WEATHER · FORECAST CHANGE

Where will it rain on Oct 5, 2026: PBE moves “Miami” down to 41% from 60%

New forecast guidance reached the model between 11:00 UTC and 17:00 UTC on Oct 4. Same model version, later data cutoff: here is exactly which inputs changed.

Published October 4, 2026 at 17:03 UTCModel data as of Oct 4 17:00ZBy the PropBetEdge Predictions Desk1 min read

Photo: Marc Averette · CC BY-SA 3.0 · Wikimedia Commons

60%41%

Miami · pbe-weather-precip@1.1.0 · 11:00 UTC → 17:00 UTC

Move−19 pts

Market at capture48%

Inputs changed4primary

LIVE UPDATE · not part of the original story

Where will it rain on Oct 5, 2026?

Miami · RESOLVED

PBE · live intraday1%−40 pts since publication

KalshiNo two-sided quote

GapNo comparable quote

At KMIA: high so far 90.0°F · latest 82.4°F () · STALE

RESEARCH PBE Oct 6 04:00Z · market

UPDATE · Oct 5 06:20Z PBE 27% pbe-weather-precip-intraday@2.0.0 · data cutoff Oct 5 06:20Z
UPDATE · Oct 5 11:00Z PBE 17% pbe-weather-precip-intraday@2.0.0 · data cutoff Oct 5 11:00Z
UPDATE · Oct 5 22:00Z PBE 7% pbe-weather-precip-intraday@2.0.0 · data cutoff Oct 5 21:20Z

Open the live forecast →

Quick read

pbe-weather-precip@1.1.0 moved “Miami” from 60% to 41% (−19 pts).
Later data cutoff: Oct 4 11:00Z → Oct 4 17:00Z; same model version, so the move reflects inputs, not a model change.
Primary inputs changed: GFS MOS max 6-h chance of rain, National Blend max 6-h chance of rain, GFS MOS chance of rain in the window, National Blend chance of rain in the window.
The Kalshi mid captured with each snapshot went from 43% to 48%.

What changed in the inputs

Between the snapshot captured at Oct 4 11:00Z (data cutoff Oct 4 11:00Z) and the one at Oct 4 17:00Z (cutoff Oct 4 17:00Z), the same model version — pbe-weather-precip@1.1.0 — moved “Miami” from 60% to 41%. The model inputs that changed: GFS MOS max 6-h chance of rain from 47% to 38%; National Blend max 6-h chance of rain from 37% to 20%; GFS MOS chance of rain in the window from 64% to 54%; National Blend chance of rain in the window from 52% to 39%.

EVIDENCEStored model inputs, before and afterFrom the immutable feature snapshots attached to each forecast

Input	Before	After	Primary
GFS MOS max 6-h chance of rain	47%	38%	yes
National Blend max 6-h chance of rain	37%	20%	yes
GFS MOS chance of rain in the window	64%	54%	yes
National Blend chance of rain in the window	52%	39%	yes
Guidance lead time	23h	17h	—

Units: as labeledAs of Oct 4 17:00ZSource: PBE feature snapshots

This page reports which stored inputs differ between two immutable forecasts. It does not claim why the underlying weather guidance changed.

The forecast path

PROBABILITY HISTORY“Miami”: pbe-weather-precip@1.1.0 snapshots vs the marketSame model version only; the dashed line is the stored Kalshi mid

Kalshi mid (stored observations)PBE pbe-weather-precip@1.1.0 snapshots

Units: % probabilityAs of Oct 4 17:00ZSource: PBE forecast archive; Kalshi

Data table

Time (UTC)	Source	Value %
Oct 4 09:15Z	PBE pbe-weather-precip@1.1.0	61
Oct 4 09:30Z	Kalshi mid	41
Oct 4 09:45Z	Kalshi mid	36
Oct 4 10:15Z	Kalshi mid	33
Oct 4 10:30Z	Kalshi mid	32
Oct 4 10:45Z	Kalshi mid	35
Oct 4 11:00Z	PBE pbe-weather-precip@1.1.0	60
Oct 4 11:00Z	Kalshi mid	43
Oct 4 11:45Z	Kalshi mid	40
Oct 4 12:00Z	Kalshi mid	44
Oct 4 12:15Z	Kalshi mid	48
Oct 4 12:30Z	Kalshi mid	48
Oct 4 12:45Z	Kalshi mid	48
Oct 4 13:00Z	Kalshi mid	44
Oct 4 13:30Z	Kalshi mid	43
Oct 4 13:45Z	Kalshi mid	42
Oct 4 14:00Z	Kalshi mid	42
Oct 4 14:15Z	Kalshi mid	40
Oct 4 14:30Z	Kalshi mid	43
Oct 4 14:45Z	Kalshi mid	39
Oct 4 15:00Z	Kalshi mid	39
Oct 4 15:15Z	Kalshi mid	39
Oct 4 15:45Z	Kalshi mid	40
Oct 4 16:00Z	Kalshi mid	40
Oct 4 16:30Z	Kalshi mid	49
Oct 4 16:45Z	Kalshi mid	48
Oct 4 17:00Z	PBE pbe-weather-precip@1.1.0	41
Oct 4 17:00Z	Kalshi mid	48

The Kalshi mid captured with each snapshot went from 43% to 48%.

The two snapshots

	Earlier	Later
Captured	Oct 4 11:00Z	Oct 4 17:00Z
Data cutoff	Oct 4 11:00Z	Oct 4 17:00Z
Guidance runs	NWS GFS MOS 2026-10-04T06Z, NWS National Blend of Models 2026-10-04T06Z	NWS GFS MOS 2026-10-04T12Z, NWS National Blend of Models 2026-10-04T12Z
PBE	60%	41%
Market at capture	43%	48%

Evidence & method

What this story is based on. Every value above was read from these immutable records.

Miami

Model: pbe-weather-precip@1.1.0 · RESEARCH
Snapshot: 991d5488-03c0-4761-ae1d-1fbf9de26bb9
Compared with: f79bd098-9d11-4159-b191-72f5926a11f0
Captured: Oct 4 17:00Z
Data cutoff: Oct 4 17:00Z
Market observation: 48% · captured with the snapshot
Source families: NOAA/NWS via Iowa Environmental Mesonet archive · NOAA Regional Climate Centers ACIS · NOAA/NWS api.weather.gov

RESOLUTION RULE

Resolves on: The Weather Company (weather.com/kalshi) — The Weather Company daily precipitation for CLIMIA (weather.com/kalshi)
Independent check: NWS Daily Climate Report CLIMIA (WFO MFL); NOAA RCC-ACIS USW00012839
Measurement: Total liquid-equivalent precipitation for the climate day 2026-10-05, 00:00-24:00 local standard time (UTC-5)
Rounding: Hundredths of an inch; trace (T) counts as 0.00
Exceptions: Missing daily precipitation value counts as 0 inches (resolves NO) · Preliminary data may be revised; the exchange may hold expiration until a non-erroneous revision, else last fair price · Venue event strike_date (2026-10-06T03:59:59Z) is not the climate-day boundary; the reported daily value governs

If the total precipitation at CLIMIA in Miami in Oct 5, 2026 is strictly greater than 0 inches, then the market resolves to Yes.

MODEL LIMITATIONS

Forecasts are published only before the climate day opens (no intraday observations in v1)
IEM NBM archive gaps -> GFS-only fallback tier
Settles on The Weather Company values; verified against NWS CLI 112/112

Weather · measurable rain — inputs: NWS GFS MOS + National Blend of Models 6-h PoP for the exact climate day, 1991-2020 station climatology (ACIS). This model on the research board →

Market prices are a benchmark only and never enter a PropBetEdge model. Research-stage probabilities; not advice.