Football prediction glossary: AI & data terms explained

How to read this glossary

Each definition focuses on the essentials, then provides a “match-language” translation. If you want the full methodology: Football prediction AI (pillar page).

Probabilities & match scenarios

Probability

A probability measures how likely an event is (between 0% and 100%). Example: “home win 55%” means the most likely scenario, not a guaranteed result.

1X2

Standard match reading: 1 = home win, X = draw, 2 = away win.

Double chance (1X, X2, 12)

Groups two possible outcomes (e.g., 1X = home win or draw). In probabilities, it is simply an addition of scenarios.

DNB (Draw No Bet)

“No draw”: we reason as if the draw outcome is neutralized. Translation: if a team wins, which one is more likely?

Margin / uncertainty

Even with a strong model, football remains uncertain (few goals, rare events). The right mindset: a probability gives a direction, not certainty.

Goals, scores, and “match-level” reading

Over / Under (e.g., 2.5)

Compares total goals to a threshold. Over 2.5 = 3 goals or more; Under 2.5 = 0, 1, or 2 goals. A simple way to describe a tight vs open match.

BTTS (Both Teams To Score)

“Both teams score”. Useful to describe matches where both teams have a non-negligible chance of scoring.

Exact score

Probability of one specific scoreline (e.g., 2–1). Exact score is very sensitive: a small detail can shift the outcome cell (2–0, 1–1, 3–1...). Best practice: look at the top 3–5 scorelines, not a single score in isolation.

Dedicated read: Exact score: rarity & uncertainty

Statistics & indicators

xG (Expected Goals)

An estimate of how likely a shot is to become a goal (chance quality). xG helps analyze performance (who created the best chances), without confusing “performance” with “result”.

Form

An indication of a team’s recent momentum (results, goals, streaks). Caution: a short streak can be misleading—treat it as a signal, not a rule.

Streak (wins, losses, unbeaten runs)

A sequence of similar results. Helpful for context (confidence, pressure), but fragile when the sample is small.

Practical page: Team form & streaks

Statistical models & AI

Statistical model

A mathematical tool that turns data into probabilities. A model does not “guess”: it estimates likelihoods using observed regularities.

Poisson

A classic model to estimate rare event counts (such as goals). Useful as a baseline, but football models often add adjustments (home advantage, styles, correlations).

Simulations

A technique that generates many “virtual” matches from probabilities, to obtain a distribution of scores and scenarios (win/draw/loss, over/under, etc.).

AI (in the Foresportia sense)

Here, AI helps detect patterns and improve robustness, as a complement to the probabilistic framework. It does not remove uncertainty: it helps produce more consistent estimates.

Methodology: AI: methodology

Reliability, calibration, and drift

Calibration

Checks whether probabilities “behave” correctly. If we announce 60% over a large set of similar matches, we expect about 6 out of 10. If that is not the case, we adjust.

Dedicated article: Probability calibration

Drift (model drift)

When football “changes”: styles, pace, rules, transfers, scheduling... Drift means past data explains the present less well. We detect it and recalibrate.

Read next: Drift, bias & seasonality

Regularization

A safeguard to avoid being over-confident when data is scarce. Example: a short winning run should not turn a team into a “super favorite”.

Article: Football prediction with limited data

Conclusion

This glossary is a foundation for reading the Foresportia blog without getting lost in jargon. The principle is simple: describe scenarios using probabilities, acknowledge uncertainty, and rely on prediction models + calibration to remain consistent.