Forecasting football
Search for who will win, and every page answers with a bookmaker's price. A price explains nothing. Here is what actually moves a football probability, and how to read the one you are given.
A price is not a probability
Look up the favourite for any big competition and the pages answer with a number like 5.50. That is a price, not an estimate. It carries the bookmaker's margin and the way money happened to land across bettors, and it comes with no reasons attached.
A probability says something else: across many comparable situations, this is the share in which the outcome happens. It can be checked, it can be argued with, and it rests on things that can be named. That difference is set out on forecast, prediction, bet.
What actually moves the number
- Form, over an honest window
- Three matches say almost nothing: at that volume luck outweighs quality. A run becomes evidence once it covers enough games for the noise to partly cancel.
- Absences, by role
- Not all absences are equal. Losing a first-choice goalkeeper or the only creative midfielder changes the shape of the side; losing a squad player changes nothing. The role matters, not the count.
- Fixture congestion
- A club playing European and domestic football three days apart does not field the same team as one that had a week. The effect is real and rarely visible in a published price.
- Home advantage
- It exists, and it is smaller than people assume. Seasons played behind closed doors shrank it without removing it, which tells you part of it is the crowd and part of it is everything else — travel, routine, the pitch.
- The format
- A knockout tie and a league phase do not produce the same uncertainty. Over one match the better side loses often. Over ten rounds, much less so.
Why football is hard to forecast
Football is low-scoring. Two or three goals decide a match, and one of them can come off a deflection. When the total is that small, the share of the result owed to chance is mechanically large: the gap in quality has fewer opportunities to show than in basketball, where a hundred possessions average the luck out.
This is why 60 % on a single match is already a strong number. An overwhelming favourite in football is not 95 %, it is 70 or 75 %. Anyone quoting you more than that on one fixture is selling a certainty that does not exist.
A favourite losing is not a failed forecast
If an outcome is given 60 %, the other one is given 40 %. Four times in ten, the other one happens. That is not an error: it is what the number said.
A forecast is not judged on one match, it is judged on a run. If everything called at 60 % happens roughly six times in ten, the forecaster is well calibrated even when they are "wrong". How to read that is on understanding probabilities.
What Predictly does, and does not do
What it does
You ask in one sentence. Predictly works out which fixture or competition you mean, researches what is being reported today across news and official sources, drops repeats of the same wire story, weighs what argues each way, and returns a probability with its confidence level and the documents it read.
What it does not do
- It runs no statistical model of its own: no recomputed xG, no season simulation. It reads what is published and weighs it.
- It takes no stake and quotes no odds. It is not a betting service.
- It refuses a match already played, a question whose subject names two clubs or two competitions, and a case with too little evidence to call.
- It publishes no accuracy figure, because no verified record exists yet.
The eleven steps in detail are on the methodology.
Questions it can answer
- Will this club reach the semi-finals of this competition this season?
- Will this manager still be in post at the end of the season?
- Will this club qualify for European football this year?
- Will this player pass this goal total over the season?
Each one is about a future fact, checkable, and covered enough in the press for there to be something to weigh. That is the shape of question described on how it works.