Last Sunday in Madrid, Atlético and Villarreal drew 2-2. Even game, says the scoreboard. Two apiece, everyone goes home with a point.
The ledger disagrees. Atlético created 0.71 xG all afternoon. Villarreal created 3.60. One side built five times the chances of the other — and the scoreboard shows none of it.
A football match keeps two records. The scoreboard is what happened: goals, points, the table. The ledger is what the chances say: every shot priced by how often shots like it score, added up over ninety minutes. The scoreboard decides who won. The ledger carries the information about what tends to happen next.
Reading football well is mostly knowing which record to ask which question.
What the ledger actually is
xG — expected goals — is chance accounting. Every shot gets a value: a penalty is worth about 0.79 because penalties score 79% of the time; a hopeful hit from outside the box is worth about 0.04. Add up a team's shots and you get their xG for the match: not a prediction, an account of the chances they built. (New to xG? Start with What is xG?)
The ledger has a second page most people skip: xGA, expected goals against — the same accounting applied to what a team allows. A defence that gives up 1.7 xG a game is leaking chances whether or not the goals have arrived yet. The scoreboard can flatter a defence for weeks. The ledger books the leak the day it starts.
Here are three ways to read it — the same three we used to pick Sunday's games.
Tool one: attack meets leak
Both-teams-to-score is not really a question about attacks. It's a question about matchups: does each attack meet a defence that leaks?
Newcastle v Liverpool, Sunday lunchtime. Over their previous ten league games:
- Newcastle created 1.56 xG a game — and conceded 1.67 a game at the other end.
- Liverpool created 1.48 — and conceded 1.42.
Cross them over. Newcastle's 1.56-a-game attack was walking into a defence allowing 1.42. Liverpool's 1.48 was walking into one allowing 1.67. Both matchups sat comfortably above 1.4 expected goals. Both sides also had conversion roughly in line with their chances (Newcastle scoring 1.3 a game off 1.56 created; Liverpool 1.6 off 1.48) — which matters, because it means the xG numbers weren't being distorted by a hot or cold finishing streak. The account was clean.
Full time: 2-2. Newcastle created 1.42 on the day, Liverpool 2.23. Both attacks cleared 1.4 again; both defences let it happen. The last kick of the game — a 99th-minute penalty — levelled it, and the scoreboard got its drama. But the shape of the game was in the ledger before kick-off.
Tool two: the total prices the total
Over/under is simpler: it's not about who scores, only how much football happens. So ask the ledger for one number — combined chances per game.
Manchester City v Bournemouth: City's games plus Bournemouth's games were producing a combined 3.07 xG per match across their last ten. That's the raw material of a goals game.
Full time: 2-1. Total xG created on the day: 2.99. Three goals from three-ish expected. The total priced the total, almost to the decimal.
It won't always land that neatly — single games are noisy. But the logic holds: chance volume is what an over/under read is made of. If two teams' fixtures keep producing 3.0 expected goals, you don't need anyone's opinion about "attacking intent". The account already says it.
Tool three: the leaky end
Back to Madrid, because this is the deepest lesson of the three.
We flagged Atlético v Villarreal before kick-off for one reason: the leaky end. Atlético were allowing 1.63 xG a game over their last ten — the leakiest defence on the pitch, leakier than Villarreal's 1.25. When the most porous thing in a fixture is a defence, both-teams-to-score has structural support: someone reliable is doing the leaking.
Full time: 2-2, both scored. And the ledger's version of the draw: 3.60 created against Atlético, 0.71 against Villarreal.
Now the part that separates ledger-readers from scoreboard-readers. Atlético scored twice from 0.71 xG. That is a finishing spike — two low-probability chances both going in. Spikes happen; they are why football is watchable. But they don't repeat on demand, and building expectations on them is how people end up "unlucky" for a month. The 3.60 conceded is different: that's structure — a defence that let a mid-table side build five times its own output. Next week, the spike resets to zero. The structure travels.
When a scoreline and a ledger disagree, the scoreline is the anecdote and the ledger is the evidence.
When the ledger lies
An honest reading guide has to include the ways the account misleads — because it does, in known ways.
Small windows. Five games of xG is weather, not climate. We don't put a signature on fewer than three data points, and ten games is where a read starts deserving the name.
Different books. A promoted team's ledger was written in a different division, against different opposition. Numbers earned in the second tier don't convert 1:1 to the top flight — which is why, in the Broovo app, a promoted side's comparison against an established one is withheld rather than priced at face value. (Two promoted sides meeting each other is another matter: same book, same yardstick — that read stands.)
The two records, kept honestly
Every match page in Broovo is built on this split. The scoreboard is there — it's the thing everyone already knows. The work is the ledger: chances created and allowed, per game, over a window you choose, with the derivation one tap away and the caveats on the surface where they belong.
The percentage you see on a read is only the conclusion. The ledger is the reason.
Related reading
- What is xG? — the series opener: every shot has a chance, xG is their sum.
- The coin flip that isn't — how a both-teams-to-score read is built.
Notes
- All match figures are Broovo's own serving data (multi-provider consensus source), 23–24 August 2026. xG models differ by provider; figures are consistent within Broovo.
- Penalty constant ≈ 0.79: Opta / The Analyst.
