Golmodel prediction guide

How to read football predictions and score probabilities

A football prediction describes uncertainty. It does not tell you what must happen. Golmodel uses a statistical model to estimate match outcomes and scorelines, then lets you compare the prediction with the actual result.

What do 1, X and 2 mean?

1 means a home win, X means a draw and 2 means an away win. These are three separate outcomes for the same match. A 60% home-win probability would still leave 40% for a draw or an away win. That example explains a probability; it is not a prediction for a current fixture.

The match-result prediction is evaluated on the 90-minute result, including stoppage time, rather than the team that advances after extra time or penalties. In knockout football, predicting the match result and predicting the team that qualifies are different questions.

What does the Poisson model do?

A Poisson goal model starts with an expected number of goals for each team. Attack and defence strength help inform these estimates. The model assigns probabilities to goal counts and combines them into possible scorelines. Summing the scorelines gives probabilities for home wins, draws and away wins.

Golmodel’s published methodology includes a Dixon-Coles correction for low-score outcomes and a blend with market consensus probabilities. Model estimates and market estimates are different inputs. Agreement between them is not a guarantee, and neither input removes uncertainty about a football match.

Expected goals are not a promised score

A model value such as 1.6 expected goals is an average over possible outcomes, not a claim that a team will score a fraction of a goal. A pre-match expected-goal estimate is also different from a post-match xG total calculated from the shots taken during a game.

The single most likely scoreline can have a low probability because probability is spread over many possible scores. A likely match winner and an exact-score prediction should therefore be evaluated separately.

Why double chance needs its own result

Double chance covers two of the three match outcomes: home or draw, away or draw, or either team to win. Covering two outcomes usually makes a correct pick more likely than choosing just one. Its success rate should not be compared directly with a single-outcome success rate.

In Golmodel, inspect the single pick and double-chance pick separately, including the matches where the prediction was wrong. A useful record includes all evaluated fixtures, the observation period and the sample size.

How to interpret the track record

The app keeps the record. Performance counts every evaluated featured match, the single pick strictly and the double chance separately, week by week; Results lists each prediction next to the final score. Those figures are reported by Golmodel; they are not independently audited, do not describe exact-score accuracy and do not guarantee future performance.

A stronger assessment compares the same fixtures against a simple baseline, such as always choosing the favourite, and checks whether stated probabilities match observed frequencies. A small sample or a selected run of successful picks cannot establish long-term predictive quality.

Use Golmodel before and after the match

• Open an upcoming fixture and read the outcome probabilities and score estimate.
• Follow your teams and choose the notifications you want to receive.
• Return after the match and compare the original prediction with the result.

Model predictions and the weekly track record are free. Optional Premium adds AI-assisted written analysis and additional context for selected matches. Golmodel is an independent information and entertainment app. It accepts no wagers and does not guarantee outcomes.