The most common questions, answered.

Four institutional sources, all live: NOAA's GFS at a 16-day horizon and the European Centre's ECMWF at 10 days, both via Open-Meteo; the UK Met Office's UKMO at 7 days, also via Open-Meteo; and hourly NWS data straight from api.weather.gov for US locations. NOAA NCEI's GHCND station archive supplies a 10-year historical prior that the Bayesian step folds into the live consensus.

The probability core uses an Abramowitz-Stegun Normal-CDF approximation. It first weights the four model temperatures (ECMWF 35%, GFS 25%, UKMO 20%, NWS 20%) into a single consensus, then sets sigma as a function of how far away resolution sits (0.8° inside six hours, climbing to 5.5° beyond ten days). Each outcome — exact value, range, "or higher", "or lower" — gets its own CDF integration. Edge is simply our probability minus the market price, and the Bayesian fold-in tightens sigma whenever the models and the historical record line up.

Sizing follows the Kelly criterion (f* = edge ÷ implied odds), then trims to a quarter of that figure for safety, with a hard ceiling of 15% of bankroll on any single trade. The actual order is the smaller of (bankroll × kellyFraction) and whatever maxPerTrade you've set in config. Defaults out of the box: $500 bankroll, $50 maxPerTrade.

Everything settles on Polygon (chain ID 137) through Polymarket's CLOB at clob.polymarket.com. pUSD is the trading asset post-CLOB-V2 (Polymarket auto-wraps USDC.e at deposit), and on the first wallet pair-up the engine signs approvals for the Exchange (0x4bFb41d5), NegRiskExchange (0xC5d563A3) and NegRiskAdapter (0xd91E80cF) contracts using ethers.js together with viem.

Five exits run in parallel and any one of them lifts the position: (1) Profit target — 60% or more of the original edge captured, (2) Convergence — the book has come within 2% of our probability, (3) Trailing stop — at least 40% given back from peak unrealized P&L, (4) Drawdown — position down 15% from entry, (5) Time decay — under two hours from resolution with any open profit. The lift goes straight through CLOB.

Almost always one of five things — walk this list before assuming the engine is broken:

1. Pair the wallet first — orders cannot route without a paired Polygon wallet. Head to Settings and confirm the connection completed.

2. Give the book some time — weather books are not crypto books. Twenty minutes between fires is normal; when conditions converge you may see ten in a row inside two minutes.

3. Verify the proxy (funder) address — a mistyped proxy is the silent killer. If it's off by a single character on-chain execution will never settle.

4. Sanity-check the risk config — open Settings and confirm bankroll, maxPerTrade and minEdge are values you actually meant. While you're there, make sure the proxy holds enough pUSD to fund the order sizes you've configured.

5. Top up POL for gas — this is the most-missed step. Polygon's native token pays the gas; even a single dollar of POL in the wallet is enough for many fills.

Sort of. The platform itself has no subscription and no upfront fee — you only pay a small per-trade levy, and you get to pick the model: $0.015 on every executed order regardless of outcome, or $0.03 only on winners. Pick whichever fits your style.

No, and any tool that says otherwise is selling something dangerous. The engine is actively developed — fresh shipping happens almost every day — and our internal benchmarks tend to land between 18% and 48% over a 24-hour window. But don't budget for instant results: weather markets resolve on weather time, not crypto time, and luck swings outcomes around the mean. Stake only what you can afford to lose.

Yes — close out anything you opened by hand before starting. The exit engine only governs positions it opened itself, because every fill is logged at the moment it happens: the exact entry price, the size, the side (BUY_YES or BUY_NO), the market ID, and the timestamp. That ledger is what makes accurate P&L and clean exits possible.

Positions you opened manually (or through another tool) have none of that on file. The engine would have to back-fit an entry price from historical data, which produces inaccurate P&L and can fire exits at the wrong moment.

Because nothing persists between sessions, here is exactly what happens:

Engine-opened positions — fully governed by the exit monitor, with the real entry on the books and proper auto-lifts.

Hand-opened positions — surfaced in the Open Positions panel for your reference, but the engine will not touch them.

To avoid any ambiguity, close any pre-existing weather positions before pressing Start.

If your mental model is overnight wealth, this is the wrong tool — important to be straight about that. Weather books simply do not move the way crypto books do. Most positions need to wait 24 to 48 hours for the underlying weather to actually happen and the market to settle.

The returns are real, but they compound slowly. The engine is built for repeatable, edge-based gains — much closer to a quietly compounding book than a lottery ticket.

Whether it makes you rich is a function of your bankroll, your settings, how long you let it run and what the market gives you. What we can say is that with realistic expectations and a willingness to let it do its job, the numbers usually speak for themselves over time.

The engine is shipping nearly every day — small fixes, new strategies, deeper instrumentation. Nothing about it is "set and forget."

The current focus is broadening the strategy library and benchmarking the next wave of frontier reasoning models — GPT-5, Gemini 3.1 and DeepSeek R2 — so traders can choose the engine that matches what they care about most: heavier analytical horsepower, faster turn-around, or a more conservative bias.

The pipeline is full and the cadence is fast.

If you've already found StormBot, you're early. No marketing has gone out yet — you're ahead of the crowd.