Comparing Momentum and State: Rethinking Football’s Hidden Dynamics

Core Question

Opta's Match Momentum and FSL are both trying to explain what a scoreline leaves out. Are they measuring the same thing at different resolutions — or is FSL describing something momentum genuinely can't see?

Exploring the Question Through FSL

Football is mostly 1–0s and 2–1s, and a scoreline on its own can't tell you much about how a match unfolded. Match Momentum was built to answer that: a possession-value model scores every on-ball action, each team's most threatening action in each minute is weighted towards the present, and the gap between the two resulting values is the momentum reading. It's a genuinely useful tool — a team can dominate an entire half without a shot to show for it, and momentum is one of the few metrics that would actually notice.

FSL is trying to answer a different question with a similar-looking shape. Momentum asks who's more likely to score right now. FSL asks what condition a team is in right now, and what that condition makes possible or impossible next. Threat is an output; a state, in FSL's terms, is an attempt to describe the behavioural condition surrounding it. Put simply: momentum describes changing threat, FSL attempts to describe changing — or persistent — behavioural condition. That second half matters more than it looks. FSL isn't only interested in what a team is becoming; sometimes the more interesting claim is that a state hasn't changed at all, even while the score, or the threat, clearly has.

Momentum can show a flat period as clearly as it can show a swing — that's not a blind spot. What it has no vocabulary for is what kind of flatness it is: deliberate stillness, contentment, exhaustion, or a team that can still see the picture but has run out of ways to act on it. Threat and condition can move independently of each other, and it's the periods where they diverge that FSL is actually interested in.

Opta's own account of why momentum exists is instructive: traditional box scores and even shot-based metrics like xG capture events in isolation, and momentum was built to give a more chronological, holistic account instead. That's a progression worth extending rather than stopping at. A scoreline tells you what happened. Box-score statistics tell you what events occurred. Momentum tells you which team was creating more threat as the match unfolded. FSL is a proposal for a further layer: what behavioural condition appears to accompany that unfolding, and whether those conditions are independently recognisable. It's tempting to describe that further layer as explaining why momentum swings the way it does — that overstates it. The more accurate description is: what condition appears to be present, and what that condition makes possible or impossible next. That's a narrower claim than "why," and a more honest one.

Alternative Explanations

A few honest alternatives are worth naming rather than arguing past. FSL's states could simply be a qualitative gloss on what possession share, territory, and momentum already show together — in which case "a long, passive middle" or "Stall" isn't a new layer, it's a narrative label stuck onto numbers that already existed.

There's also a real risk that behavioural condition is just momentum smoothed over a longer window — that what looks like a qualitatively different state is actually the same threat signal, averaged. If that's true, FSL isn't describing a different thing, it's describing the same thing at lower resolution and with more words.

And there's an outcome-bias problem specific to how FSL has been illustrated so far. The Argentina–France and Czech Republic–Scotland readings used here are both drawn from Illustrations, which are retrospective and were chosen partly because the behavioural story reads clearly with the benefit of hindsight. It's easy to retrofit a plausible-sounding state description onto a lull once the outcome is known. Whether the same distinctions would hold on matches nobody has picked for their narrative shape is untested, and is exactly the kind of thing that outcome-blind coding is meant to guard against.

What Evidence Would This Need?

Take a sample of matches not selected for how well they illustrate FSL — an unselected sample, not more famous finals and collapses. Compute Match Momentum for each, minute by minute, as normal.

Separately, have coders classify the same matches into FSL states using frozen coding rules, blind to the momentum values and, where practical, blind to the score and eventual outcome. Establish inter-rater reliability before any comparison runs.

Then test two things. First, whether there are stretches where momentum is flat but independent coders still reliably distinguish different FSL states from one another — evidence that condition carries information threat doesn't. Second, and just as important, whether FSL state classifications end up correlating almost perfectly with momentum bins once plotted against each other — which would mean the vocabulary is redescribing momentum rather than adding to it.

The bar for a positive result should be explicit and set in advance: FSL states need to predict something — behavioural repeatability, a later state change, coder agreement on "this passage matters" — that momentum and possession share alone don't already predict. If they don't, the language isn't yet earning its complexity, whatever the illustrations suggest.

Potential Implications

If FSL states turn out to carry information momentum doesn't, that matters for a few different audiences, each provisionally. For broadcasters and analysts already using momentum graphs on air, it would suggest a genuine second layer worth adding alongside the swing — not instead of it — for explaining spells where the numbers look settled but something about the team's mode has quietly shifted. For coaches, it would offer an earlier signal than threat metrics currently provide: a way of naming a change in a team's condition before it shows up as a change in the chances they're creating or conceding. For analysts specifically, it raises the possibility of flagging false stability — a spell of high momentum or clean possession numbers that's actually fragile, versus one that's genuinely under control, a distinction momentum alone has no way to draw.

None of that follows automatically. It follows only if the evidence above actually shows FSL states carrying information momentum doesn't already carry — and whether that's true, on matches that weren't chosen to make the case, is still an open question.