We run 120+ ensemble members across 3 models against the live weather books on Polymarket, for the 15 stations we read best. The finished forecast is on your dashboard with every run, and it is filed with its outcome the next morning whether it won or lost.
Over the last 30 days, across the fifteen stations we sell.
Every station is ranked by how often the settling bucket was one of the two we named. Single is the top 3, Pro the top 15.
None of them require anything from you except an account.
122 versions of tomorrow's atmosphere across 3 models, each started from a slightly different guess about today. Where they end up gives the probability. How far apart they end up says how much that probability is worth trusting.
Our probability is compared with what the market is charging at that moment, under the same rounding the exchange settles on. A forecast on its own is trivia. A forecast next to a price is either an opportunity or a reason to stay out.
Every station on your plan, every run, 4 times a day: the mean, the spread, the integer the exchange will read off, our probability and the price. When the gap is wide enough the message says so and the call goes on the record.
Week by week: how often the settling bucket was one of our three, and one of our two, across the 15 stations we sell.
The same three that lead the ranking above, measured the same way.
A top-two rate is not a hit rate on a bet: it says the settling bucket was one of the two we named, and the two are adjacent, so it measures how tightly the forecast is bracketed rather than promising anything about a single position. The full breakdown, every station, is on the track record. Past rates describe what already happened and predict nothing.
Priced in dollars, settled on chain in stablecoin or bitcoin at the rate quoted. When the term ends, access stops.
3 cities you choose, from the 15 we sell. For following a few markets properly rather than fifteen you skim.
All 15 we sell, with the afternoon revisions. Everything we are willing to sell.
No card, no auto-renewal, no retention call.
Single or Pro, one, three or six months. Priced in dollars, from $19.99.
A single-use address and the exact amount, locked for fifteen minutes.
USDC, USDT or bitcoin, on ten networks. You send; we never hold keys.
The network confirms. No card, nothing stored, nothing to cancel later.
The dashboard unlocks and you get your bot code. When the term ends, access ends.
Forecasts are easy. Accountable, priced forecasts are not.
| Capability | Free weather app | Signal group | PolyTemp |
|---|---|---|---|
| Probability, not a single number | No | Rarely | Yes |
| Model spread shown | No | No | Yes |
| Priced against the live book | No | Sometimes | Every call |
| Losing calls kept on the record | N/A | Almost never | All of them |
| The forecast arrives on quiet days too | Yes | No | Every run |
| Takes custody of your money | N/A | Often | Never |
| Auto-renews without asking | Yes | Yes | No card held |
No. PolyTemp publishes forecast data on your dashboard and nothing else. We never take custody of funds, never connect to your exchange account, and never execute a position. What you do with a forecast is entirely your decision.
The fifteen we are best at. Every station is ranked by its verified record, and the top fifteen are the ones we sell. Pro is all fifteen and follows the ranking as it moves. Single is three of them, and you choose which three; each extra city is $9.99 for the term, up to fifteen. What you chose stays yours even if the ranking shifts. The current order is on the track record page for anyone to check.
A phone app gives you one number. We run an ensemble, so you get a probability and a spread, meaning how much the model disagrees with itself, and then we price that probability against the live market under the same rounding the exchange settles on. A forecast without a market price attached is not a signal.
Every forecast is frozen before the day starts and filed with its outcome the next morning, win or lose. The record is public on the track record page before you pay anything, and the station ranking that decides what each plan includes is published with the sample behind every figure.
Prices are in US dollars. At checkout we quote the amount in the coin you picked at the rate of the moment, hold that rate for fifteen minutes, and open your plan as soon as the payment confirms on chain.
Access stops and nothing renews, because we hold no card and no allowance. Your dashboard tells you seven days out, in case you want another term.
Twenty-nine, on four continents, and the plans sell the best fifteen of them. Each one is the specific station the market settles on, which is not always the airport you would guess: London is City, not Heathrow, and New York is LaGuardia, not Central Park. The full list, with the ones outside the top fifteen, is on the coverage page.
Subscribing to forecast data is straightforward almost everywhere; trading event contracts is not. Prediction markets are restricted or banned in several jurisdictions and the rules change often. Check your own before you subscribe. That part is on you, and we would rather say so plainly.
Our best 3 for $19.99, the full dashboard, and a record you can read in its entirety before you decide whether the second month is worth it.