PredictCast on wxmap.io · Research blog

Market tape · August 20, 2026

Las Vegas 5:58 PM — the 1° that was not small

A 107–108°F YES filled at 71¢ two minutes before the posted Las Vegas flip-safe clock. Twelve minutes later METAR running max printed 109°F. The contract went to zero. Informational only

Not financial advice. Personal trading journal about public Kalshi weather tape and NWS climate products. Not affiliated with Kalshi or the NWS.
Bottom line

This was a late desert spike on a 109°F CLI day, not a “models cooled” event. The fill at 5:58 PM CDT was 3:58 PM PDT — two minutes before Las Vegas flip-safe 4:00 PM PDT / 6:00 PM your time. Hourly METAR still showed ~107°F while the climate MAX later listed 109°F at 3:00 PM PDT. A 1°F rise at 107–109°F is a smaller percentage of the temperature than 1°F at 60°F, but it is a full bracket width on Kalshi’s 2° ladder. Advice that treated a 0.5° ensemble wiggle as “cooled” while the airport was still making highs was the wrong narrative.

5:58 PM CDT
YES fill 107–108 @ 71¢
3:58 PM PDT
2 min before 4:00 PM safe
109°F @ 3:00 PM
CLI climate MAX (PDT)
6:10 PM CDT
R5 bust push · run 109°F

Was it past the safe time?

No — it was two minutes early in city-local time, and the desk UI already said so. Las Vegas flip-safe hour from the 30-year ASOS study is 16:00 PDT (warm season). The card showed safe 4:00 PM PDT · 6:00 PM your time. The operator was on Central Daylight Time. 4:00 PM PDT is 6:00 PM CDT. The Kalshi fill notification is 5:58 PM — that is 3:58 PM in Las Vegas.

The integer-hour gate treats 16:00–16:59 as “ready.” At 15:58 PDT the hour is still 15, so a city-clock implementation should have said WAIT. Buying at 5:58 PM CDT was either jumping the displayed 6:00 PM line, or treating “close enough” as ready. Either way, the posted clock was not yet past.

Even the research 16:00 hour is thin for this kind of day. KLAS JJA flip risk at 16:00 is about 3.4%, and the p90 hour of max is already 16:00 — meaning one in ten summer days is still making the high after 4 PM local. Extreme heat in the Mojave is exactly where that tail lives.

The twelve minutes

Why “ensemble cooled” was the wrong sentence

R2 fires when cycle-to-cycle μ falls by at least 0.5°F and you are long a warmer-than-μ bracket. That threshold is a fixed degree. At 60°F, 0.5°F is about 0.8% of the temperature. At 109°F it is about 0.5%. The atmosphere in a 109°F boundary layer does not “lock” because the blend ticked from 108.2 to 107.7.

Meanwhile the live METAR was still sitting on the running max. That is the opposite of a cool-down: the high is still being written. A liquidate that says “ensemble cooled” while the thermometer is making new highs is advice people correctly distrust — and then the bust push arrives one minute later, too late to sell 71¢ paper.

Member tails matter here too. If a cluster of GFS/ECMWF members still sit at or above 109 while the mean drifts down, the mean is not a veto of 107–108 risk. It is a reminder that the warm solution is alive.

CLI vs hourly METAR, again

Same family of error as the Central Park whale tape and the Denver flash crash: settlement is the climate MAX, not the last METAR. On August 20 the CLI peak time (3:00 PM PDT) is almost an hour before the fill. If that 109°F had been in the running max the desk uses, 107–108 would already have been R5-dead and R3 would not have been a low-edge buy (target would be 110, not 108).

The 109°F lived in the sub-hourly climate sensor trail. Hourly METAR lagged at 107. Advice that arms off hourly METAR on a 107°F Las Vegas afternoon is trading with a delayed thermometer.

What happened
−100% of the 71¢ YES
Entered 107–108 two minutes before city-safe time, on a still-rising 107°F METAR, on a CLI 109°F day. Bust to 109°F at 6:10 PM CDT.
What the new rules do
No buy · earlier exit
Desert + ≥100°F waits until 17:00 local. Still-making-highs blocks the low-edge buy. 0.5°F Δμ is not a cool signal at 108°F. R5 now liquidates when you are on the ceiling and still heating, not only after 109 prints.

What we changed in the advice engine

  1. Heat-scaled trend. R1/R2 |Δμ| scales with temperature so 0.5°F at 60°F is the reference. At 109°F you need ~0.9°F of ensemble move before we call it a cool/warm cycle.
  2. Desert extreme-heat clock. LAS/PHX on a ≥100°F running max use flip-safe 17:00 local, not 16:00.
  3. Still making highs. If live temp is within 0.5°F of the running max before 18:00 local on desert/extreme-heat days, R3 will not BUY the low-edge 2° ticket — including 107–108 at 71¢ with run 107.
  4. Warm member tail. If μ+σ (or the CI high) sits above the focus ceiling, that is a veto, not a “mean cooled so buy the cool bin.”
  5. Ceiling + still heating = R5. Holding 107–108 when run is already 108°F and live is still there is a one-tick bust. Liquidate then, not after 109.
  6. R2 copy. If obs are still heating, the headline is no longer “ensemble cooled.” It says the high is still being written.

None of this makes 109°F Vegas afternoons safe. It stops treating 1°F at 109°F like a rounding error, and it stops arming the obs-edge buy while the desert is still adding heat.

What we still cannot promise

A 107→109 jump in a single METAR minute will always outrun a human. Sub-hourly CLI can lead hourly METAR. Flip-safe hours are climatology, not a force field. This post is a record of one afternoon and a change to the rules that produced the card — not a claim that the next 109°F day will be tradable.

Informational only — not financial advice. Not affiliated with Kalshi or the NWS.