Methodology

A percentage on Ledger is an estimate of how likely a result is. "40%" means that in about 40 of every 100 similar matches we would expect that result. It is not a promise that the result will happen.

The home win, draw and away win figures on a match page are free. They describe the result at the end of normal time: extra time and penalty shoot-outs are not counted.

Where the figures come from

The figures come from one of two statistical models. Both work from past match results only. Bookmaker prices are not an input, so a change in the odds does not change these figures.

A match page labelled "Bookmaker odds, margin removed" does not show these figures: see Bookmaker odds below.

The goal model

When it has enough data, the model works out an average number of goals for each team in the match, turns the two averages into the chance of every possible score, and adds those up into a home win, a draw and an away win. A small correction makes low scores such as 0-0 and 1-1 come out more realistically. It uses:

• Goals scored and conceded by both teams in past matches, with recent matches counting more than older ones.
• The strength of the opponents in those matches.
• Whether each team plays at home or away, and the competition the match is played in.

The model's figures are then checked against how often each result actually happened in past matches, and adjusted where they ran consistently high or low. In technical terms, this is a Poisson goal model with the Dixon-Coles low-score correction, calibrated on past results.

Enough data means two things. The competition the match is played in needs at least 80 past matches in our data, so the model can tell how many goals are usual in it. Each team needs at least five past matches in competitions that have reached 80.

The rating fallback

When the goal model cannot be used for a match, for example because the competition has fewer than 80 past matches in our data, the figures come from a simpler team rating instead. Each team's rating goes up or down after every result, by more for a win against a stronger side or by a wide margin, and home advantage is included. The gap between the two ratings sets the chance of each result. These figures are not adjusted against past matches, and this fallback gives only the home win, draw and away win figures.

The rating fallback also needs at least five past matches for each team, counted across all the competitions we follow. A team with fewer gets figures from neither model, so its matches have no analysis. If neither model has current figures for both teams, no analysis is published.

The match page also shows recent form, past meetings between the two teams and injuries for context. These are not inputs to the model.

Bookmaker odds

A match page labelled "Bookmaker odds, margin removed" shows the chances that the bookmaker's prices for that match imply, with the bookmaker's margin taken out so that the chances of all the results add up exactly. Those numbers come from the prices, not from the model, and the page says when the prices were taken. These figures are updated when new prices are taken, until kick-off; once the match is over, the free page labels them "Bookmaker odds before kick-off".

Leagues covered

Match analysis covers selected leagues.

When it is published, and what can change it

Every analysis is published before kick-off and can be updated until then. At kick-off it is locked: it is not changed after the match starts or after the result is known.

Before kick-off, the home win, draw and away win figures change only when:

• the models are refitted after match results are added or corrected (we check for changes every hour);
• a refit moves the match from the rating fallback to the goal model, for example once its competition reaches 80 past matches;
• we change the model itself. Every analysis records the version of the model that produced it.

New prices, team news, injuries, line-ups, form and past meetings do not change these figures. Other parts of the match page, such as the injury list, can be updated as new information arrives.

The list above does not cover a page labelled "Bookmaker odds, margin removed": its figures are not the model's (see Bookmaker odds above).

Limits

• The model can be wrong.
• The model does not take team news, such as injuries, suspensions or line-ups, into account.
• The rating fallback is simpler than the goal model and is not adjusted against past results.
• Results vary. An unlikely result happens regularly across many matches.

New to match pages? How to read a Ledger match page

Ledger is for people aged 18 or over. About Ledger

18+ · Ledger publishes analysis, not betting advice, and cannot guarantee any result. Gamble responsibly · BeGambleAware.org · Responsible gambling