How Our Goal Prediction Algorithm Works
Most prediction sites throw dozens of markets at you — 1X2, both-teams-to-score, correct score, handicaps, accumulators — and hope something sticks. We do one thing: total goals. Over or under a line, usually 2.5, sometimes 3.5. This page explains how we get to that call, what the numbers on each prediction mean, and what we don't claim to know.
What goes into a pick
1. Recent form
We look at each team's last 5 matches: how many goals were scored in them (by both sides) and how many finished over or under the line. That gives 10 recent matches per fixture, and you can see every one of them on the prediction page, with the score and the date.
2. Head-to-head
We look at the last 5 meetings between the two teams, if they have met: how many of those games went over or under the same line. Old meetings say less about how two teams play today than their current form does, so head-to-head counts for less.
3. The call
From those numbers, together with our analyst's view of the matchup, we pick one side of one line: for example, "Over 2.5" or "Under 3.5". Most picks follow the numbers. Some don't, and those are marked as an analyst pick (see below).
What the confidence figure means
The confidence figure shows how much of that recent evidence lands on the side of our pick:
- 80% of the figure comes from the 10 recent matches of the two teams: the share that finished on the side of our pick.
- 20% comes from their last 5 meetings: the share that finished on the same side. If the teams haven't met, the figure is based on recent form only.
An example: if 9 of the 10 recent matches and 4 of the last 5 meetings went over 2.5, an "Over 2.5" pick shows 0.8 × 90% + 0.2 × 80% = 88% confidence.
So the confidence is a measure of how strongly the recent numbers back the pick. It is not the probability that the pick wins, and it doesn't take the bookmakers' odds into account. A 60% confidence pick can still lose, and so can a 90% one: football is not that predictable.
Analyst picks
Recent form can mislead. A team's last five games may have been in another competition, against much stronger or much weaker opponents than the one it meets next. A Norwegian side's league games and a Gibraltar side's European qualifiers, for example, don't tell you much about how the two will play each other.
When our analyst expects a match to go against the recent numbers, we still publish the pick, but we don't show it with a low confidence figure. It is marked Analyst pick, the prediction page says that it goes against the recent numbers, and the numbers themselves stay on the page so you can judge for yourself. Analyst picks are judgement calls; treat them that way.
Reading a prediction
- Over / Under + line — our call and the goal threshold it's measured against, e.g. "Over 2.5."
- Confidence — how much of the recent evidence backs the pick, as explained above. Or Analyst pick when the pick goes against it.
- Why this pick — a short explanation written from the numbers on the page: what the form and the head-to-head say, and whether they back the pick.
- Best odds — the best price we found for the pick. The odds are shown for information; they are not part of the confidence figure.
- Status — an hourglass means the match hasn't kicked off yet, a pulsing dot means it's live, and a final score with a ✓ or ✕ means it's settled. We show all three, not just the wins.
Track record
We log every prediction we publish, whether it wins or loses. If a pick loses, it stays in the record. The win rate on the homepage covers the last 30 days, and each country page shows the record for that country. That's the only way a track record means anything.
Markets we cover
Over/Under 2.5 and 3.5 goals are our core markets: each fixture we cover gets one pick on one of the two lines. We're looking at 1.5 goals and both teams to score, but we'd rather add a market slowly and get it right than launch everything at once.
What this isn't
A prediction here is an estimate, not inside information, not a guarantee, and not financial advice. We don't have access to anything a bookmaker doesn't — we're reading the same public data, just focused entirely on one narrow question instead of trying to predict everything about a match. Treat every pick as one input into your own decision, not a replacement for it.
If betting stops being fun, or starts feeling like something you can't step away from, that's worth paying attention to. Set limits before you bet, not after. You must be 18 or over to use betting services referenced on this site. Free, confidential help is listed on our responsible gambling page.