FSL Foundations

The behavioural model underneath the language. Read this once — not mid-match.

Vocabulary gives you the words: Settle, Drive, Collapse, and the rest. This page is about where those words come from, and how they fit together underneath the surface.

FSL is primarily a descriptive framework for observing football. It necessarily contains some theoretical assumptions — Identity, Will, and Memory are explanatory claims, not just description — but its immediate purpose is to improve observation, not to provide a complete causal theory of the game.

It's built at a different level of confidence than Vocabulary. The States are close to something a coder could apply to footage — there's a manual behind them, coders scoring the same passage of play and checking whether they agree. Nothing on this page has that yet. What follows is the theory of why the States hang together the way they do — offered in Alexander's spirit, to sharpen attention rather than to prove a mechanism. Where a claim is closer to a guess than a finding, this page says so.

Outcome, Event, Behaviour, State

Before Agency × Clarity, it helps to be explicit about the layer FSL is actually working at, since it's easy to blur with what conventional analytics already measure. Football analysis can be organised into four layers, each one level earlier than the last:

Outcome — goal or miss.
Event — the shot itself, with its location, angle, and probability. This is what xG and most shot-based metrics describe.
Behaviour — the coordinated sequence that produced the shot: the passing, movement, and spacing leading up to it.
State — the underlying behavioural organisation governing that sequence. This is what FSL's vocabulary names.

Two possessions can produce shots of identical xG — the same Event, the same Outcome probability — while differing substantially in the Behaviour and State that produced them: one from sustained positional control, one from a scrambled clearance and a fortunate deflection. Existing metrics are very good at describing Event and Outcome. FSL is an attempt to describe State — one layer earlier than event-based analytics currently reach, and a genuinely different layer from Behaviour itself, since two possessions can share a Behaviour-level description (a give-and-go, a switch of play) while sitting in different underlying States.

This is FSL's core ontology, not something specific to any single Reading — every State, Force, and Transition described elsewhere on this page and on Vocabulary operates at the State layer defined here.

State is introduced here as a descriptive construct. Identity, Will, and Memory extend the framework into explanatory claims about why States arise and change — they sit one layer deeper than the vocabulary itself, and are closer to mechanism than description. Treat State as the organising lens; treat Identity, Will, and Memory as the model’s current hypotheses about what shapes that lens over time.

Agency × Clarity

Every State in FSL sits somewhere on two axes.

Clarity is how well a team is reading the moment — whether the picture is shared and stable, or noisy and misread.
Agency is how much a team can act on what it sees — whether it has options and initiative, or is being played rather than playing.

Cross the two and you get four rough territories:

High Agency × High Clarity — the team acts with intention and sees the picture. Drive, Ignite, Slice, and Command live here: coordinated behaviour, functioning as designed.
High Agency × Low Clarity — intention without perception. The team is still acting, still initiating, but on a picture that no longer matches reality. Overheat and Scramble come from here — the difference is temperature, not axis: Overheat still has initiative and is losing control of it; Scramble has already lost it and is reacting.
Low Agency × High Clarity — the team sees the moment clearly but cannot act on it. Settle, Contain, and Connect sit here when they work; Stall is what this territory looks like when it stops working.
Low Agency × Low Clarity — structure dissolving, behaviour turning reactive. Fog and Drift live here, with Collapse at the far end: Drift is early and soft, Fog is later and inert, Collapse is the end-state.

The reason Stall and Contain can look identical from the stands, or Fog and Scramble can both read as “the team’s lost it,” is that they’re neighbours on this grid. The grid doesn’t resolve the ambiguity — but it tells you what question to ask: is this a choice, or a wall? Is this inert, or is it running to catch up?

There’s a deeper ambiguity underneath both of those, worth flagging on its own: low clarity can mean two different things, and FSL doesn't yet distinguish them. One is a team or player misreading information that was genuinely available — a fixable perceptual failure. The other is a moment that's irreducibly hard to predict, where no amount of attention would resolve it faster. FSL's vocabulary was built for the first case. Where a Reading finds itself needing the second, that's a sign the language is being asked to explain something it doesn't yet have the tools for.

What Kind of Thing Is a State?

It's worth being explicit about what kind of claim a State is. Does a State cause the behaviour that follows it? Does it emerge from behaviour rather than drive it? Or is “State” simply the best summary an observer can construct from what they're watching, with no claim that it exists independently at all?

FSL currently reads differently depending on which page you're on — the Loop (Identity → Will → Forces → States → Outcomes) implies causation; the hedging elsewhere implies something looser. This is an open question. The language can work under any of these interpretations, but the research plan will eventually need to decide which one is actually being tested. Until then, treat State as a descriptive construct whose deeper ontology is still being clarified.

This distinction matters beyond terminology. If State fails to hold up under testing, that wouldn't invalidate the vocabulary as a way of watching football more carefully — it would simply mean State doesn't (yet, or ever) earn status as a scientific construct. The two outcomes are separable: a language can be perceptually useful long before it becomes empirically validated.

Will

A Transition — the moment a Force actually moves a team from one State to another — can happen two ways.

Without Will, a Transition is accidental. A team drifts into a new State because pressure or emotion or circumstance pushed it there.

With Will, a Transition is deliberate. A team shifts state on purpose — a captain slowing tempo, a manager substituting for calm.

This is one of the harder ideas to apply in real time. After the fact, deliberate Transitions often look different — the timing is right, the shift holds, the team seems to know what it’s doing. In the moment, that’s harder to read, and FSL doesn’t yet have a reliable way to tell Will from luck. That’s an open problem.

Identity

Identity is a team’s home State — the behavioural pattern it gravitates back to, especially under pressure. Not a fixed setting so much as a centre of gravity.

Identity shows up at multiple levels:

Club / Country Identity — long-term behavioural culture.
Team Identity — shared behaviour of the current group.
Manager Identity — the pattern the manager brings.
Player Identity — individual tendencies.

These levels pull against each other. A player’s Identity can override the team’s under pressure. A manager’s Identity might take years to displace a club’s. Which level shows up strongest is not yet governed by an FSL rule. The interesting moments are usually when two levels visibly conflict.

Memory

Memory is the residue a match leaves behind. Outcomes accumulate and lean on Identity, reinforcing patterns a team trusts or slowly eroding them.

Of everything on this page, Memory is furthest from being tested. The current research plan doesn’t measure it directly. It’s a sketch — a natural extension of the model, not a validated claim.

Outcomes and the Loop

Outcomes are the visible evidence a match produces — the consequences of however the States actually flowed. An Outcome is worth asking a few questions of: did clarity hold? Were Forces chosen or accidental? Did Identity stay intact? Was Will present? Did a Failure State take over? Did the system stay coherent?

Put Identity, Will, Forces, States, Outcomes, and Memory together and they describe a loop:

Identity → Will → Forces → States → Outcomes → Memory → Identity

This loop is the whole of FSL in miniature — but it’s also the least tested part of the language. Only States and Transitions currently have a coding manual. The Loop is what the model implies; it isn’t yet validated. Read it as the shape the theory is reaching for.

If you want to see how any of this could be tested, the Evidence & Methods page has the coding manual, hypotheses, and reliability requirements. If you just want the words to keep open during a match, Vocabulary is the shorter, lighter version.