reaches the coach hours after the match. FIS-native sensors transmit raw IMU live. Bilateral load drift, sprint counts, and foot preference data are on the coaching tablet during the game — actionable while there is still time to act.
Identity confidence
Sensor-equipped players stay correctly identified throughout the match — even through tight clusters, tackles, and crowded penalty areas where vision alone struggles. The sensor layer and the vision layer reinforce each other continuously, so every player stays named from first whistle to last.
| FIS output | Without sensors | With FIS sensors |
|---|---|---|
| Kick and touch detection | Vision proximity estimate — misses touches in crowds and through-balls | Confirmed foot contact with which foot, real-time, every player |
| Preferred foot ratio | Not available | Live left/right kick count per player, updated each touch |
| Sprint and decel count | Frame-rate-limited vision estimate, error-prone | IMU-derived, available in Pass 1 during the game |
| Turn detection | On-ball turns guessed from trajectory change; off-ball turns not tracked | All turns detected via gyro; plant foot identified on each cut |
| Gait symmetry | Not available | Stance and swing time per foot — asymmetry between left and right signals compensation or fatigue |
| Bilateral load drift | Not available | Left/right load ratio tracked across the match — drift from 50/50 toward one side is a live fatigue and injury-risk signal |
| Goal confidence | Audio and vision only | Multiple independent signals — significantly higher confidence, fewer missed events and false positives |
What FIS Measures
FIS produces two classes of data: vision-derived metrics available for every tracked player, and sensor-derived metrics for players wearing FIS IMU sensors. Vision metrics cover distance, territory, touch count, possession, and shot map. Sensor metrics add kick velocity and foot preference, bilateral load, gait symmetry, sprint and decel count, turn quality, decision speed, and how a player’s choices and technique change when they’re under defensive pressure.
A third class sits above the other two: team-level metrics built by combining data across every tracked player at once. Passing networks, pressing and closing speed, team shape and compactness, and duel win-rates all come from looking at the full squad together rather than one player at a time.
Full metric definitions and report schemas are in the technical documentation.
How FIS Compares
FIS shares a category with AI tracking cameras and GPS wearables. The table below covers the systems most likely to come up in conversation. The key differentiators are data ownership, operating cost, and whether sensors feed the live coaching dashboard or only post-game analysis.
| Feature | FIS | Trace / TraceCam 2 | Pixellot Air NXT | I-TRX | Playermaker |
|---|---|---|---|---|---|
| Tracking method | Vision — full field + bilateral foot sensors | Vision — AI pan/zoom camera | Fixed panoramic stitch | Wearable sensor + follow camera | Bilateral foot sensors only |
| Live sensor data | Yes — coach tablet during the game | No | No | No | No — post-game sync only |
| Bilateral foot tracking | Yes — one sensor per foot | No | No | Single sensor only | Yes — one sensor per foot |
| Live sideline stream | Yes — local WiFi + 5G, no app | Yes — cloud delivery | Yes — paid tier | No | No |
| Automated highlights | Yes — audio + vision + sensor | Yes — post-game AI | Yes — AI auto-clips | Buffer only | No |
| Subscription cost | <$5/month | Required SaaS | Required SaaS | None | Required SaaS |
| Data ownership | Full local | Vendor cloud | Vendor cloud | Device only | Vendor cloud |
| Player development reports | Yes — weekly, monthly, seasonal | Per-game only | Per-game only | No | Yes |
| Practice mode | Yes — drill detection, rep tracking, load alerts | No | No | No | Yes |
Design Principles
Capex over recurring cost
No cloud inference, no SaaS subscriptions. Models run locally. Uploads go to services we control. Target operating cost: <$5/month.
Human review, not auto-publish
Whistle detection and goal confidence scoring produce review queues. Humans decide what gets published. The system saves time; it does not replace judgment.
Self-comparison only
Player development reports compare each player only to their own prior performance. Rankings and peer comparisons are the coach's job - the system provides the data, not the verdict.
Data owned by the program
All footage, sensor logs, and derived metrics are stored on operator-controlled infrastructure. No third-party platform owns or can delete the data. The development record belongs to the program, not to any vendor.
System Requirements
FIS is designed for single-operator deployment at any properly lined outdoor pitch — no mains power, no venue infrastructure, no technical crew. Battery powered, self-contained networking, under 15 minutes to set up. Full technical requirements are documented in the section index.