Percentages show predictor consensus, not measured chances of winning. Double-chance methods are added; separate fight percentages are multiplied, assuming independence. Shared predictors and overlapping legs mean these suggestions are not independent bets. AI confidence is a separate opinion.