A team's recent results are the most visible thing about it and among the least predictive. Measures built from shot quality forecast what comes next far more reliably.
Results carry a lot of noise
Football decides matches on a handful of goals, so a deflection or a fine save moves a result that was otherwise finely balanced. The table records the outcome and forgets the margin.
Two performances that looked nearly identical can be recorded as a win and a defeat. Anyone reading only the results column sees a swing in form that never happened on the pitch.
This is why the last few scorelines make a poor basis for forecasting. They contain real information wrapped in a large amount of luck, with no label separating the two.
What shot-quality measures are counting
Each attempt on goal is assigned a value based on how often comparable attempts have been converted historically, using distance, angle, body part and the situation it came from.
Adding those values across a match gives a figure for how much a team's chances were worth. It is a description of the openings created, not of who put them away.
The measure is deliberately blind to the identity of the finisher. That is a limitation in some contexts and precisely the point in this one.
Why the noisier-looking number predicts better
A team takes many shots in a match but scores very few goals, so the chance-quality total is built from a much larger sample than the scoreline is.
Larger samples are steadier, and steadier measures carry forward. A side's chance creation this month resembles its chance creation next month more closely than its results do.
Finishing, by contrast, swings hard and reverts. A hot streak in front of goal tells you far less about the next ten matches than the volume and quality of openings.
Where the measure breaks down
Squads that are genuinely better or worse at finishing exist, and a model that assumes everyone converts at the league average will misprice them consistently.
Game state distorts it too. A side defending a lead concedes territory deliberately, so the raw totals describe the scoreline as much as they describe the teams.
Set-piece specialists and unusual tactical setups also sit awkwardly inside a model built on league-wide averages. The number needs context rather than blind trust.
How to use it in a preview
The productive comparison is between a team's results and its underlying numbers over a run of matches. A wide and persistent gap is the interesting signal.
A side winning while creating little is usually heading for a correction, and one losing while dominating chances is usually heading for the opposite.
Neither statement is a forecast of the next result. Both change the prior you bring to it, which is all any single input should be asked to do.