Every city, its own record.

Settled days. A city has two ladders a day, its daily high and its daily low: a set of brackets of which exactly one settles Yes on Kalshi's result. The table counts a market day once every bracket of it has settled; the day-by-day view also lists the days still open. The record starts with the robot's first read, on 2026-09-18.

Ruled out means the station's confirmed reading had cleared the bracket by a full degree on some read: the dead call. Kalshi settles each market on The Weather Company's reported figure for the station, not on these hourly observations, so a ruled-out bracket that settled Yes is counted here, not hidden (settlement).

The leading bracket is the one priced highest at the chosen moment, by its midpoint, or by its bid when nobody offers it; a bracket nobody bids on is never the leader, and a tie goes to the first ticker. It is the favourite of study C-010 (the studies), counted here city by city over every settled day; this table is not that call's grade, which reads only the markets dated October 1 to 31. Its mean midpoint is what the price said on average, and how often it paid is what happened. There is no profit-and-loss column: this measures the price, it does not rank cities.

The moments. Each market closes at midnight local standard time, the end of the National Weather Service's climate day. So 24 hours before the close is the start of the measured day, 12 hours is noon, 6 hours is 6 pm and 1 hour is 11 pm, local standard time (an hour later on the clock while daylight time is in force). The buttons choose the moment for the leading bracket and the price's score.

The price's score is the Brier score: the average squared gap between each bracket's midpoint and what happened, 1 if it paid and 0 if not, so 0 is perfect. Beside it, an even split prices every bracket of a ladder at one over its size. The 95% interval resamples whole market days a thousand times, with a fixed seed, because brackets that settle on the same day share its weather. The calibration page has the same score for every city at once.

When the day is decided. For each settled ladder the robot read from before its day began: the local clock time at which the bracket that went on to settle Yes first had a midpoint of 90% or more while both a bid and an ask were resting (a lone bid of 80 averages with 100 to read as 90%, so one-sided reads are not counted), counted only when that came before the close (the table shows the median; the day-by-day view, the middle half of them); and how many hours before the close the ladder's first bracket was ruled out. The ladders dated 2026-09-18 are left out of these two, because the robot's first read of them came that afternoon. Times only: no price of any bracket is shown.

Not read. Under 20 settled ladders a figure is shown, muted, and not read: its interval is too wide to say much. With 40 rows, about two will sit outside a 95% interval by chance alone. These are measurements of the price and of the weather's timing, not advice on any contract.

All four are defined in cities_views.sql: counts, sums over a ladder and times only. Market dates are each city's own date. No bids, asks, sizes or per-bracket prices are published, and nothing here is offered as a download.