Polymarket Weather Edge: Temperature Odds vs Forecast & METAR

Price every bracket of Polymarket's daily high and low temperature markets against real weather data, the same way the markets settle, and see where the order book disagrees with the forecast.

Polymarket lists daily "Highest temperature in NYC on October 30?" style markets for about 50 cities, each with around 11 brackets (68-69°F, 17°C, 78°F or higher, …). This Actor reads those markets, pulls the latest airport reports and forecasts for the station each market settles on, and returns a probability for every bracket next to its bid and ask. Open, public data only. No login, no browser.

$0.003 per analyzed event (about $0.30 for 100 events). Invalid input, unsupported events and API errors are free.

Who it is for

• Weather traders who want a second opinion on every bracket before placing an order.
• Same-day monitoring: once the day has started, the observed high or low so far rules brackets in or out. Run it on a schedule and watch for markets that have not caught up yet.
• Research and dashboards: a tidy table of model vs. market for every city and day, ready for Sheets, a database or an AI agent through the Apify API or MCP.

How it works

1. Markets. Open temperature events come from Polymarket's public Gamma API. The Actor reads the settlement station from each market's rules (for example KLGA for New York, EGLC for London).
2. What has already happened. For days that have started, it fetches that station's METAR reports from NOAA's Aviation Weather Center and applies the market rules: routine reports only (SPECI ignored), the precise tenths-of-a-degree value from the remarks when present, the station's local calendar day, and halves rounded up. The highest (or lowest) value so far is a hard floor (or ceiling) for the result.
3. What is still to come. For US airports it uses the National Weather Service gridpoint forecast, including the official daytime high and overnight low. MET Norway's global forecast is used everywhere and is the only forecast outside the US. Both are nudged toward the latest report, with the correction fading over the following hours.
4. Probabilities. The remaining-day extreme is modeled as a normal distribution around the forecast. Uncertainty grows with lead time and is wider when only the global model is available or the two forecasts disagree. Combined with the observed extreme, this gives the probability of every whole-degree outcome and so of every bracket.
5. Edge. For each bracket: edgeBuyYes = probability - best ask and edgeBuyNo = best bid - probability (buying NO costs 1 - best bid). Brackets with edge at or above Minimum edge are listed as opportunities.

How to use it

1. Pick Cities (or leave empty to scan everything) and how many Days ahead to cover. Or paste specific event URLs.
2. Optionally restrict to highs or lows and adjust Minimum edge and Uncertainty multiplier.
3. Click Start. A full scan of all open events usually takes under a minute.

{
    "cities": ["nyc", "london", "chicago"],
    "kind": "both",
    "maxDaysAhead": 1,
    "minEdge": 0.05
}

Output

One dataset item per event. The Overview view shows the essentials; Detail has everything, including every bracket.

{
    "eventSlug": "highest-temperature-in-nyc-on-october-30-2026",
    "city": "New York City",
    "kind": "high",
    "date": "2026-10-30",
    "unit": "F",
    "station": "KLGA",
    "dayStatus": "upcoming",
    "observed": { "available": true, "reports": 0, "extremeSoFar": null },
    "forecast": { "nws": 69.7, "metno": 67.9, "expected": 69.7, "sigma": 1.64, "source": "nws", "leadHours": 21.1 },
    "modelFavorite": { "bracket": "70-71°F", "probability": 0.41 },
    "marketFavorite": { "bracket": "70-71°F", "price": 0.6 },
    "bestOpportunity": { "bracket": "70-71°F", "side": "NO", "price": 0.41, "modelProbability": 0.59, "edge": 0.18 },
    "brackets": [
        { "label": "70-71°F", "modelProbability": 0.41, "marketPrice": 0.6, "bestBid": 0.59, "bestAsk": 0.61, "edgeBuyYes": -0.2, "edgeBuyNo": 0.18 }
    ]
}

Also saved in the run's key-value store:

• OPPORTUNITIES: one flat, edge-sorted list of every bracket and side above the minimum edge across all events.
• RUN_SUMMARY: events found, analyzed, charged, errors and request counts.

marketPrice is the price Polymarket displays: the bid/ask midpoint, or the last trade when the spread is wider than 10 cents. Edges always use the executable bid and ask.

Reading the results

• Same-day markets are where the data is strongest. Once the high is in, the remaining question is mostly settled by the reports.
• Day-ahead edges are a disagreement, not a guarantee. The market often prices information the public forecasts do not have yet, and a raw global model can run a degree cold or warm at a particular airport. Treat a large day-ahead edge as a prompt to look closer, especially outside the US where only the global forecast is available.
• The default uncertainty was checked against two weeks of settled events: day-ahead market favorites (about 51% implied) won about 57% of the time. Raise the Uncertainty multiplier if you want more cautious probabilities.
• Hong Kong settles on the Hong Kong Observatory daily extract, and a few cities settle on Weather Underground. Those items carry a note; Hong Kong is forecast-only.

Pricing

Pay per event, nothing else:

Event	Price
Temperature event analyzed	$0.003
Actor start	$0.00005

Errors and skipped events are never charged. Max events caps what one run can analyze, and a run stops cleanly when it reaches the maximum total charge you set.

Data sources and attribution

• Market data: Polymarket public Gamma API.
• Observations: NOAA Aviation Weather Center (aviationweather.gov), public domain.
• US forecasts: NOAA National Weather Service API (api.weather.gov), public domain.
• Global forecasts: weather data from MET Norway (api.met.no), licensed under CC BY 4.0.

The Actor sends an identifying User-Agent and stays well below each service's published rate limits. Add your own email in Contact email if you run it very often.

Disclaimer

This Actor is not affiliated with, endorsed by or sponsored by Polymarket, NOAA, the National Weather Service or MET Norway. Probabilities come from a statistical model and public forecasts and can be wrong. Nothing here is financial advice. Prediction markets are restricted in some countries; check what applies to you before trading.