Understand football forecasts. Ask better questions.

GUIDE 01

How to read a football forecast

Start with the exact claim. A prediction about the match winner is different from a prediction about total goals. Check the competition, teams, kickoff time and timezone, then read the market definition. Whether extra time counts depends on the stated market rules.

Five questions to ask

1. When was it published? Distinguish a forecast made before kickoff from commentary after the event.
2. What evidence supports it? Look for relevant data and an explanation, rather than confident language alone.
3. What is the context? Consider recent results alongside opponent strength, home and away conditions, squad changes and sample size.
4. What is uncertain? Unconfirmed lineups and missing information should be acknowledged. A numerical estimate is not a guarantee.
5. Has it changed? Corrections should explain what changed and when, while preserving the original prediction.

What should a match preview contain?

A clear preview names the fixture and market, states the estimate if available, explains the strongest evidence and lists the main reasons it could be wrong. It separates confirmed information from assumptions. “Not enough evidence” is a valid conclusion.

GUIDE 02

Decimal odds and implied probability

Convert decimal odds into a raw implied probability using 100 ÷ decimal odds. Odds of 2.00 imply 50%; odds of 1.50 imply approximately 66.7%; odds of 3.00 imply approximately 33.3%.

This converts a price; it does not independently estimate the true chance. Prices may include a bookmaker margin and can change over time.

Keep estimates separate from prices

If an analyst estimates a 60% chance while a price implies 50%, that difference is only as useful as the estimate. An inaccurate model can create an apparent advantage where none exists. This comparison alone is not a reason to place a bet.

Probability is not certainty

A high-probability outcome can fail. A low-probability outcome can happen. Assess repeated forecasts over a meaningful record rather than treating one result as proof of a system.

GUIDE 03

How to judge a forecasting record

Winning screenshots tell you little without the predictions that lost. Look for a complete record: publication time, fixture, market, estimate, outcome and any correction. Where odds are shown, their source and observation time matter.

Use consistent settlement rules

Explain how postponed, abandoned and void matches are recorded. Keep them visible and distinguish them from wins and losses. State which outcomes are included in each statistic so readers can reproduce it.

Win rate is only one measure

A high win rate can come from selecting very likely outcomes. It does not establish profitability. Financial performance also depends on prices, stakes and costs. Forecast quality concerns how well estimates match outcomes.

Check calibration and sample size

For a well-calibrated model, events assigned a probability near 60% should occur around 60% of the time across a sufficiently large, comparable set. Small samples can vary widely. Include the number of forecasts and date range whenever presenting a percentage.

These guides explain evaluation methods. They do not claim that MPB has an independently verified performance record.