TipsterCompetition
Plain-English guide to the methods behind our football tips: Poisson goal models, Elo ratings, consensus, trends, machine learning, LLMs and exchange data.
No single method beats football. What works is a toolkit: sound maths for the baseline, machine learning for the patterns people miss, language models for clear explanations, and a habit of testing everything. This page walks through each part in plain English, with a one-line reason why it matters.
Everything starts with data about the match: history, performance, form, momentum, head-to-head and league standings. Every method below works on these facts. Our language model never works from memory or imagination, which is explained on the AI previews page.
A Poisson model estimates how many goals each team is likely to score, using their attacking and defensive records. From those expected goals it works out the probability of every scoreline, and from the scorelines, the probability of a home win, draw, away win, both teams to score, or over and under 2.5 goals.
Why it matters: it turns "Arsenal look strong" into a number we can compare directly with the bookmaker's price.
Elo is a rating system borrowed from chess. Every team has a rating that rises when it beats stronger opponents and falls when it loses to weaker ones. We maintain our own in-site Elo ratings, so they reflect the leagues and teams we cover rather than someone else's assumptions.
Why it matters: it gives an honest, constantly updating measure of team strength that adjusts to momentum.
Our tipster models do not all think alike. Wisdom of the crowd means looking at the consensus of our models: when several independent approaches point at the same selection, that agreement carries more weight than any one voice. Our consensus tipster is built on this idea, as you can see on the AI tipsters page.
Why it matters: averaging independent views cancels out individual blind spots.
Some patterns repeat. We study how selections have performed historically at different odds ranges and in different situations, so a tip is judged not just on who is likely to win but on whether the price is a fair one for that kind of bet.
Why it matters: a good team at a poor price is a bad bet, and odds ranges help us tell the difference.
Machine learning finds relationships in data that are too subtle or too numerous for hand-built rules. Our founder has more than 15 years' experience in the field, and ML actively improves our previews and predictions.
We run machine-learning experiments most weeks looking for gaps in the market. Plenty of experiments find nothing, and we are happy to say so. Only the ideas that hold up make it into the models.
Why it matters: it keeps the models improving instead of standing still.
Large language models do the writing. They turn the facts and calculations into a readable preview, and they have lifted both quality and readability. They are given the facts to work from and are checked afterwards, as covered in how we build AI previews.
Why it matters: clear explanations let you judge a tip for yourself instead of taking it on trust.
As of 5 November 2026 we also collect Matchbook exchange odds. Exchange prices are set by punters betting against each other, with a much smaller margin than a typical bookmaker, which makes them a useful reference for what the market really thinks.
Why it matters: a sharper market price is an extra yardstick for spotting where a bookmaker's odds are out of line.
Research never stops. Most weeks we test new features, new model types and new ways of combining signals, looking for gaps in the market. New ideas are measured against the existing models and kept only if they genuinely add something.
Why it matters: markets adapt. A method that works today can stop working, so constant testing is the only safe policy.
No model can see the future. Injuries, red cards and sheer luck decide matches, and betting carries risk whatever the method. Please read our responsible gambling page, and stake only what you can afford to lose.
Methods are only credible when the record backs them up. Every tip is recorded at real bookmaker odds and sealed, as set out in how our tips work. Check the results page and the tipsters page, or return to the about hub.