Vision pass · activeVisual strategy · Match context

BlueX product · Visual Strategy AI

Visual strategy, directly on the video.

BlueX combines AI video editing and specialised off-ball movement analysis so football staff can see, review, and share strategy without translating raw coordinates.

BlueX visual strategy AI placing tactical movement directly on football footage

Visual Strategy AI

The analysis appears where the moment happened.

Horus intelligence system

Coordinates matter. Context changes the game.

A tracking export can stop at x, y, and z. BlueX starts there, then connects every observation to identity, role, phase, action, value, and source evidence so strategy teams never have to read coordinates.

Football match with player, ball, and pitch detection translated into tactical context

Detection before interpretation

Your football staff should never have to read x, y, or z.

HorusNet decodes once and fans the same frames into specialist heads. BlueX turns the resulting coordinates into searchable football context before any tactical claim is made.

Semantic depth

Not a coordinate warehouse. A queryable model of the match.

01

Coordinates

x / y / z / time

Image and pitch positions remain the immutable observation layer, not the final product.

02

Identity

player / team / role

The observation becomes a known actor with roster context, role, and confidence.

03

Football context

phase / zone / action

Possession, pitch zone, pressure, movement, action, and outcome make the row football-readable.

04

Decision evidence

value / risk / cause

Game models connect the moment to danger, contribution, sequence, and reviewable causal evidence.

Pitch geometry

Horus Field

Detects pitch lines and camera geometry so every observation can share one match coordinate system.

Player identity

H-Oracle View

Connects a track to the roster while ambiguous identity remains explicitly unresolved for review.

Player observation

Horus Player

Extracts player, goalkeeper, and referee positions as synchronized frame-level evidence.

Ball observation

HorusOrb

Finds the small, fast ball and preserves x, y, height, and trajectory evidence across the match.

High-speed football detection with motion trails, player boxes, and pitch geometry

Processing target

Minutes

Full-match context is designed to be ready in minutes, with the unified processing pipeline targeting a 20-minute full-match run.

Dozens of football analysts manually reviewing and tagging match video at workstations

Operational shift

Free the team from manual video-tagging labor.

Parallel processing supports up to five video angles. Analysts review football evidence instead of cutting footage, reconciling spreadsheets, or redefining events from scratch.

Runtime is a processing target, not a guaranteed SLA. It varies by source length, view count, and enabled model heads.

100% timeline
Full-match scope, not sampled highlights.
Up to 5 angles
One synchronized, parallel processing run.
Ask the match
Find what happened, when, and why.

Automated delivery

From match to report. One pipeline.

Horus carries the same evidence from source video through detection, semantic context, natural-language retrieval, and coach-ready delivery. No manual tagging handoff is required between stages.

  1. 01

    Match

    Upload the full match from one to five synchronized video angles.

  2. 02

    Detect

    Pitch, player, identity, ball, and time stay aligned in one processing run.

  3. 03

    Context

    Coordinates become phases, actions, zones, roles, pressure, value, and evidence.

  4. 04

    Ask

    Ask what happened and when. Horus searches the whole match, not a sampled clip set.

  5. 05

    Report

    Return time-linked clips, visual strategy, and a reviewable coach-ready report.

AI-native football database

Strategy teams do not read coordinates. They query the match.

Raw observations stay immutable while semantic and reasoning layers remain recomputable. Confidence, taxonomy, model version, evidence, and review state travel with every output so correction improves the system instead of overwriting history.

Football space creation analysis showing a decoy run opening a central pocket

01 · Space creation

Who moved the defense and what opened behind them?

Connect the decoy run, defender displacement, and the pocket created for the next action.

Live semantic analysis

Auto · 5.2s

Showing 01 · Space creation: Who moved the defense and what opened behind them?

Match structure

Shape, pressure, and phase

H-Block reads defensive structure and pressure. H-Phase and H-Phase-Share turn the timeline into comparable phases of play.

H-Block · 89.5%H-Phase · 83.4 / 82.7%H-Phase-Share · 93.4%

Value and risk

What the movement was worth

The system measures the probability and field value around a moment instead of stopping at a detected event.

HxG · 0.754 AUCHxDanger · 0.753 AUCHxT · 0.791 AUC

Specialist context

Game states need specialist models

Horus-GK evaluates goal-kick outcomes, while H-Action and Horus Calib remain clearly identified research and synchronization layers.

Horus-GK · 0.821 reach AUCH-Action · research seatHorus Calib · sync layer

Football intelligence outputs

What staff receives, not another coordinate export.

The product returns searchable match evidence, off-ball and spatial metrics, and game-model outputs in language analysts, coaches, and directors can use immediately.

Full-match retrieval

Ask for the moment. Get the evidence.

Search the full timeline by player, phase, action, outcome, or tactical question and return time-linked evidence without manual scrubbing.

PossessionsPhases of playActionsSequencesEvidence clips

Off-ball intelligence

See what happens off the ball.

Connect support runs, player orientation, field of view, focus of attention, pressure, zones, and team shape to the visible action.

Support runsFOV / FOAPressureZonesTeam shape

Game-model outputs

Move from location to consequence.

Translate standardized match context into structure, phase, danger, value, and specialist restart intelligence for decision-makers.

H-BlockH-PhaseHxGHxDangerHxTHorus-GK

Applied computer vision

The vision work behind the service.

BlueX also maintains applied vision capabilities that support movement reconstruction and specialised measurement workflows.

BlueX pilot

Bring one match. See the game beyond the ball.

We will prepare a focused walkthrough using the workflows that matter to your staff.

Cross-sport vision

From the pitch to the court.

BlueX adapts its visual strategy approach to racket sports. The method is designed to extend from football's pitch and phases of play to the strokes, rallies, and court positions of tennis and badminton.

  1. Smaller target. Faster exchange.

    A tennis ball or badminton shuttle is smaller and can move faster. The visual read has to follow a tighter, quicker path.

  2. From field to court.

    Pitch geometry becomes court lines, service boxes, lanes, and zones that keep each exchange in context.

  3. From 22 players to 2 or 4.

    The view narrows from a full team to singles and doubles, following the players and the space their exchange creates.

  4. From pitch analysis to strokes and rallies.

    Football's phases and actions become stroke sequences, rally patterns, positioning, and point construction.

Tennis player preparing a stroke on a hard court

Tennis

Read the point at stroke speed.

Designed to extend into serves, returns, rallies, recovery, and court position.

Badminton players contesting a fast rally on an indoor court

Badminton

Follow the shuttle through the rally.

Designed to extend into shuttle movement, recovery steps, rally shape, and the space around each shot.

Racket sports

The same vision system, adapted to the sport in front of it.

Explore racket-sports match analysis, or talk with Baceline about the tournament operations around it.