How to read football match probabilities (without a stats degree)

You don't need a statistics degree to read a match analysis — you need to know what four or five numbers mean and, more importantly, what they don't. Here's the whole vocabulary.

1X2: the three numbers that sum to 100

Every match has three possible results: home win, draw, away win. A model assigns a probability to each, and they add up to 100% because one of them must happen. Read 55% / 25% / 20% as: in 100 replays of this match, home wins ~55, it's level ~25, away wins ~20.

The trap is treating the biggest number as a fact. 55% is the favourite — and it still loses 45 times in 100. The number isn't a prediction of this match; it's the rate across many matches like it. Internalise that and you're already reading it better than most.

Confidence isn't the size of the number

Here's the part that separates an analyst's eye from a punter's. A 75% favourite feels "confident" — but if the market rates the same side 75%, the model is just agreeing with everyone. No new information, no edge. Meanwhile a modest-looking 46% pick can be the strongest read on the card if the market only gives that side 39%.

So when you look at a probability, ask two questions, not one: how likely and how far from the market. The gap is the read. Modal shows both side by side for exactly this reason.

The derived markets: same match, different question

The 1X2 isn't the only thing a score model tells you. Because a Dixon-Coles model produces a probability for every scoreline, you can add up the cells to answer other questions from the same underlying picture:

• Over/Under 2.5 goals — sum the scorelines with 3+ total goals versus those with 0–2. It's the model's read on whether the game is open or tight.
• BTTS (both teams to score) — sum every scoreline where both numbers are 1 or more. High when two attacking, leaky sides meet; low when one side is expected to keep a clean sheet.
• Most likely score — the single highest cell, though as we've written before, even that is usually under 15%.

Because they all come from one grid, they never contradict each other — a quiet lesson in why a real model beats a mystery box that quotes each market from a different back-of-envelope.

What the numbers can't tell you

Probabilities quantify uncertainty; they don't remove it. A calibrated model is honest about how open a match is — and the honest answer, for most matches, is "closer than the headline suggests." Read the percentages as a distribution of what could happen, keep an eye on the gap to the market, and you're reading football the way the analysis is meant to be read. It's information, not instruction — the decision stays yours. 18+.

Want to see it live? Every match on /predictions is laid out exactly this way — 1X2, the market beside it, and the derived markets from one grid.