Methodology

Garuda Lions's predictions come from a statistical model, not opinion or "insider tips". The pipeline in three steps:

1. Team ratings. Attack/defence strengths are estimated from a rolling, recency-weighted window of 10–15 matches. Where expected-goals records cover that window they are built from expected goals; where they do not, they fall back to goals actually scored and conceded. Either way the result is an expected-goals rate (λ) for each side in a fixture.
2. Score distribution. A Dixon-Coles adjusted Poisson model turns those two λ values into a full home/away scoreline grid — see our mathematical methodology page for the detail.
3. Market blending. Where bookmaker odds are available, the model probability is blended with the de-vig'd market-implied probability. The market is treated as an informed input, not noise to ignore.

Every number you see on a match page — 1X2, over/under, BTTS, correct score, handicap — is read straight out of this pipeline. Nothing is hand-tuned per fixture.

We publish our accuracy track record honestly, including where the model falls short.