Micro Weather Sniper Trader
This is a template.
The default signal uses NOAA (US) and Open-Meteo (global) weather forecasts to find mispriced temperature bins on Polymarket -- remix it with additional weather sources, ensemble model averaging, or multi-day position building.
The skill handles all the plumbing (market discovery, forecast fetching, bin matching, trade execution, safeguards). Your agent provides the alpha.
Strategy Overview
Polymarket lists "highest temperature" bin markets for cities worldwide (e.g. "Will the highest temperature in Warsaw be 10°C on April 10?"). These bins are priced by market participants who often lack access to professional weather forecast data.
This skill fetches actual weather forecasts from NOAA (US cities, ~85% accuracy for 1-2 day predictions) and Open-Meteo (global cities, free API, similar accuracy). It then compares the forecast temperature against each market bin:
- If the forecast matches the bin and the market price is low (p <= 20%): BUY YES -- the market is underpricing a likely outcome.
- If the forecast disagrees with the bin and the market price is high (p >= 80%): SELL NO -- the market is overpricing an unlikely outcome.
Example: Open-Meteo forecasts Warsaw high = 10°C on April 10. The bin "Warsaw 10°C" is priced at p=20%. Edge = 0.85 - 0.20 = 0.65. Conviction = 0.65 / 0.85 = 0.76. Size = max($2, 0.76 * $5) = $3.82. The skill buys YES at $3.82.
Edge
Weather forecasts from NOAA and Open-Meteo are publicly available but not widely used by Polymarket participants. NOAA 1-2 day high temperature forecasts have documented accuracy of approximately 85%. This creates a structural information edge:
- The forecast provides a probability estimate (~85%) that the temperature will fall in a specific range
- Polymarket bins are priced by retail participants without systematic forecast data
- When the forecast-implied probability (85%) diverges from market price (e.g. 20%), the difference is pure edge
- Weather markets resolve daily, providing rapid feedback and capital recycling
- Micro-sizing ($2-$5) means each trade has minimal risk while maintaining consistent exposure
Signal Logic
- Discover active weather markets via keyword search across all supported cities plus
get_fast_markets() and get_markets(limit=200) bulk scan
- Parse each market question: extract city name, temperature range (bin), and resolution date
- Fetch NOAA forecast (US cities) or Open-Meteo forecast (global cities) for each relevant city
- For each market: check if the forecast temperature falls within the bin
- If forecast matches bin AND
p <= YES_THRESHOLD (0.40): buy YES -- conviction = (0.85 - p) / 0.85, size = max($2, conviction * $5)
- If forecast disagrees with bin AND
p >= NO_THRESHOLD (0.80): sell NO -- conviction = (p - 0.15) / 0.85, size = max($2, conviction * $5)
- Spread gate: skip if spread >
MAX_SPREAD (15%)
- Place up to
MAX_POSITIONS (10) micro trades per run
- All trades include
signal_data with forecast temperature, city, and edge for backtest support
Remix Signal Ideas
- Ensemble averaging: Combine NOAA + Open-Meteo + AccuWeather forecasts and weight by historical accuracy per city
- Confidence scaling: NOAA is more accurate for D+1 than D+3 -- scale conviction by forecast horizon
- Multi-bin strategy: When forecast is between two bins (e.g. 9.9°C between 9°C and 10°C bins), buy YES on both at half size
- Exit management: Monitor positions and sell when price reaches
EXIT_THRESHOLD (take profit) or forecast changes
- Historical calibration: Track forecast accuracy per city per season to dynamically adjust NOAA_ACCURACY
Safety & Execution Mode
The skill defaults to paper trading (venue="sim"). Real trades only with --live flag.
| Scenario | Mode | Financial risk |
|---|
python trader.py | Paper (sim) | None |
| Cron / automaton | Paper (sim) | None |
python trader.py --live | Live (polymarket) | Real USDC |
autostart: false and cron: null mean nothing runs automatically until configured in Simmer UI.
Required Credentials
| Variable | Required | Notes |
|---|
SIMMER_API_KEY | Yes | Trading authority. Treat as a high-value credential. |
Tunables (Risk Parameters)
All declared as tunables in clawhub.json and adjustable from the Simmer UI.
| Variable | Default | Purpose |
|---|
SIMMER_MAX_POSITION | 5 | Max USDC per micro trade (ceiling at 100% conviction) |
SIMMER_MIN_TRADE | 2 | Floor for any trade (min USDC regardless of conviction) |
SIMMER_MAX_SPREAD | 0.15 | Max bid-ask spread (15%) |
SIMMER_MIN_DAYS | 0 | Min days until resolution (0 = allow same-day weather) |
SIMMER_MAX_POSITIONS | 10 | Max concurrent micro positions |
SIMMER_YES_THRESHOLD | 0.40 | Buy YES when forecast matches bin and p <= this |
SIMMER_NO_THRESHOLD | 0.80 | Sell NO when forecast disagrees with bin and p >= this |
Dependency
simmer-sdk is published on PyPI by Simmer Markets.
