How Our Football Predictions Work

Statistical distribution model — expected goals, win probabilities, BTTS and Over 2.5

Overview

Fixture360 generates pre-match probability estimates using a statistical distribution model — the same statistical framework used by quantitative analysts at professional sportsbooks. Given two teams and their historical scoring records, the model produces:

Expected goals (xG) — the mean number of goals each team is likely to score (λ home and λ away).
Match outcome probabilities — home win %, draw %, away win %.
Both Teams to Score (BTTS) — probability that both teams find the net.
Over 2.5 Goals — probability that the match produces three or more goals.

All figures are calculated once per fixture before kick-off and stored. They are not updated in-play.

Step 1 — League average goals

Everything starts with the league baseline — the average number of goals scored by home teams and away teams across the competition. We maintain a separate average for each side because home advantage is real and consistent across world football: home teams typically score around 30% more goals than away teams in the same league.

These league averages are refreshed daily using completed fixtures from the current season, blended with the previous season's figures when the current season is young (fewer than 19 games per team).

            League avg home goals (μH) = total home goals ÷ total home games

            League avg away goals (μA) = total away goals ÷ total away games

Step 2 — Team attack and defence strengths

Each team is assigned two strength values, calculated relative to the league average:

            Attack strength  = (goals scored per game) ÷ μH or A

            Defence strength = (goals conceded per game) ÷ μA or H

A value above 1.0 means the team is above the league average; below 1.0 means below average. For example, an attack strength of 1.35 means the team scores 35% more goals per game than the league norm. A defence strength of 0.80 means they concede 20% fewer goals than average — a strong defensive record.

Home teams use their home games only for attack and their opponents' away goals when measuring defence. Away teams use the mirror image. This preserves the home/away split throughout the model.

Step 3 — Expected goals (λ)

The expected goals for each team in the specific match-up is calculated by multiplying the relevant strengths together:

            λhome = Home attack strength × Away defence strength × μH

            λaway = Away attack strength × Home defence strength × μA

For example: if the home team has an attack strength of 1.40, the away team a defence strength of 0.90, and the league averages 1.35 home goals per game, then:

λ home = 1.40 × 0.90 × 1.35 = 1.70

The home team is expected to score 1.70 goals in this particular fixture.

Step 4 — Scoreline probability matrix

With λ home and λ away in hand, we use a probability mass function to calculate the probability of each team scoring exactly 0, 1, 2, 3, 4, 5, or 6 goals independently. Every combination of scorelines (e.g. 0-0, 1-0, 2-1 …) is then cross-multiplied to produce a 7×7 probability matrix covering 49 possible scorelines.

            P(goals = k) = (e-λ × λk) ÷ k!

            P(score = h–a) = Phome(h) × Paway(a)

Because the matrix is truncated at 6 goals, the probabilities are normalised (divided by their sum) to ensure the final home/draw/away percentages always add up to 100%.

Step 5 — Outcome and market probabilities

Probabilities for each betting market are obtained by summing the relevant cells of the scoreline matrix:

Home win — sum of all cells where home goals > away goals.
Draw — sum of all cells where home goals = away goals.
Away win — sum of all cells where away goals > home goals.
BTTS — sum of all cells where both h ≥ 1 and a ≥ 1.
Over 2.5 goals — sum of all cells where h + a ≥ 3.

Season blending and data phases

Early in a season, a handful of games is not enough to accurately measure a team's true strength. To avoid unstable predictions, Fixture360 uses a linear blend of current-season and previous-season data:

PrevSeason — fewer than 5 current-season games per team. Predictions rely entirely on last season's data.
Blended — 5 to 18 games. Current and previous season data are blended, with the current season weighted progressively higher as more games are played.
CurrentSeason — 19 or more games per team. The model uses current-season data exclusively.

The data phase is displayed on each fixture's predictions tab so you can see exactly how the model was derived.

Limitations

The model does not account for team news — injuries, suspensions, or rotation decisions.
It does not incorporate in-play events, weather, or venue-specific factors beyond the home/away split.
Cup fixtures involving teams from different leagues use the available league data for each team, which may not reflect cup-specific form.
Newly promoted or relegated teams have no previous-season data in the same division — predictions in the first 5 games of their debut season may be less reliable.

Despite these limitations, this type of statistical model remains one of the most empirically validated approaches to estimating football match outcomes, used widely in academic research and quantitative sports analysis.

Disclaimer

All predictions are generated by a statistical model and are for informational and entertainment purposes only. They are not financial or betting advice. No model is infallible — use these figures as one input among many, not as a definitive guide to outcomes. See our responsible gambling information if you intend to use statistical data to inform betting decisions.