05Family 04 of 05

The hard parts

Six problems decide whether a design survives a real match. None of them is about the model. All of them are about what a value in the data is allowed to mean.

Your question The study that answers it Two clocks, one match? One clock for the match Two cameras, one pitch? Space is not time No player identities? Players without names No ball on the frame? When the ball goes missing In play, or stopped? Alive or dead, then who A touch, or a turnover? Events are not possession
The decision map of this family. Each question leads to the study that answers it.
  • 6studies in this family
  • 2about joining a second camera
  • 4about reading one tracking feed

One passage of play, six strikes

The truth of this passage is simple: home, a stoppage, home. Each hard part damages a naive reading in its own way, and each is answered by one rule.

Six strikes in turn: clock drift, a change of view, a slot swap, a lost ball, a stoppage, a tag. The naive reading ends with nine possessions. The robust one ends with two, which is the truth.
Strike The trap Study
1 One offset for a whole broadcast One clock for the match
2 Trusting geometry without timing Space is not time
3 Reading a slot change as an event Players without names
4 Missing ball means dead ball When the ball goes missing
5 One mechanism for both questions Alive or dead, then who
6 A touch is a turnover Events are not possession

Where each one bites

The six do not enter at the same place. Knowing the stage tells you where the fix belongs, and which of them a design can ignore.

Clocks and cameras exist only when a second camera is joined. The other four are present in a single tracking feed. Whatever the stage, the damage arrives at the last one as a split or a merged possession.

How an architect chooses here

  • Fix the representation before the model. Every trap above is a wrong meaning given to a value.
  • Say what silence means. Missing, unobserved and unknown are three states. Collapsing them is a design decision, usually a bad one.
  • Test with the failure, not around it. Hide the ball, shuffle the slots, shift the clock, then score.

What this family cannot do

It cannot tell you that the answers work. That takes whole matches the design has never seen, and it is the business of proof.

In this family6

  • Every scored frame is a tactical frame. Any other source must be mapped onto that clock before it can testify, and a replay makes the mapping run backwards. Which evidence can f...

    • Synchronisation
    • Broadcast
    • Replays
  • Calibration says where a pixel lands on the pitch in one frame. It says nothing about when. Which calibration to trust, where it stops being true, and why geometry and timing ar...

    • Homography
    • Calibration
    • Multi-camera
  • Tracking slots are not players. A slot can change bodies mid-move and no football has happened. Design around identity, or repair it. Never trust it by default.

    • Tracking slots
    • Set features
    • Re-identification
  • What should a design believe on a frame with no ball? Three treatments of the blank frame, and two schools of evidence: measure the ball when it is there, or infer possession fr...

    • Missing ball
    • Ball-aware
    • Ball-free
    • Fielded
  • Whether the ball is in play and which team controls it are two different football facts. Designs that answer them with one mechanism let each corrupt the other.

    • Dead ball
    • Restarts
    • Two questions
  • Events and tags are sparse moments. Possession is a state on every frame, including all the frames between them. What a touch, a header or a tagged phase is allowed to change, a...

    • Events
    • Tactical tags
    • Sparse to dense
    • Fielded