Japan v Afghanistan

500 complete T20 matches are played ball by ball between the two named XIs, in your browser, using each player's record by phase of the innings and by type of bowling — the same simulator behind the Match Center's Predict Winner. The toss is not known, so each side bats first in half of them.

The forecast is not simply the share of those matches a side wins. Tested on over 1,700 T20s it had never seen, that share proved far too sure of itself — a side it gave 70% won little more than half the time. So the share is weighed for how sure it has proved and combined with each side's team strength (the strike rates of the top six and the economy of the attack). A percentage here means what it says: of the sides given about 60%, about 60% won. Under 55% is too close to call.

This is a pre-match read: it does not look at the score, unlike the win probability on the Live tab, and it does not change as the match is played. A player the model has never seen is stood in as a newcomer — a little below the typical player — and named beneath the result. Running it again moves the number by about a point.