Polymarket Catastrophe Trader

Trades Polymarket prediction markets on hurricane seasons, earthquake probabilities, wildfire forecasts, and extreme weather records. Exploits two structural edges — availability bias correction (retail anchors to vivid recent disasters rather than 40+ years of NOAA base rates) and seasonal data quality timing (signal is only actionable when models are actively running in real time).

Catastrophe & Extreme Risk Trader

This is a template. The default signal is keyword-based market discovery combined with conviction-based sizing and catastrophe_bias() — two structural edges that work without any external API. The skill handles all the plumbing (market discovery, trade execution, safeguards). Your agent provides the alpha.

Strategy Overview

Catastrophe prediction markets are uniquely mispriced because retail traders anchor on the most recent vivid disaster rather than historical base rates. The availability heuristic is the dominant pricing force: the first named storm of the season spikes subsequent storm markets 20–40%, even when NOAA's seasonal forecast hasn't changed by a single storm. After a major wildfire, every "will X state break a fire record?" market overshoots. After a quiet start to a season, markets underprice the base rate. Two structural edges compound:

1. Availability bias correction — NOAA, NHC, NIFC, and USGS publish decades of calibrated base rates. Named Atlantic storm counts have 40+ years of forecasting data. Global temperature records are measured to ±0.01°C simultaneously by three independent agencies. The edge is in knowing these numbers when retail is trading on vibes and recent memory.

2. Seasonal data quality timing — The signal is only actionable when models are actively running. During hurricane peak season (Aug–Oct), NHC issues advisories every 6 hours and model ensembles update in real time. A named-storm-count market in February is priced on stale pre-season data; the same market in September is priced against daily NHC output. The edge doubles when real-time data is flowing.

Signal Logic

Default Signal: Conviction-Based Sizing with Catastrophe Bias

1. Discover active catastrophe and extreme weather markets on Polymarket
2. Compute base conviction from distance to threshold (0% at boundary → 100% at p=0/p=1)
3. Apply catastrophe_bias() — hazard type data quality × seasonal calendar timing
4. Size = max(MIN_TRADE, conviction × bias × MAX_POSITION) — capped at MAX_POSITION
5. Skip markets with spread > MAX_SPREAD or fewer than MIN_DAYS to resolution

Catastrophe Bias (built-in, no API required)

Factor 1 — Hazard Type / Data Quality

Hazard type	Multiplier	The structural reality
Named storm count / above-normal Atlantic season	1.25x	NOAA seasonal outlooks calibrated over 40+ years; above/below-normal ~70% accurate at 90-day lead; retail over-reacts after first storm (20–40% spike)
Global temperature record (hottest year/month)	1.20x	Measured to ±0.01°C by NOAA, Berkeley Earth, NASA GISS simultaneously; trajectory clear months before year-end; retail doesn't check
Billion-dollar disaster count	1.20x	NOAA tracks since 1980; trend clearly upward from climate change + expanding insured assets; retail anchors to average-year intuition
Wildfire season severity (acres burned, state records)	1.20x	NIFC YTD acres vs 10-year average: strong 2–4 week leading indicator; Palmer drought index leads fires by weeks; data public, updated daily
Major hurricane (Cat 3+) landfall	1.10x	NHC 2–5 day track cone probabilities annually verified; retail overprices landfall from visual cone; actual landfall-specific probability far lower
Tornado season record / violent outbreak	1.10x	SPC seasonal outlook reliable at 3-month scale; specific outbreak timing within season harder to predict
FEMA disaster declaration	0.85x	Political and bureaucratic discretion adds real noise beyond meteorological signal
Earthquake (M7+, specific region/window)	0.80x	Fundamentally unpredictable on quarterly timescales; USGS hazard models are long-run annual rates
Tsunami / volcanic eruption	0.75x	Triggered by underlying seismic/geologic events that cannot themselves be predicted; lowest edge in catastrophe markets

The Availability Bias Rule — The first major event of a season creates a retail pricing spike that is almost always an overreaction. The NOAA seasonal forecast before and after that first storm is essentially unchanged, but the market price jumps 20–40%. Fading these spikes — or, better, entering before them — is the core mechanism of the named storm edge. The base rate, not the headline, is the signal.

The Earthquake Exemption — Unlike weather hazards, earthquakes have no seasonal signal and no meaningful short-term forecasting capability. USGS can give you a 1-in-500 annual probability for a M7+ event in a specific fault system. They cannot tell you if it will happen in Q3. Trade earthquake markets at maximum caution (0.80x), and tsunami/volcanic markets at the floor (0.75x).

Factor 2 — Seasonal Calendar Timing

Condition	Multiplier	Why
Atlantic hurricane + Aug–Oct	1.25x	NHC issuing daily advisories; GFS/ECMWF updating every 6h; data richest
Atlantic hurricane + Jun–Jul/Nov	1.10x	Active season; storms possible; below peak frequency
Atlantic hurricane + Dec–May	0.85x	Off-season; no active systems; base rate near zero
Western wildfire + Jul–Sep	1.20x	NIFC daily updates; drought indices current; red flag warnings active
Western wildfire + May–Jun/Oct	1.10x	Fire weather building or receding
Western wildfire + Nov–Apr	0.90x	Most fires absent; markets on stale winter-season data
Tornado alley + Mar–Jun	1.15x	SPC issuing daily outlooks; storm reports accumulating
Tornado + Jul–Feb	0.90x	Off-season; tornado markets thinly priced
Winter storm + Dec–Feb	1.10x	GFS/ECMWF ensemble agreement highest in peak months
Temperature record + Oct–Feb	1.15x	Oct–Dec trajectory clear; Jan–Feb prior-year data finalised

Combined Examples

Market	Type mult	Season mult	Final bias
"Will there be 20+ named Atlantic storms?" — September	1.25x	1.25x (hurricane peak)	1.35x cap
"Will 2026 be the hottest year on record?" — November	1.20x	1.15x (temp record)	1.35x cap
"Will Western US wildfire season exceed 10M acres?" — August	1.20x	1.20x (wildfire peak)	1.35x cap
"Will there be a Cat 5 hurricane landfall by Oct?" — March	1.10x	0.85x (off-season)	0.94x
"Will FEMA declare a major disaster in Florida?"	0.85x	1.0x	0.85x — always cautious
"Will there be a M8.0+ earthquake in Pacific NW by Dec?"	0.80x	1.0x	0.80x — floor territory
"Will there be a Pacific tsunami in 2026?"	0.75x	1.0x	0.75x — near MIN_TRADE

Keywords Monitored

text

hurricane, tropical storm, category 5, Atlantic season, named storm,
tornado, wildfire, acres burned, fire season, earthquake, magnitude,
tsunami, volcanic eruption, 100-year flood, FEMA, disaster declaration,
billion-dollar disaster, polar vortex, bomb cyclone, derecho, heat dome,
hottest year, warmest year, temperature record, blizzard, ice storm,
above-normal season, NOAA, NHC, Category 4, major hurricane

Remix Signal Ideas

• NOAA National Hurricane Center: Named-storm seasonal forecast gives a directly tradeable probability for storm-count markets — compare NOAA's official probability to Polymarket price; the lag after a quiet start to the season can be 15–25%
• NIFC Wildfire Statistics: Year-to-date acres burned vs 10-year average — when YTD is tracking 40% above average by July, "above-normal fire season" markets are structurally underpriced
• USGS Earthquake Hazards API: Real-time seismic data M2.5+ globally — for post-earthquake aftershock markets, the USGS Omori decay law gives probability estimates of M6+ aftershocks within days of a major event
• Berkeley Earth / NASA GISS: Annual global temperature anomaly updated monthly — when October anomaly is already 0.2°C above the prior record, "will 2026 be hottest year?" is a near-certainty the market underprices

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 — nothing runs automatically until you configure it in Simmer UI.

Required Credentials

Variable	Required	Notes
SIMMER_API_KEY	Yes	Trading authority. Treat as 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	25	Max USDC per trade (reached at 100% conviction)
SIMMER_MIN_VOLUME	5000	Min market volume filter (USD)
SIMMER_MAX_SPREAD	0.10	Max bid-ask spread (10%) — catastrophe markets are thinner
SIMMER_MIN_DAYS	7	Min days until resolution — seasonal markets need time to develop
SIMMER_MAX_POSITIONS	7	Max concurrent open positions
SIMMER_YES_THRESHOLD	0.38	Buy YES if market price ≤ this value
SIMMER_NO_THRESHOLD	0.62	Sell NO if market price ≥ this value
SIMMER_MIN_TRADE	5	Floor for any trade (min USDC regardless of conviction)

Dependency

simmer-sdk by Simmer Markets (SpartanLabsXyz)

• PyPI: https://pypi.org/project/simmer-sdk/
• GitHub: https://github.com/SpartanLabsXyz/simmer-sdk