How Taktig estimates matches

The Poisson goal model

1. League baseline. From completed matches in the same competition we compute average home goals (H) and away goals (A). The gap between them is the home advantage.
2. Team strengths. Each team's attack = goals scored ÷ goals expected from the league average; defence = goals conceded ÷ goals expected. Recent matches count more (180-day half-life), and each rating is pulled towards average by a prior worth four matches, so a few freak results don't dominate.
3. Expected goals. Home xG = H × home attack × away defence; away xG = A × away attack × home defence.
4. Scorelines. Goals for each side are modelled as Poisson distributions. We compute every scoreline from 0–0 to 10–10, then sum them into home win, draw, away win, over/under 2.5 goals and both teams to score. The three outcomes are normalised to exactly 100%.
5. Uncertainty. Each estimate shows a data-quality label (how many matches back it) and, for Premium, an approximate range from rating uncertainty. With fewer than three completed matches per team, we show no estimate at all.

Known limitations: the model does not know about injuries, suspensions, lineups, transfers, weather or motivation, and assumes goals are independent. Cup competitions use only matches within that cup.