T20 ratings Track record

Model v market

Every forecast here was published before the first ball and scored after the last. Where the betting market priced the same thing, we score it too, so you can see who was closer.

Loading the record…

Early prices: picks posted before the market settles

A day or two before a match, bookmakers' prices have not yet caught up with what the sharpest markets know. We post a pick when a bookmaker's price beats our fair price by at least 3%, between 48 and 12 hours before the start, and score it against the fair price at the start. The fair price at the start is the sharpest market's price when the match begins, with the bookmaker's margin taken out. If our price keeps beating it, the model is finding something; one match on its own tells you very little. We're recording these to test them. Don't bet on them yet.

Loading the picks…

Before the live record: five seasons, tested blind

Frozen before each season from 2022 to 2026, the model projected 27,584 player-innings in 1,228 matches it had never seen, and beat each player's recent form in every season. Against the betting market's player runs lines, measured over thousands of lines, its expected runs proved about as good as the market's, no better. For the 10- and 15-over runs, values built from the main markets beat the thin markets' own prices in every season tested.

Projected runsAverage projectionWhat they scoredPlayer-innings
0–52.11.68,041
5–107.47.83,547
10–1512.613.23,359
15–2017.618.63,849
20–2522.523.14,560
25–3027.227.22,824
30+33.332.51,404

Every update to the live record is a public, timestamped commit: github.com/anshbansal67/fm-live