Football Data Analytics Glossary
The 11Stat Glossary defines the core terms behind football probability modeling, such as xG, model agreement, data quality, calibration and CLV, in plain language. These are analytical concepts, not promises of any outcome.
What this glossary is for
11Stat is a football data analytics platform, not a betting site. This glossary explains the technical terms you see across our panels and reports so you can read a match's probability distribution, how closely the models agree, and how trustworthy the underlying data is.
One important note: none of these terms guarantee an outcome. Metrics like ROI and win rate are strictly simulated, paper-backtest quality measures. Everything here is for education and analysis, and is not betting advice. Understanding the difference between football predictions and data-driven analysis starts with this vocabulary.
Core probability terms
xG (expected goals)
xG estimates how many goals a team would be expected to score based on the quality of the chances it created. It sums the scoring probability of each shot. A high xG reflects strong chance quality, not a certain result.
Probability analysis
Probability analysis is the process of turning a match's possible outcomes (a win, a draw, total goals, and so on) into percentage probabilities. Instead of a single prediction it produces a distribution, assigning separate likelihoods to scorelines such as 1-0, 2-1 or 1-1.
Team form
Team form is a time-weighted summary of a team's recent performances, including goals scored and conceded, results and overall control. In 11Stat, form is computed with a multi-window method that gives recent matches more weight.
Model signal, agreement and risk
Model signal
A model signal is an analytical observation that a model highlights for a specific market (for example, match result or total goals). It is not a tip or a coupon, only a directional data output produced by the model.
Model agreement
Model agreement measures how strongly independent engines (Poisson/Dixon-Coles, Elo/form, and the advanced engine) point in the same direction. A label like '3/3 engines agree' means all three support the same outcome. Higher agreement indicates a more consistent signal, but never certainty.
Risk level
Risk level (Low/Medium/High) summarizes how volatile (high-variance) a signal is, given its probability, model agreement and data quality. A high risk level means the outcome is more uncertain.
Data quality and calibration
Data quality
Data quality measures how complete, current and reliable the available data is for a match. Missing lineup news, small samples or a brand-new season all lower it. When data quality is poor, 11Stat weakens or disables signals; for example, fixtures below the quality threshold in new tournaments are excluded from analysis.
Calibration
Calibration measures how well a model's stated probabilities match reality. In a well-calibrated model, roughly 60 of every 100 events it calls '60%' actually happen. Predicting correctly and giving the right probability are different things, and calibration audits the latter.
Market abbreviations (OU2.5, BTTS)
OU2.5 (Over/Under 2.5)
OU2.5 is the market for whether total goals in a match will be above or below 2.5. 'Over' means 3 or more goals; 'Under' means 2 or fewer. 11Stat presents this not as advice but as a probability derived from the model's goal distribution.
BTTS (Both Teams To Score)
BTTS, also written KG (Var/Yok) in Turkish, is the market for whether both teams will score in a match. The model derives this probability from each side's attacking and defensive expectations.
Modeling methods and a value metric
Dixon-Coles
Dixon-Coles is a football scoreline model that extends the classic Poisson goals model to correct biases in low-scoring results (0-0, 1-0, 1-1). 11Stat's primary engine uses this approach with time-decayed weights.
Elo
Elo is a rating system that expresses a team's relative strength as a single number. It updates after each match, with surprising results moving the rating more. In 11Stat, Elo feeds the form-based secondary engine.
CLV (Closing Line Value)
CLV is an analytical quality indicator that measures where an analysis's probability lands relative to the market's closing (final) odds. Consistently being positioned better than the closing line is a strong sign that a model carries information. CLV is not a measure of winnings; it is an audit metric for model quality. CLV is not a wagering tool or profit measure; it is a pure analytical performance metric used to audit the model's historical calibration accuracy.
More market terms: HT/FT, Handicap, Correct Score
HT/FT (Half-Time / Full-Time): a market combining the half-time and full-time result; 11Stat analyzes it as a probability distribution, not a single tip.
Handicap (Asian Handicap): a market that balances probabilities via a virtual goal advantage/disadvantage; related to goal expectancy and model probabilities.
Correct Score (Score Distribution): the probability of a specific final score; instead of one score tip, 11Stat produces the full distribution via a Poisson/Dixon-Coles score matrix.
Frequently Asked Questions
Is 11Stat a betting site?
No. 11Stat is a football data analytics and probability-modeling platform. Its signals and metrics are for education and analysis, are not betting advice, and never guarantee any outcome.
Why can xG differ from the actual score?
xG reflects the expected goals from the quality of chances created, while the real score can swing due to luck, goalkeeping and finishing. xG measures performance quality, not a prediction of the final result.
If model agreement is high, is the outcome certain?
No. High model agreement means different engines point the same way and the signal is more consistent, but it never implies certainty. Football is inherently uncertain.
What happens when data quality is low?
11Stat weakens or fully disables signals for matches with missing or thin data. This prevents unreliable inputs from producing misleading analysis.
Why is CLV an important metric?
CLV measures where a model's probability sits relative to the market's closing odds, auditing whether the model genuinely carries information. It is a long-term indicator of model quality, not a promise of returns.