The strength of the teams
Every prediction starts from the strength of the two teams — how many goals each of them scores and how many it concedes. Every result is weighed against the opponent: three goals against the leaders say more than three goals against the bottom side. We also take the home advantage into account, which differs from one competition to another.
When either team has too few matches played — a newcomer to the competition, say — the match gets no prediction until the record is long enough. Better a match without a prediction than a prediction without a basis.
From goals to every market
From the strength of the two teams we arrive at the probabilities of every market — 1X2, double chance, over and under 2.5 goals, both teams to score and the correct score — by statistical and mathematical methods. European matches, in which teams from different leagues meet, call for a calculation of their own.
When we publish
For the Bulgarian First League, the big five leagues and the European competitions they come out up to a week ahead, for the other competitions for the next few days. Before kick-off we calculate the match once more. If a match is postponed or moved, its page says so.
Frozen and verifiable
At the first whistle the prediction is frozen. After the match its page shows exactly the last version published before kick-off — with the time it was published, its version number and the SHA-256 fingerprint of its numbers — so anyone can check that the prediction was not adjusted after the result.
Every prediction is scored after the match, without exception. The Prediction accuracy page shows how many came true — for the outcome, for the goals and for both teams scoring — by competition, by week and by level of confidence. The weekly reports stay available after the week is over.
Upsets
When an upset is likely, we mark the match as a possible upset.
Reading the numbers
A probability of 60% does not mean that the team will win; it means that out of ten such matches it would win about six. The fair odds are 1 / the probability, without a margin: at 50% they are 2.00, at 25% they are 4.00. The predictions are calculated probabilities, not betting advice, and they do not improve the chance of winning.