One question per piece, answered with data; the method published, the uncertainty stated, every probability scored in public.
Each piece asks one question and answers it from data. The data and the method are published with the answer, the uncertainty is stated, and every probability is scored afterwards on the ledger. Every figure comes from a notebook anyone can run, every dataset is named with its licence, and a probability is a calibrated number with its inputs, never a claim about who will win. The scores are on the forecast ledger; the approach is set out on the method page.
The top four seeds have won 25 of the 27 men's Slams since 2020, yet a third of the top eight still leave in the first week.
Over 178,275 matches since 2022 the betting favourite lost 31.2% of the time, a 90% favourite about once in 23. The pattern holds from the Grand Slams to ITF level.
Since 1990 the lower-ranked player has won 29.5% of Slam matches and 35.7% at the Masters. Five sets alone explain part of that gap, less where the pairing is lopsided.
