What you should have known, when — and what actually happened
Everything the desk knows about this station on one page: what the market of models says today, what it said about each recent day at every lead — frozen at issue time, never rewritten — and the official observation beside it. The ledger answers the three questions a forecast record exists to answer: what should you have known, how far ahead should you have known it, and what happened.
The official NWS number against the market of models for the day you pick. Ticks are each source's bias-corrected mean; the shaded mass is the market's disagreement with the official call.
Each row is one day. Reading left to right is reading backwards in time: what the market of models said about that day seven days before, then six, then five … down to the day before, and finally what the thermometer printed. Colour is the miss. Knowable from is the longest lead from which the market was already within 2 °F and stayed there — the day you could have acted on it.
Mean absolute error of each source's bias-corrected mean at this station over the ledger window, by lead, against the market's own. Bold is the best at that lead.
How many settled days were locked in — within 2 °F and never leaving — from each lead. A tall bar at D5 is a day you could have traded a week out.
Every model as a bettor against the market of models on this station's contracts (High ≥ K, flat $1 wherever it disagrees by 5 points or more). A bettor in the black here carries information the market is not using at this point.
Calibrated odds that the local calendar day is wet, and how wet, from the National Blend's 12-hour PoP and QPF re-cut to this station's day and fit against the official record. Prices in cents on a $1 binary; settles on the NWS daily precipitation, trace = 0.00.
Menu rows are frozen at issue time and never rewritten; the ledger reads the latest issue at each lead. Truth is the official ACIS daily record (whole degrees), with URMA standing in for the day or two ACIS has not yet printed. Not investment advice — a simulated market of models.