Football prediction & AI glossary

This is a plain-English reference for the terms you will see across AI football prediction tools, including ScoreGPT. The space mixes two different vocabularies: AI/ML concepts (models, consensus, confidence) and football-stats concepts (xG, both teams to score, total goals). We define both neutrally so you can read any prediction page critically.

Every definition is for information and entertainment only.

Each term has its own page with a definition-first answer block, a "how it works" and "why it matters" section, and FAQs. Start with AI football prediction if you are new, then read AI consensus prediction to understand what ScoreGPT means by "consensus" — agreement across AI models, not a poll of people. The football-stats terms — xG, BTTS and Over/Under 2.5 goals — help you read any prediction, from any tool.

These are the five core terms, each with its own definition page:

#	Term	One-line definition
1	AI football prediction	A match forecast produced by software — a statistical model, machine learning, or a large language model — rather than a human pundit.
2	AI consensus prediction	Agreement across several independent AI models on one match — how many of them back the same result.
3	Expected goals (xG)	A metric estimating how many goals a team "should" have scored based on the quality of its chances, not the final score.
4	Prediction confidence level	How strongly a model backs its own forecast — a probability or label, not a guarantee that the pick is correct.
5	BTTS & Over/Under 2.5 goals	Two common match questions: whether both teams score, and whether total goals land above or below 2.5.

On ScoreGPT, "consensus" means what the AI models agree on — never a poll of people. The table below shows the two numbers behind it.

Part	What it means
The consensus call	Five independent AI models combined on one match: their home / draw / away probabilities averaged, and the outcome with the highest average is the call.
The agreement count	How many of the five models picked that same result on their own — for example, "4 of 5".

Agreement is not accuracy: every consensus call is graded in public after full-time, wins and losses alike.

ScoreGPT is AI-native: for each match, five named frontier large language models analyse independently, then ScoreGPT combines them into one consensus call. The five models are GPT-6.1 Sol (OpenAI), Claude Opus 5.5 (Anthropic), Grok 4.7 (xAI), DeepSeek-V4-Pro (DeepSeek) and Kimi K3 (Moonshot); the consensus averages their home / draw / away probabilities and backs the outcome with the highest average.

What is different about ScoreGPT, stated as method not performance: multi-model consensus across five named models, AI-native design, and full public grading of every pick — wins and losses. We never claim a win rate or an accuracy figure, or that we are more accurate than anyone.