Method: each row is written to the database before kickoff with the exact probabilities served on the site at that moment. When the match ends, the stored numbers are compared with the actual result by a scheduled job — there is no manual step and no way to revise a past entry. The full model documentation is on the methodology page.