Lauantaivakio

A good afternoon for the coupon that the forecasting can take limited credit for.

Where it was actually lost. Blackburn v QPR and Norwich v Burnley. Both were written on the Tuesday, four days out, as close matches with a home lean, and both went the other way: QPR won 2–1 at Ewood Park, Norwich won 4–1. Both were doubled, so the coupon caught them anyway — the probabilities didn't. A forecast written four days before kickoff is working off four-day-old team news, and both of these look like exactly the kind of match where that gap matters.

Stockport, still. The Stockport forecast was a hand estimate read off form and the venue, with nothing firmer behind it. They won 5–1, and it was one of the round's costliest legs. The lesson holds regardless: a guess dressed up as a forecast is still a guess.

Coventry is the one worth staring at. The forecast backed the home side; the Veikkaus crowd backed them least of anyone. Hull won 0–1. The pool — the least sophisticated number on the page — was the only one not leaning home. One match doesn't move a prior, but it's a data point filed under "the crowd reads soft home favourites better than I do."

Draws, still. Three draws, at Bournemouth, Watford and Cardiff, and all three landed on outcomes the coupon hadn't marked, Cardiff on a double with X squeezed out. Two rounds running the excluded draw has been the miss. The volume isn't the problem any more; the reflex to drop X to fit the system is.

The structure carried it again. Favourites came in less than half the time, still short of where the model is meant to be. The coupon hit anyway because Spurs, Blackburn and Bolton all lost as home favourites and all three were doubled. Same shape as 22 August: the system does the work, and a good coupon headline isn't the same claim as good forecasting.