The Polymarket Weather Bot — Live on Polymarket CLOB

Polymarket Weather Bot.
Atmospheric volatility, traded systematically.

StormBot — the AI Polymarket bot for weather markets.

StormBot is the Polymarket Weather Bot: a storm-grade polymarket bot that listens to every tick on Polymarket's weather book, fuses five institutional forecast models (GFS, ECMWF, UKMO, NWS, plus HKO for Hong Kong) in real time, and only fires when the AI signs off. 67+ cities. Quarter-Kelly sizing. Lightning-fast CLOB execution.

Start Trading Whitepaper

Built on Polymarket CLOB · Reasoned by Claude

Live · 30D updated · just now

+$578,418

Members' total realized PnL

Engine attributed

+$418,615

Avg / member

+$661

Win rate

71.4%

Fills · 30D

6,247

Daily PnL · last 8d ▲ 12.4%

Community +24% w/w

857

Active traders this month

New · 7D

+47

Countries

38

Top trader · 30D

+$48,290

Sessions · 30D

12,408

+852

Top Traders →

NYC Highest Temp Mar 15 · BUY_YES 52-53°F Chicago Temp Mar 16 · Edge +12.4% London °C Mar 17 · ECMWF 14.2° Miami Highest Temp · GFS 82.1°F Tokyo Temp · UKMO 18.5°C Paris Forecast · σ=1.3° HIGH agreement Phoenix AZ · NWS 91°F · Kelly 4.2% Berlin Temp · Consensus 12.8°C ±0.6° NYC Highest Temp Mar 15 · BUY_YES 52-53°F Chicago Temp Mar 16 · Edge +12.4% London °C Mar 17 · ECMWF 14.2° Miami Highest Temp · GFS 82.1°F Tokyo Temp · UKMO 18.5°C Paris Forecast · σ=1.3° HIGH agreement Phoenix AZ · NWS 91°F · Kelly 4.2% Berlin Temp · Consensus 12.8°C ±0.6°

Live · Members winning right now

Wins keep landing.

Every position the bot closes in profit across the active trader base flows into this feed. Real members, real fills, real PnL — no paper trades.

857 active traders 71.4% win rate · 30D 6,247 fills · 30D

Recent Big Wins last 90s

0x4ab6…c2a82s ago

Highest temperature in Seoul · YES

+$420
0x8f3a…1d928s ago

Highest temperature in Shanghai · NO

+$185
0xd71c…8b4e14s ago

Highest temperature in Paris · NO

+$1,240
0x2e9b…f05723s ago

Highest temperature in Hong Kong · YES

+$67
0x6c1d…a73831s ago

Highest temperature in London · YES

+$98
0xb084…cf1247s ago

Highest temperature in Singapore · NO

+$312
0x59ee…3a6b1m ago

Highest temperature in Hong Kong · YES

+$54

7 big wins · last 90s View top traders →

By the numbers

A few data points from the engine.

$2.4M

Volume this month · +40% MoM

4,892

Trades executed · 95%+ success

85%

Peak win rate · up from 70%

857

Active traders · growing

67+

Global cities tracked

5

Forecast models (GFS·ECMWF·UKMO·NWS·HKO)

5s

Lowest WebSocket latency

0.25x

Quarter-Kelly position sizing

Architecture

Inside the
Storm-Signal Engine

Under the hood: a Node.js service that maintains a live socket into Polymarket's CLOB, pulls up to five meteorology APIs in parallel, runs a Bayesian probability engine over every outcome, and routes the result through Claude before any capital moves.

Five-Source Ensemble

Live pulls from GFS (NOAA, 16-day), ECMWF (European Centre, 10-day), UKMO (UK Met Office, 7-day), and NWS (US hourly) — plus HKO (Hong Kong Observatory) for Hong Kong, the official gauge that settles HK markets. A confidence-weighted blend (ECMWF 35%, GFS 25%, UKMO 20%, NWS 20%) is computed every cycle, with stragglers more than 1.5σ off the mean automatically halved; for Hong Kong, HKO leads at 50% and tapers by lead time.

Probability Core

A horizon-aware sigma (0.8° at six hours out, scaling to 5.5° beyond ten days), Normal-CDF probability per outcome, and a NOAA NCEI 10-year prior folded in via Bayesian update. Trade size is quarter-Kelly with a hard ceiling of 15% of bankroll.

Lightning CLOB Stream

Persistent socket to Polymarket CLOB. Each price-change retriggers the edge math instantly — under 2 seconds from tick to order.

Claude Final Filter

Every candidate is reviewed by Claude with full briefing — model spread, climatology, books, history. Returns BUY_YES, BUY_NO or SKIP plus confidence and sizing.

67-City Coverage

From NYC and London to Tokyo, Paris, Berlin, Sydney, Dubai. Parser reads city/date from each market title and pulls localized forecasts in °F or °C.

Five-Way Exit Discipline

Five independent triggers monitor every open position. Whichever fires first lifts the position straight through CLOB.

60% edge captured

<2% edge collapsed

40% trailing stop

15% drawdown

2h time decay

Hong Kong Resolution-Source Lock

Polymarket settles Hong Kong markets on one number — the official daily maximum at the Hong Kong Observatory (HKO). So for Hong Kong, HKO is wired in as the dominant model and its HQ gauge is polled live every few minutes on resolution day. A daily max only ratchets upward, so the moment it crosses a threshold that outcome is locked YES — the engine buys it while it is still mispriced (35–83¢ band), then runs a mirror play once the afternoon peak has passed.

From tick to trade

Scan → Reason → Strike
three loops, always running

Pair a Polygon wallet, dial in your risk settings, hit Start. From that moment the engine runs three concurrent loops in the background — one finding markets, one reasoning about them, one striking the order book — until you tell it to stop.

01

Discovery & Forecast Pull

The Gamma API is queried for every weather-tagged event (paginated up to 400). Each title is parsed for city, date and unit; coordinates looked up; then GFS, ECMWF, UKMO and NWS are fetched in parallel — plus HKO (Hong Kong Observatory) for Hong Kong, the official resolution gauge — alongside a 10-year climatology snapshot from NOAA NCEI.

02

Edge Math & Claude Sign-Off

The probability core builds a confidence-weighted consensus, derives a horizon-aware sigma, and runs a Normal-CDF sweep over every outcome. Claude reviews the briefing and replies with direction, outcome label, confidence tier, reasoning, and size.

03

Stream Listen & Auto-Strike

A persistent CLOB socket re-fires the edge math on every tick. If the absolute edge crosses the dynamic floor (3-8%), the trader sends a Fill-Or-Kill market order, falling back to a GTC limit with 2% slippage if signing fails. The five-way exit monitor takes over.

FAQ

The most common questions, answered.

Forecast models, edge math, position sizing, exits, blockchain, fees, troubleshooting — every recurring question, written out in plain language.

Open the FAQ

Trade the Storm.
Live on Polymarket.

Pair a Polygon wallet, fund it with pUSD, and let five institutional models plus Claude do the work of finding mispriced weather books.

Launch StormBot

Live Dashboard

StormBot Stats

Real-time telemetry exposed through /api/state. Every internal surface — discovered markets, model forecasts, signals, Claude analyses, socket health, open positions and fill history — is available over the same Express API the dashboard reads from.

API Endpoints

14

REST routes: /api/state, /api/scan, /api/trade, /api/bot/start, /api/bot/stop, /api/wallet/connect, /api/ai/analyze, and more

Weather Models

5

GFS (NOAA), ECMWF (European), UKMO (UK Met), NWS (US), and HKO (Hong Kong Observatory, HK resolution source) — fetched in parallel via Promise.all()

Cities Supported

67

NYC, London, Tokyo, Paris, Berlin, Seoul, Dubai, Sydney, Mumbai, Istanbul, São Paulo, and 56 more

Top Traders Tracked

10

Leaderboard wallets fetched from data-api.polymarket.com — PnL, win rate, positions, activity

Edge Threshold

3-8%

Dynamic: max(0.03, min(0.08, σ × 0.02)) — tighter when forecast is confident

Forecast Refresh

30 min

Background interval re-fetches all weather APIs and reloads StreamEngine with fresh data

AI Analysis Cycle

5-15 min

Claude Sonnet re-analyzes top 15 markets with full forecast + trade history context

Historical Lookback

10 yrs

NOAA NCEI GHCND station data for same calendar date — provides Bayesian prior

Engine Modules

StormBot Engines

The platform is composed of six small engine files, each owning one stage of the pipeline. Pure ES modules on Node.js — no build step, no UI framework dependency.

StreamEngine `stream.js`

Real-time WebSocket connection to wss://ws-subscriptions-clob.polymarket.com. Subscribes to all weather token IDs, handles price_change/last_trade_price/book events. On every tick: recalculates edge via inline Normal CDF, checks 5 exit conditions on open positions, fires trade signals.

CoreWebSocket

AIAnalyzer `ai.js`

Sends each market to Claude Sonnet with full context: question, YES/NO prices, all 4 forecasts, consensus temp, model agreement, hours-to-resolution, historical data, edge analysis, recent trade history. Returns structured JSON: decision, outcome label, confidence, reasoning, edge_pct, suggested_size_pct.

CoreClaude API

EdgeCalculator `edge.js`

Production probability engine. calculateSigma() maps forecast horizon to uncertainty (0.8° to 5.5°). calculateConsensus() does weighted blending with outlier detection. calculate() runs Normal CDF for each outcome type, applies Bayesian historical blending, outputs edge, signal, confidence, Kelly fraction, EV.

CoreMath

ForecastEngine `forecast.js`

Fetches from up to 6 real sources: GFS via Open-Meteo (16-day hourly), ECMWF via Open-Meteo (10-day), UKMO via Open-Meteo (7-day), NWS via api.weather.gov (US-only hourly + alerts), HKO via the Hong Kong Observatory Open Data API (Hong Kong only — the official resolution gauge, 9-day forecast + live running max), and NOAA NCEI CDO (10-year historical TMAX). All in parallel.

CoreWeather API

Trader `trader.js`

Connects to Polymarket CLOB via @polymarket/clob-client with ethers.js wallet. Derives API keys, auto-approves pUSD and CTF tokens on Exchange, NegRiskExchange, NegRiskAdapter contracts. Executes FOK market orders first; falls back to GTC limits with 2% slippage on signature errors.

CorePolygon

MarketDiscovery `markets.js`

Queries Polymarket Gamma API for weather-tagged events (up to 400 via pagination). Falls back to keyword search. Parses market titles to extract city (67-city map), date (4 regex patterns), temp unit, and market type. Parses outcomes into ranges, thresholds, exact values.

CoreGamma API

Documentation

StormBot Docs

The full operator's manual: setup, environment, every API route, every config switch, and the end-to-end execution path from market discovery through on-chain settlement.

Quick Start

git clone <repo>
cd stormbot
npm install          # express, ethers, @polymarket/clob-client, viem, ws, @anthropic-ai/sdk
npm start            # starts Express server on port 3778
# Open http://localhost:3778 in your browser

Environment Variables

ANTHROPIC_API_KEY=sk-ant-...    # Claude API key for AI analysis
NOAA_CDO_TOKEN=XRHLdkB...      # NOAA CDO token for historical data
PORT=3778                        # Server port (default 3778)

API Endpoints

Method	Route	Description
GET	/api/state	Full bot state: markets, signals, logs, streaming stats, AI analyses, positions
POST	/api/scan	Trigger market discovery + forecast fetch + edge calculation
POST	/api/wallet/connect	Connect Polygon wallet, auto-approves pUSD + CTF tokens
POST	/api/wallet/disconnect	Disconnect wallet and clear credentials from Store
POST	/api/trade	Execute trade: tokenId, side, amount, negRisk, tickSize, kellyFraction
POST	/api/bot/start	Start full pipeline: scan → stream → AI cycle → auto-trade
POST	/api/bot/stop	Stop streaming, clear timers, disconnect WebSocket
POST	/api/config	Update bankroll, maxPerTrade, minEdge, autoExecute, anthropicApiKey
POST	/api/ai/analyze	Analyze single market with Claude (pass marketId)
POST	/api/ai/analyze-top	Batch analyze top N markets (default 15)
GET	/api/traders	Top 10 weather trader leaderboard with PnL, positions, activity

Configuration Options

Setting	Default	Description
bankroll	$500	Total capital available for trading
maxPerTrade	$50	Maximum single trade size. Capped at bankroll
minEdge	5%	Minimum absolute edge required to generate a signal
autoExecute	true	Auto-execute trades when edge detected or AI recommends
aggressiveMode	false	Lower thresholds: 3% min edge, BUY_NO above 11¢, MEDIUM accepted
scanIntervalSec	300	Seconds between background market rescans

Multi-User Architecture

The Store module (store.js) keeps per-user state in a single JSON file at /data/users.json. Each record holds: hashed credentials and an auth token, a wallet block (encrypted PK plus proxy address), the user's risk config, counters, AI analyses, fill history and logs. BotManager spins up a dedicated StreamEngine, Trader and AI cycle per user while sharing the underlying market data layer.

Top Traders

Trader Leaderboard

The engine keeps a running watch on the ten most profitable weather wallets on Polymarket via /api/traders. The data is hydrated live from the data-api leaderboard, the Gamma profile endpoint, and the positions/activity feeds.

How the Tracking Works

For each wallet, the /api/traders endpoint hits four upstream data sources in parallel: Polymarket's leaderboard API (the authoritative source for PnL and volume), the Gamma profile endpoint (username and avatar), the positions API (all resolved and active positions with cashPnL), and the activity API (the trailing 30 trades). The merged result is sorted by total PnL, descending.

Wallets on the Watch List

The engine tracks the ten most profitable weather wallets on Polymarket, hard-coded in cities.js. The list is curated by cumulative PnL and weather-market win rate — not just raw volume.

Data Points Per Trader

Profile — username, profile picture (from leaderboard or Gamma)
PnL — total profit/loss from leaderboard API
Volume — total trading volume in USD
Win Rate — calculated from positions: (winning / total) × 100
Active Positions — current open positions with size > 0
PnL History — cumulative chart from sorted resolved positions
Recent Trades — last 5 trades with market title, side, size, price

AI-Augmented Copy Signals

In auto-execute mode, Claude evaluates the exact books these top wallets are active in. The model receives the full forecast briefing, our edge math, and recent fill history — including a flag for whether the engine is already holding the position so it does not double-enter. The result is a copy-signal layer informed both by the activity of the best traders and by Claude's independent meteorological reasoning.

Anti-Churn Guardrails

Claude's prompt always carries the recent fill history for the market under review. The model is explicitly instructed to SKIP if the engine just exited the same outcome at a profit and the price has not moved, to be cautious about re-entering after a loss unless the inputs have meaningfully changed, and to avoid bouncing in and out of the same outcome.

Technical Deep-Dive

StormBot Whitepaper

The full mathematical specification: probability core, ensemble weighting, sigma dynamics, the Bayesian fold-in, and quarter-Kelly sizing — written for traders who want to verify every step.

1. Consensus Temperature Algorithm

The EdgeCalculator.calculateConsensus() function computes a quality-weighted forecast temperature from up to 5 models. Base weights: ECMWF 0.35 (best global model), GFS 0.25, UKMO 0.20, NWS 0.20 — and for Hong Kong, HKO 0.50 (the resolution gauge), tapered by lead time. Outlier penalization: if a model's forecast deviates more than 1.5× the standard deviation from the raw mean, its weight is halved.

Model agreement is scored in 5 tiers based on inter-model standard deviation: VERY_HIGH (<0.5°), HIGH (<1.0°), MODERATE (<2.0°), LOW (<3.5°), VERY_LOW (≥3.5°).

2. Dynamic Sigma (σ) Calculation

Forecast uncertainty σ is the most critical parameter. Base σ from forecast horizon: 0.8° (≤6h), 1.0° (≤12h), 1.3° (≤24h), 1.8° (≤48h), 2.2° (≤72h), 3.0° (≤120h), 3.8° (≤168h), 4.5° (≤240h), 5.5° (beyond). Model count penalty: 1.0× (≥4), 1.1× (3), 1.25× (2), 1.5× (1). Spread penalty: 1.0 + (stdDev / 5.0). Final σ clamped to [0.5, 8.0].

3. Bayesian Historical Blending

When NOAA NCEI historical data is available, the forecast is Bayesian-combined with the 10-year historical distribution for the same calendar date. Combined precision = forecast_precision + historical_precision, where precision = 1/σ². If forecast and history disagree by more than 1.5 × historical_σ, the combined σ is widened by 30% to account for unusual conditions.

4. Probability Calculation

For each market outcome, the Normal CDF (Abramowitz-Stegun approximation) calculates the probability that the actual temperature falls within the outcome's range:

Exact value: P = CDF(52.5) − CDF(51.5) using ±0.5° bin
Range: P = CDF(53.5) − CDF(51.5)
"or higher": P = 1 − CDF(54)
"or lower": P = CDF(50)

Edge = forecastProb − marketPrice. Trades signaled when |edge| exceeds max(0.03, min(0.08, σ × 0.02)).

5. Kelly Criterion Sizing

For BUY_YES: kellyFraction = edge / ((1/marketPrice) − 1). For BUY_NO: |edge| / ((1/noPrice) − 1). Quarter-Kelly applied (×0.25) and clamped to [0, 0.15]. The bot never risks more than 15% of bankroll on a single trade; typical positions are 1-5%.

6. Trade Execution

The Trader class connects via @polymarket/clob-client with signature type 2 (POLY_GNOSIS_SAFE). Primary: FOK (Fill-Or-Kill) market order for immediate fills. Fallback: GTC limit at best ask/bid + 2% slippage, rounded to tick size. Dead orderbooks (404) cached and skipped. Concurrent execution on same tokenId prevented via Set lock.

7. Claude AI Prompt Engineering

Each market is analyzed by Claude Sonnet with a structured prompt: market question, YES/NO prices, all model forecasts (5 for Hong Kong, where HKO is flagged as the resolution source), consensus temp, model agreement, hours until resolution, historical averages, edge analysis per outcome, and recent bot trade history. Two rule sets: BALANCED (5% min edge, prefer BUY_YES at 5-40¢) and AGGRESSIVE (3% min edge, BUY_NO above 11¢). Output enforced as JSON with exact outcome labels.

Security & Architecture

StormBot Security

How the engine treats your keys, which contracts it touches, and the layered risk controls that govern every order it routes.

Wallet & Key Handling

Wallet handling lives entirely on the client. Your private key is used locally to derive the address and to sign the Polymarket API credential trio (key, secret, passphrase) through the ClobClient. Signature type 2 (POLY_GNOSIS_SAFE) is used with the proxy address as funder. The key never crosses the network and is never persisted on any server we operate.

Auto-Approved Contracts

Contract	Address	Approval
pUSD	0xC011a7E12a19f7B1f670d46F03B03f3342E82DFB	ERC20 approve(MAX) — V2 collateral
CTF	0x4D97DCd97eC945f40cF65F87097ACe5EA0476045	ERC1155 setApprovalForAll
Exchange	0x4bFb41d5B3570DeFd03C39a9A4D8dE6Bd8B8982E	pUSD + CTF
NegRiskExchange	0xC5d563A36AE78145C45a50134d48A1215220f80a	pUSD + CTF
NegRiskAdapter	0xd91E80cF2E7be2e162c6513ceD06f1dD0dA35296	CTF only

Balance Monitoring

On wallet connect, the bot checks POL balance (for gas) and pUSD balance (on proxy address) via Polygon RPC. Warnings if POL < 0.01 or pUSD < $1. Before every trade, the bot verifies the trade amount doesn't exceed available pUSD and tracks balance locally after each fill.

Risk Management Stack

Quarter-Kelly sizing — never risks more than 25% of the optimal bet size
15% max fraction — hard cap on any single position relative to bankroll
maxPerTrade cap — configurable dollar limit per trade (default $50)
Dead orderbook detection — skips tokens that return 404
Concurrent execution lock — prevents double-ordering on same tokenId
Price range filter — skips markets where YES is below 2¢ or above 98¢
Time filter — skips markets resolving within 2 hours
AI anti-churn — Claude receives recent trade history
5-layer exit system — profit, convergence, trailing, drawdown, time

Disclaimer

StormBot is experimental software. Polymarket trading carries real financial risk — you can lose some or all of what you stake. Forecasts are probabilistic, not promises. AI analyses are outputs, not advice. Wallet operations execute on the client, so the keys never reach a server. You are solely responsible for complying with the regulations that apply where you live, and prior results never guarantee future outcomes.

Legal

Privacy Policy

How StormBot treats your wallet data, your trading activity, and any information that touches the platform.

1. What Data the Platform Touches

StormBot runs client-side. We do not store private keys, wallet credentials or personal data on any server. The data below is held locally in your browser or on infrastructure you control:

Wallet data — Polygon wallet address and proxy address used to sign and route on-chain orders. Private keys never leave the device.
Trading data — fill history, AI analyses, open positions, edge detections and risk config live locally.
Telemetry — scan/trade/tick/edge counts kept locally for the session.

2. How That Data Is Used

Wallet credentials sign and route Polymarket orders on Polygon.
Risk config drives position sizing, edge thresholds and the automation envelope.
Recent fill history is fed to Claude as context to prevent churn.
Logs and counters render into the dashboard for live monitoring.

3. Data Storage

All data is handled client-side. StormBot does not store user data on any remote server or cloud database. There is no third-party analytics and no external data transmission beyond the APIs required for bot operation.

4. Third-Party Services

Polymarket — CLOB, Gamma, WebSocket, Data APIs for market discovery, trading, streaming.
Open-Meteo — Weather forecast API for GFS, ECMWF, UKMO. No auth required.
NOAA / NWS — US weather (api.weather.gov) and historical climate (ncei.noaa.gov).
Hong Kong Observatory (HKO) — Open Data API (data.weather.gov.hk) for Hong Kong forecasts and the live resolution-gauge reading. No auth required.
Anthropic — Claude API for market analysis. Your API key is used; market data sent.
Polygon RPC — Public Polygon node for balance checks and approvals.

5. Data Retention

Logs capped at 200 entries per user (oldest discarded). Trade history capped at 100. AI analyses at 20. Live edge detections at 50. Persists until the server data directory is manually cleared.

6. Security

StormBot operates client-side — your private keys are never transmitted to or stored on any server. All signing and trade execution happens locally on your device.

7. Your Rights

You can disconnect your wallet at any time via the dashboard. Disconnecting clears all associated session data.

8. Changes to This Policy

This privacy policy may be updated as StormBot evolves. Changes will be reflected on this page.

Last updated: March 2026

Legal

Terms of Service

The terms that govern your use of StormBot.

1. Acceptance of Terms

By accessing or using StormBot ("the Service"), you agree to be bound by these Terms of Service. If you do not agree, do not use the Service. StormBot is a polymarket weather trading bot that executes trades on Polymarket prediction markets using automated strategies.

2. What the Service Does

Locates weather prediction markets on Polymarket via the Gamma API
Pulls live forecasts from the GFS, ECMWF, UKMO and NWS models — plus the Hong Kong Observatory (HKO) for Hong Kong, the official resolution gauge
Computes probabilistic edge through Bayesian statistics and Normal-CDF integration
Routes each candidate through Anthropic's Claude for a trade/no-trade verdict
Settles fills on Polymarket's CLOB via the Polygon blockchain
Watches every open position against five independent exit triggers

3. Risk Disclosure

Polymarket trading carries real financial risk. You can lose some or all of the capital you put in.

Market risk — weather books are volatile and prices can move quickly.
Model risk — every forecast is a probability estimate, not a promise.
AI risk — Claude reasons over the data given; not infallible.
Execution risk — orders can fail from thin liquidity, network congestion.
Smart-contract risk — interacting with Polymarket contracts inherits chain risk.
Software risk — experimental code; defects can translate to financial loss.

4. Not Financial Advice

Nothing surfaced by StormBot — signals, AI write-ups, edge numbers, fills — constitutes financial, investment or trading advice. They are algorithmic outputs. Every trading decision and its consequences are yours alone.

5. Eligibility

You are of legal age in your jurisdiction
You are not located in a jurisdiction where prediction market trading is prohibited
You have read and agree to Polymarket's Terms of Service
You understand the risks of cryptocurrency trading and blockchain interactions
The funds you use are not derived from illegal activity

6. User Responsibilities

Securing your device and browser environment
Protecting your private key and wallet credentials
Monitoring your bot's activity and positions
Ensuring adequate POL for gas and pUSD for trading
Configuring appropriate bankroll, maxPerTrade, minEdge for your risk tolerance
Complying with all applicable laws and regulations

7. Limitation of Liability

StormBot is provided "as is" and "as available" without warranties. To the maximum extent permitted by law, the developers shall not be liable for any direct, indirect, incidental, special, consequential, or punitive damages arising from your use of the Service.

8. Intellectual Property

StormBot's source code, algorithms, documentation, and website design are the intellectual property of the StormBot team. The polymarket weather trading bot name and branding are proprietary.

9. Termination

You may stop using StormBot at any time by disconnecting your wallet. We reserve the right to modify, suspend, or discontinue the Service at any time.

10-12. Modifications · Governing Law · Contact

Terms may be updated at any time. Continued use after modifications constitutes acceptance. Governed by applicable law; disputes resolved through binding arbitration. Contact via the project's GitHub repository or community channels.

Last updated: March 2026

Polymarket Weather Bot. Atmospheric volatility, traded systematically.

StormBot — the AI Polymarket bot for weather markets.

Wins keep landing.

A few data points from the engine.

Inside theStorm-Signal Engine

Five-Source Ensemble

Probability Core

Lightning CLOB Stream

Claude Final Filter

67-City Coverage

Five-Way Exit Discipline

Hong Kong Resolution-Source Lock

Scan → Reason → Strikethree loops, always running

Discovery & Forecast Pull

Edge Math & Claude Sign-Off

Stream Listen & Auto-Strike

The most common questions, answered.

Trade the Storm.Live on Polymarket.

StormBot Stats

API Endpoints

Weather Models

Cities Supported

Top Traders Tracked

Edge Threshold

Forecast Refresh

AI Analysis Cycle

Historical Lookback

StormBot Engines

StreamEngine stream.js

AIAnalyzer ai.js

EdgeCalculator edge.js

ForecastEngine forecast.js

Trader trader.js

MarketDiscovery markets.js

StormBot Docs

Quick Start

Environment Variables

API Endpoints

Configuration Options

Multi-User Architecture

Trader Leaderboard

How the Tracking Works

Wallets on the Watch List

Data Points Per Trader

AI-Augmented Copy Signals

Anti-Churn Guardrails

StormBot Whitepaper

1. Consensus Temperature Algorithm

2. Dynamic Sigma (σ) Calculation

3. Bayesian Historical Blending

4. Probability Calculation

5. Kelly Criterion Sizing

6. Trade Execution

7. Claude AI Prompt Engineering

StormBot Security

Wallet & Key Handling

Auto-Approved Contracts

Balance Monitoring

Risk Management Stack

Disclaimer

Privacy Policy

1. What Data the Platform Touches

2. How That Data Is Used

3. Data Storage

4. Third-Party Services

5. Data Retention

6. Security

7. Your Rights

8. Changes to This Policy

Terms of Service

1. Acceptance of Terms

2. What the Service Does

3. Risk Disclosure

4. Not Financial Advice

5. Eligibility

6. User Responsibilities

7. Limitation of Liability

8. Intellectual Property

9. Termination

10-12. Modifications · Governing Law · Contact

Polymarket Weather Bot.
Atmospheric volatility, traded systematically.

Inside the
Storm-Signal Engine

Scan → Reason → Strike
three loops, always running

Trade the Storm.
Live on Polymarket.

StreamEngine `stream.js`

AIAnalyzer `ai.js`

EdgeCalculator `edge.js`

ForecastEngine `forecast.js`

Trader `trader.js`

MarketDiscovery `markets.js`