Glossary: 40 AI and Data Terms Every Fan Should Know

Articles about AI in sport are full of jargon. This glossary explains 40 of the most common terms in plain English, grouped by topic. Each entry is short on purpose; where we cover a term in depth, we link to the article.

Core AI terms

Term	Meaning
Algorithm	A precise set of steps a computer follows to solve a problem.
Artificial intelligence (AI)	Software that performs tasks we associate with human intelligence, such as recognising images or making predictions.
Machine learning	A branch of AI in which a program learns patterns from examples instead of following hand-written rules.
Model	The result of training: a mathematical object that turns inputs into a prediction.
Training	The process of adjusting a model so that its predictions fit past examples.
Training data	The examples a model learns from, such as thousands of past shots.
Test data	Examples kept aside to check how the model performs on cases it has not seen.
Feature	One piece of input information, such as shot distance or a team's rating.
Label	The known answer for a training example, such as "goal" or "no goal".
Neural network	A model made of layers of simple connected units, good at complex patterns such as images.
Deep learning	Machine learning with large neural networks of many layers.
Large language model (LLM)	A neural network trained on huge amounts of text to generate language. See Large Language Models and Sports Journalism.

Building and testing models

Term	Meaning
Overfitting	When a model learns the noise in its training data and performs worse on new data.
Cross-validation	Testing a model several times on different slices of the data to get a more reliable score.
Backtest	Running a model on past events as if it had been used at the time.
Data leakage	When information that would not have been available at prediction time slips into training, making results look better than they are.
Logistic regression	A classic method that turns a weighted sum of features into a probability.
Decision tree	A model that predicts through a series of yes-or-no questions.
Random forest	Many decision trees trained on different samples, with their answers combined.
Gradient boosting	Building models one after another, each focusing on the errors of the previous ones.
Ensemble	Any combination of several models. See Consensus Models.
Hyperparameter	A setting chosen before training, such as how deep a decision tree may grow.

Judging predictions

Term	Meaning
Probability	How likely something is, from 0 (impossible) to 1 (certain).
Calibration	Whether predicted probabilities match reality: events given 30% should happen about 30% of the time.
Brier score	The average squared difference between predicted probabilities and what happened. Lower is better.
Log loss	A scoring rule that punishes confident wrong predictions especially hard.
Accuracy	The share of predictions that were right. Misleading when one outcome is much more common than others.
Precision	Of the cases a model flagged, the share that were real.
Recall	Of the real cases, the share the model flagged. See How AI Is Used in Injury Prevention.
Baseline	A simple benchmark a model must beat, such as "always pick the home team".

Sports data terms

Term	Meaning
Event data	A record of every on-ball action: passes, shots, tackles, with their locations.
Tracking data	The positions of every player and the ball, many times per second. See Computer Vision in Football.
Computer vision	AI that extracts information from images and video.
Pose estimation	Working out the positions of a person's body parts from images.
Expected goals (xG)	The probability that a shot becomes a goal, based on similar past shots. See xG Explained.
Expected threat (xT)	How much moving the ball from one area to another raises the chance of scoring.
Elo rating	A rating that goes up or down after each match depending on the result and the opponent's strength. Try the Elo rating calculator.
Poisson model	A model that treats goals as random events at an average rate to estimate scorelines.
Implied probability	The probability contained in betting odds; for decimal odds, 1 divided by the odds.
Per 90	A statistic divided by minutes played and multiplied by 90, so players with different playing time can be compared.

Key point: When a term appears with a number, such as "accuracy of 75%", always ask what it is compared with. A number without a baseline tells you very little.

Core AI terms

Building and testing models

Judging predictions

Sports data terms

Further reading