Route 06 · Winter Odds · the Season Book
What history prices for 52 cities this winter.
For every city and threshold the book shows three things: what the odds are (a fair-value percentage), what could happen (the full range of past winters), and, from April, what did happen (settled and scored). Prices start from the history layer: the 1991–2020 base rate, what the strong El Niño winters did, both moved to today's warmer climate by each station's own trend, and a fair value between them. From November 5 they are blended with the calibrated model layer, family by family, by the skill each layer has shown in El Niño winters.
How a line works
Contract rules- One truth source, one station, one window. Every line names its station's threaded record in RCC-ACIS and its window (e.g. snowfall, July–June; heating degree days, Nov–Mar).
- Frozen before the window opens. The catalogue is fixed before the season starts and never re-priced after the fact.
- Trace counts as zero. A season with too many missing days is marked incomplete and never priced.
- Settlement is on the value published on the named settle date. Later corrections do not reopen a line.
- Base rate: the share of 1991–2020 winters that cleared the strike. Under it, the same winters shifted to the 2026–27 climate by the station's own trend.
- Strong El Niño: how many of the 9 strong El Niño winters since 1950 cleared it, with a 95% Wilson interval.
- Trend-adjusted: the same strong El Niño winters, each moved to the 2026–27 climate by the station's own trend over 1950–2025: a straight-line shift for temperature, degree days and ice days (never below zero); for snow, each winter scaled by how much the station's average snowfall has changed, so a snowless winter stays snowless; precipitation is not adjusted, because a trend there did not improve the backtest.
- Fair value: the trend-adjusted strong El Niño rate pulled toward the trend-adjusted base rate, (k + a·p₀)/(n + a), because nine winters are too few to price on alone. The weight a = 10 was set by the leave-one-winter-out backtest, which also found this trend-adjusted price better than the untrended one (decided September 29, 2026). "Untrended" shows the price without the shift.