How Bookie Verdict's football predictions work

Most prediction sites give you a pick and nothing else. We would rather show our working: our chance for every result, the price you can get, and a public record of how every pick turned out.

Here is how it fits together, step by step.

1. Expected goals for each side

Everything starts with expected goals (xG): how many goals each team should score in this match on average. We build that from how well each side has been attacking and defending, adjusted for home and away, and for who they have been playing.

For example, before Chelsea v Tottenham on 24 October our model had Chelsea at 2.2 expected goals and Tottenham at 1.1.

2. From expected goals to every scoreline

Goals in football are rare, random events, so we can turn expected goals into a chance for every scoreline: 0-0, 1-0, 1-1, 2-1 and so on. Add those up and you get the chances for every market we cover:

• home win, draw and away win
• over and under 2.5 goals
• both teams to score
• the most likely correct scores

3. A sanity check from the market

Bookmaker prices carry a lot of information: team news, injuries and the views of thousands of bettors. So we do not ignore them. Our final chance is a blend of our own numbers and what the market is saying. That stops the model from drifting too far on a hunch of its own, while still letting it disagree when the data supports it.

4. Picking the verdict

For each match we rank the picks by how likely they are and how they compare with the best price available.

• Value pick: our chance is higher than the chance the price implies. A price of 2.00 implies 50%; if we make it 55%, that is value.
• Most likely pick: when nothing shows value, we say so and give you the most likely sensible outcome instead.

You will see a top pick plus a 2nd and 3rd pick for every game, each with our chance next to it.

5. A public record of every pick

Every pick is logged before kick-off and settled after full time on our track record. We show wins and losses, by market, competition and team, so you can judge the model on its results rather than on our claims.

The model also learns from those results. Each settled game feeds back into how it weighs form, goals and the market, so it keeps getting checked against reality.

What a prediction is not

A 65% pick still loses about a third of the time. No model, ours included, can guarantee a winning bet, and accumulators multiply the risk because every leg has to win. Use our numbers to judge chances and prices more clearly, and only bet what you can afford to lose.