We were wrong in August. Less wrong by September.
Dear readers,
In May we wrote down a prior for the first frost: 40 days after the equinox, give or take 14. It was wide on purpose, because the weather models disagreed with each other and with us.†
The data arrived in late August, and the posterior narrowed to 31 days, give or take 4. We were wrong by about nine days, which is less wrong than we were in July. We buried accordingly.
“The model did not become confident. It became less uncertain, which is different and more useful.”
Our biggest miss was a rumor that the oak on the north side would fruit early. Under our prior the rumor was unlikely, and it turned out to be false anyway. We have updated the prior, and separately, our feelings about rumors.
As always, the three beliefs that moved furthest from their prior are listed in the appendix, which the squirrel has buried somewhere safe.
We remain long acorns and short complacency. Winter is 58 days away.