ScoutStream AI
Product

An objective score, computed from film nobody had time to watch.

ScoutStream AI is a perception pipeline and an evaluation model. The pipeline reads every athlete in the frame. The model turns that into a number your staff can defend, a development map the athlete can use, and a load picture your medical staff can act on.

01 / Scoring

The Pro-Ready Score v2.0

One 0–100 figure, built from five weighted pillars. Every pillar is decomposable — you can always see which inputs moved the number.

Tactical Intelligence
25%
  • Pitch-space positioning
  • Off-ball movement
  • Defensive line discipline
  • Scanning frequency from head-turn
Technical Execution
25%
  • Action success rate under pressure
  • First-touch retention
  • Both-foot usage
Athletic Output
20%
  • Total distance
  • Top speed
  • Accelerations / decelerations over 3 m/s²
  • High-intensity distance
Physiological Response
15%
  • HR recovery between efforts
  • Intensity distribution
  • Workload relative to minutes played
Biomechanical Efficiency & Durability
15%
  • Gait symmetry
  • Knee valgus on cut and land
  • Deceleration mechanics
  • Fatigue-related form decay
02 / Fairness

Age normalization

Raw output is normalized against biological maturation, not just chronological age. Two athletes born in the same month can be three years apart in maturity status, and the more mature one will win almost every raw physical metric on the sheet.

Comparing a 14-year-old's numbers to a 19-year-old's — or to an early-maturing peer in the same birth year — systematically over-ranks early developers. This is the #1 failure mode in youth talent ID: programs select for current size and speed, then spend years wondering why the class didn't convert.

Normalizing against maturation status separates "good now" from "good for where they are". The late developer stops being invisible.

Callout
The Pro-Ready Score reports both a raw pillar output and a maturation-normalized output. Scouts see the gap between them. A large positive gap is the strongest single signal we produce — it means the athlete is competing at this level while still physically behind the group.
03 / Discovery

Talent Similarity Index

Every athlete-performance produces a movement embedding — a dense vector encoding how that athlete moves, positions, accelerates, and executes, independent of the scoreline or the level of opposition.

Search the index for athletes whose movement signature matches a known profile at the same age. Give it a player you already rate and it returns the closest matches across every game the system has ever processed.

This is the hidden-talent engine. It surfaces athletes nobody is watching because they play for an unfashionable program in a market no scout flies to — not because they lack the profile.

Query
similar_to(athlete_id=1184, age_band="U16", k=25)
#14 · U16 · Midfield0.94
#7 · U16 · Wide0.91
#22 · U15 · Midfield0.89

Cosine distance over movement embeddings, restricted to the same age band and maturation window.

04 / Durability

Biomechanical Load & Asymmetry Module

3D pose estimation gives you asymmetry, landing mechanics, and deceleration load with no sensors on the athlete — nothing to charge, distribute, collect, or lose, and nothing that changes how a 14-year-old plays.

  • Left/right gait and stride symmetry, tracked across the season
  • Knee valgus angle on cutting and landing events
  • Deceleration mechanics and braking load per exposure
  • Second-half form decay versus that athlete's own first-half baseline

Every flag is relative to the athlete's own baseline first, and their age-group distribution second. That keeps the output useful for a staff that knows the athlete better than any model does.

Important
Decision support for your medical and performance staff. Not a medical device, not a diagnosis, and not an injury prediction.

See the score computed on your own team.

Discovery call, then a three-game pilot on your footage. You compare our rankings against your staff's.

Book a Discovery CallSee pricing