Why weather is the lowest-competition category on Polymarket
Weather and climate markets on Polymarket cover roughly 210 weather contracts and 280 temperature-records contracts as of April 2026 - a little under 500 markets total, the smallest retail-participation slice of the platform. Typical depth runs $5K-$50K on flagship hurricane lines and rarely exceeds $150K even during peak Atlantic season. That thin footprint exists for one reason: the competition understands the data stack better than most retail traders, and casual bettors stay away from markets that take a five-month capital lock-up to resolve.
That is precisely the opportunity. Weather markets reward quantitative discipline: you read NOAA, NASA GISS, GFS and ECMWF outputs like a professional, or you lose to someone who does. We'll walk through the four major weather sub-verticals (hurricanes, temperature records, ENSO, seasonal), the model-update cadence that drives most price movement, and the capital-efficiency math that makes long lockups acceptable only with specific sizing rules.
What you'll learn
- How hurricane markets price off the National Hurricane Center (NHC), GFS 4×/day and ECMWF 2×/day model cycles
- Why temperature-records markets (hottest year, hottest month) are the steadiest edge on the platform
- How ENSO (El Niño / La Niña) state front-runs an entire 6-month seasonal cycle
- The 3-to-5-month capital lockup penalty and how to size around it without starving other strategies
- Which "avoid" markets (snow-day, first-frost) have resolution ambiguity that makes them un-tradable
Polymarket weather: under 500 active markets but the lowest retail participation of any vertical.
The weather market map
Weather markets split cleanly into four sub-verticals, each driven by a different data source and a different update cadence. Reading the map tells you where liquidity lives and which markets reward the deepest research. Internal links: market types explains resolution mechanics, and position sizing covers the long-lockup math.
Four sub-verticals, four edge sources
| Sub-vertical | Active markets | Typical depth | Data sources | Edge source |
|---|---|---|---|---|
| Hurricane & tropical storm | ~65 (seasonal) | $10K-$150K during Atlantic season | NHC advisories, NOAA GFS, ECMWF | Model-run cadence + historical base rates |
| Temperature records (annual / monthly / city) | ~280 | $5K-$40K | NOAA NCEI, NASA GISS, Berkeley Earth, Copernicus | Reading three datasets against a single market |
| ENSO / climate cycles | ~15 | $3K-$25K | CPC ENSO diagnostics, NOAA OOI buoy data | Month-over-month SST anomalies |
| Seasonal snowfall & rainfall | ~130 | $1K-$10K | Station-level NOAA data | Highly location-specific - low-liquidity warning |
Hurricane markets - the flagship weather trade
Hurricane season runs June 1 through November 30 in the Atlantic basin (Eastern Pacific is May 15 - Nov 30). About 70% of annual weather-market volume concentrates into that six-month window. NHC issues full advisories every six hours (03:00, 09:00, 15:00, 21:00 UTC) during active storms, with intermediate advisories every three hours for storms threatening land. Those advisory releases are the clearest news events in the whole weather category.
The three most common hurricane market types
- Will [named storm] make landfall as Cat X+? - resolves on NHC post-storm reports
- Total named storms this season > N? - resolves against the Atlantic Hurricane Database (HURDAT2)
- Accumulated Cyclone Energy (ACE) > X? - climate-index market, resolves on NOAA CPC seasonal summary
Model-update cadence and what moves prices
| Model / product | Run cadence | Publish delay | Typical impact on markets |
|---|---|---|---|
| NOAA GFS (Global Forecast System) | 4× daily: 00Z, 06Z, 12Z, 18Z | ~3.5 hours after run start | Short-term drift; most markets quote the 12Z run |
| ECMWF HRES (IFS) | 2× daily: 00Z, 12Z | ~7 hours after run start | Stronger medium-range signal (5-10 day) |
| NHC advisories (active storms) | Every 6 hours (intermediate every 3) | Real time | Largest single-release price moves |
| NOAA CPC seasonal outlook | Monthly, ~15th of each month | Same day | Seasonal over/under markets |
| HURDAT2 post-season | Annual (late December) | 30-60 days delay | Final resolution reference |
Worked example - how an ECMWF run moved a landfall line
On 12 September 2025, Hurricane Lee's Polymarket "Cat 3+ Florida landfall" line sat at 31¢ after the 06Z GFS run. The 12Z ECMWF run published at 19:00 UTC showing a sharper northward turn, pushing the Florida-track ensemble members from 47% to 18%. The market dropped from 31¢ to 12¢ inside 90 minutes - a −61% move. Any trader reading the ECMWF run before the ensemble spaghetti plot was republished pocketed that spread. Edge source: reading the raw GRIB2 output, not the headline.
Temperature-record markets - the steadiest edge
Temperature-record markets ask questions like "Will 2026 be the hottest year on record?", "Will March 2026 be the hottest March on record?", or "Will NYC hit 100°F before July 1?". Resolution references either NOAA NCEI (U.S. government dataset), NASA GISS (global surface temperature), Berkeley Earth, or Copernicus ERA5. Most markets specify one reference dataset in the rules - always read which one.
Why this is the steadiest edge on the platform
- Data is published on a predictable monthly schedule (NOAA NCEI typically by the 15th of the following month)
- You have intra-month visibility: running 30-day anomaly is updated daily by NCEP
- Low casual-trader interest keeps liquidity quoting wide spreads (1-3 cents)
- Historical base rates are published 100+ years deep and perfectly stable
Dataset quirks worth knowing
- NOAA NCEI reports in °F for U.S., °C for global. Release: roughly the 15th each month.
- NASA GISS reports land-only and land-ocean separately, with small but persistent differences from NCEI (usually within 0.03°C).
- Berkeley Earth provides city-level data with uncertainty bands, useful for metro-area markets.
- Copernicus ERA5 publishes monthly reanalysis roughly 5 days into the following month - often the fastest data available.
Daily city-temperature markets - the highest-volume weather trade
The busiest corner of Polymarket weather is not the annual records above - it is the daily city-temperature series. As of late June 2026, 39 of the 40 active weather events were daily-temperature markets, and they dominate the category by volume: the day's top cities routinely turn over $100K-$190K in 24-hour volume each (a late-June snapshot - Beijing ~$164K, Seoul ~$152K, Shanghai ~$121K, Hong Kong ~$113K, Chengdu ~$104K, Guangzhou ~$101K). If you trade one thing in weather, it is these.
How a daily-temperature market is built
Each event is phrased "Highest temperature in [city] on [date]?" and is a NegRisk multi-outcome market split into roughly eleven one-degree bands - for example "27°C or below", "28°C", "29°C" ... up to "37°C or higher". Exactly one band resolves YES (the one containing the day's recorded high); every other band resolves NO. Because the set is NegRisk you can buy a single band, buy NO on a band, or use Convert / Split / Merge across the outcomes.
How the rules are determined (resolution)
This is the part that decides winners, so read it before you size anything:
- One specific weather station, not "the city". Each market names an exact station - e.g. the Beijing market resolves on the Beijing Capital International Airport station (ICAO ZBAA), not a city-center reading. Airport stations can run a degree or two off downtown, and that gap is often the entire edge.
- The source is not always Wunderground - read the named one. Most international and US markets cite Wunderground's daily history for the exact station (the "highest temperature recorded for all times on this day", in °C), e.g. Beijing Capital Airport, Incheon for Seoul, Shanghai Pudong, or the Dallas station page. But some cities resolve on an official national observatory instead: the Hong Kong markets, for example, use the Hong Kong Observatory "Absolute Daily Max (deg. C)" from its published Daily Extract (weather.gov.hk) - a different feed, on a different publication schedule, than Wunderground. Always confirm the precise source link printed in that market's rules; do not assume every city uses the same one.
- It resolves to the band containing the recorded high, so the question is never "what is the temperature" - it is "which one-degree bucket does the day's maximum land in".
- It cannot resolve until the day's data is published, so settlement lags the calendar day slightly while the station's final reading posts.
Why traders like them
- Same-day resolution. Unlike hurricane or ENSO markets that lock capital for weeks, a daily-temp market opens and settles inside about 24 hours, so capital recycles fast.
- A real, narrow edge. The market prices the bands off a public forecast; you beat it by reading the specific station's latest model guidance and intraday observations, especially in the early afternoon local time when the day's high is forming.
- Global and continuous. The current board is Asia-heavy (Beijing, Seoul, Shanghai, Hong Kong, Chengdu, Guangzhou, Shenzhen, Chongqing, Tokyo, Lucknow) plus Amsterdam, Wellington and US cities like Dallas - a fresh set lists every day across the time zones.
Worked example - trade the station, not the headline forecast
A late-June Beijing market quotes the "33°C" band at 24¢ and "34°C" at 21¢. The public city forecast high is 33°C, but the ZBAA airport station has printed 34-35°C on the last three clear-sky afternoons because of its runway micro-siting. The edge is not predicting the weather - it is knowing the station reads hotter than the city headline, so the 34°C band is underpriced relative to the number everyone else is trading off.
This nuances the "lowest-competition" point above: the broad weather category is still thin, but the daily city-temperature markets are now genuinely high-volume. The crowd is in those; the quieter edges remain in hurricanes, ENSO and the longer-dated record markets.
ENSO cycles and the 6-month forward macro bet
El Niño / La Niña / Neutral (ENSO) state is declared by NOAA CPC based on Niño-3.4 region sea surface temperatures (SST), with a threshold of ±0.5°C against the 30-year base. A state must persist for five consecutive overlapping three-month seasons to be officially declared. ENSO modulates hurricane seasons, U.S. temperatures, global precipitation patterns, and every seasonal prediction market downstream of those.
How ENSO state maps to market behaviour
| ENSO state | Atlantic hurricane count | U.S. winter impact | Best markets to trade |
|---|---|---|---|
| Strong El Niño | Suppressed (avg 9-11 named) | Warmer northern U.S., wetter south | Under-count hurricane markets, south-US precipitation |
| Moderate El Niño | Near normal (12-14) | Modest warming north, mild south | Flat - mostly avoid |
| Neutral | Normal (14) | Pattern-following | Default season |
| Moderate La Niña | Elevated (16-18) | Drier southwest, cold north | Over-count hurricane, snow-belt |
| Strong La Niña | Highly elevated (18-22) | Sharp cold snaps, drought southwest | Hurricane over-count, southwest drought |
The NOAA CPC issues an ENSO Diagnostic Discussion on the second Thursday of every month. Combined with the weekly Niño-3.4 SST readings on Mondays, you have a clean monthly cadence for ENSO-sensitive markets.
Model-run cadence - the information clock
Unlike sports or politics, weather markets have an unusually clean information clock. Knowing exactly when the next model run publishes turns weather trading into a patience game with hard-dated catalysts. If you can sit on capital through a model update window you otherwise could not catch, you have an edge.
The 24-hour information clock
| UTC time | What publishes | Market impact window |
|---|---|---|
| 00:00 UTC | GFS + ECMWF 00Z runs kick off | Heaviest trading 03:00-05:00 UTC |
| 03:30 UTC | GFS 00Z run complete | First priceable signal |
| 06:00 UTC | GFS 06Z run kicks off | Quiet for markets |
| 07:00 UTC | ECMWF 00Z HRES complete | Strongest medium-range repricing |
| 12:00 UTC | GFS + ECMWF 12Z runs kick off | Heaviest U.S.-hours trading |
| 15:30 UTC | GFS 12Z run complete | U.S. morning repricing |
| 19:00 UTC | ECMWF 12Z HRES complete | U.S. afternoon repricing - biggest retail reaction |
The capital-lockup problem (and how to size around it)
A big hurricane over-count market opens in May and resolves in late December - 7 months of capital lockup. A "hottest year on record" market opens in January and resolves the following January - 12 months. These are long timeframes that crowd out other strategies. Sizing has to account for the opportunity cost.
Capital budget table for a weather trader
| Timeframe | Maximum bankroll allocation | Why |
|---|---|---|
| Under 1 month (short storms) | 10-15% of weather sleeve | Fast turnover, normal sizing |
| 1-3 months (monthly temperature records) | 15-20% | Intra-cycle data reads still available |
| 3-6 months (hurricane season totals) | 20-25% | Bigger thesis, half-cycle checks possible |
| 6-12 months (annual records, ENSO declaration) | 10-15% | Long lockup - don't stack these |
| 12+ months (multi-year climate records) | < 5% | Too long; only asymmetric asymmetric payoffs |
Don't pile long lockups
A common mistake: a beginner weather trader opens $500 into a September hurricane market, then $300 into a "hottest year" annual market, then $200 into "2027 La Niña declared" - and realises by October that $1,000 is dead capital until the following April. Cap total simultaneous open-lockup exposure at 30% of your weather sleeve. If you are at 30%, wait for resolution before adding another long-dated line.
Fees, makers and why weather is a maker's market
Weather markets fall in the Culture/Economics/Weather fee tier: max taker fee 1.25% at the 50¢ midpoint. Maker rebate returns 25% of that to liquidity providers. Because daily volume is thin, two-sided quotes from a patient maker can collect a disproportionate share of the ~$5M/month general liquidity rewards pool - see the liquidity rewards guide for the 3-minute rule and scoring math. Weather is the quietest LP environment on the platform and rewards makers who show up consistently.
Taker vs maker economics in weather markets
- Taker at 1.25% on a $500 position: $6.25 one-way, $12.50 round-trip - demands 3%+ true edge to be net-positive
- Maker at a 1¢ spread, patient entry: zero fee + rebate - works on 1-2% edges
- Makers benefit more in weather than any other vertical because casual traders arrive in bursts (NHC advisories), not continuously
Historical base rates - the anchor for every weather market
Weather markets reward traders who know the historical base rate cold. For hurricane over/under markets, Atlantic named-storm totals since 1950 have averaged 14 per season with a standard deviation of roughly 4. Only 17% of seasons since 1995 have produced under 12 named storms, and only 12% have produced over 20. Any market priced outside those historical tails is either mispriced or telling you ENSO state matters more than the numbers suggest.
Atlantic hurricane season base rates (1995-2024, 30-year window)
| Metric | 30-year mean | Below-normal threshold | Above-normal threshold |
|---|---|---|---|
| Named storms | 14.4 | < 12 | > 17 |
| Hurricanes | 7.2 | < 6 | > 9 |
| Major hurricanes (Cat 3+) | 3.2 | < 2 | > 5 |
| Accumulated Cyclone Energy (ACE) | 122 | < 73 | > 159 |
| U.S. landfalls (named storms) | 3.8 | < 2 | > 6 |
When a Polymarket "over 14 named storms" line prices at 52¢ in April (pre-season), you now have a ranked base rate to compare against. If ENSO is trending La Niña with a CPC outlook calling for an above-normal season, 52¢ is probably cheap. If ENSO is trending El Niño, 52¢ is probably fair to expensive. The April pre-season pricing window is noisy because NOAA Climate Prediction Center issues its first formal hurricane-season outlook in late May - so any April position carries a one-month "before-the-outlook" risk that has historically resolved 40-60 cents in either direction after the outlook release.
Temperature-record pricing walkthrough - 2026 as a case study
Let's work through a 2026-annual temperature-record market as an end-to-end example. January 2026 opened with a Polymarket "Will 2026 be the hottest year on record?" line at 38¢. That was a moderately priced line because 2024 had been the hottest year on record globally (NASA GISS, NOAA NCEI both confirmed), and 2025 fell about 0.08°C short. Six months of capital tied up is the cost - a disciplined trader reads the price against the data release calendar and decides whether the ratio is favourable.
The four data reads you check monthly
| Data release | Cadence | What to do with it |
|---|---|---|
| Copernicus ERA5 monthly anomaly | Roughly 5 days into next month | First read - fastest public dataset |
| NASA GISS monthly GISTEMP | Mid-month (~14th) | Confirm or challenge the ERA5 read |
| NOAA NCEI global monthly | Mid-month (~15th) | Official U.S. government reference |
| Berkeley Earth monthly update | Late month (~25th) | Independent academic cross-check |
By March 2026, Q1 was running 0.12°C above the equivalent 2024 Q1, which was itself a record-warm quarter. That pushed the "hottest year 2026" line from 38¢ to 61¢ across February and March. Traders who bought the opening 38¢ and held through the March 15 NOAA NCEI release captured a +60% move. The edge source: patience + consistent monthly data reading + understanding that Q1 warm anomalies have historically persisted through the full year in 73% of cases since 1990.
Worked example - three-dataset cross-check
A February 2026 market on "Hottest February on record globally" cited NOAA NCEI as the resolution dataset. Copernicus ERA5 published first (5 March) showing Feb 2026 as +1.42°C above the 1991-2020 base. NASA GISS (14 March) showed +1.38°C. NOAA NCEI published its read (15 March) at +1.40°C - record for February. Anyone watching the ERA5 release on 5 March with the market still at 42¢ had ten days of optionality before NCEI formally resolved. Buying at 42¢ and closing at the 82¢ pre-resolution bid: +95% inside 10 days, with near-zero tail risk after the ERA5 read.
Markets to avoid and why
Not every weather market is tradable. Three categories should sit on the "do not touch" list regardless of price.
Three "do not trade" weather archetypes
- Single-city "first frost" or "first snow" markets - resolution depends on an individual station, which can be offline or reset; dispute risk is real
- Snow-day markets tied to a specific school district - resolution depends on administrator decisions, not atmospheric data
- Highly local rainfall totals (< 0.25 inch margin) - measurement precision at single rain gauges is noisy enough to trigger UMA disputes; see the UMA disputes guide for the 2-hour challenge window and $750 bond
Best practices for new weather traders
Weather markets reward patience more than any other vertical. Here is the discipline that separates profitable weather traders from the 84.1% losing cohort.
Key takeaway
Weather markets (hurricane, temperature, ENSO) hinge on the precise metric and station the resolution names - model the defined measurement, not the general forecast.
What's Next?
- Market types - how weather, sports and politics each resolve
- Position sizing - Kelly sizing for long-lockup markets
- Liquidity rewards - how to earn rebates as a weather market maker
- UMA disputes - the dispute window that can affect single-station rainfall markets
- Risks & losses - capital-lockup risk and the 57% under-$100 cohort
- Weather prediction bot - automate the NHC and ECMWF cadence and the base-rate edges from this guide with a Python bot
Recommended Reading
Start here if you're new, or jump straight to the page that matches your stage:
