How weather markets actually resolve: every station, audited
A temperature market resolves against one thermometer. Not the city, not an average, not what your phone says. One instrument, at one airport, reported to a fixed precision, over the local calendar day. Get any part of that wrong and you can be right about the weather and still lose. We read the resolution text for all thirty nine live temperature markets and tabulated what each one actually settles on. Four of them settle somewhere most people would not guess.
The four fields that matter
Every rulebook specifies the same four things, and every one of them can bite.
The station. Named explicitly, usually with a link to the data source. This is the single most important line in the document and the one people skim.
The unit. Celsius for most of the world, Fahrenheit for the United States markets. This changes the ladder spacing, which changes the difficulty.
The precision. Thirty eight of the thirty nine say the source measures to whole degrees, so that is the precision used for resolution. One does not.
The window. The highest reading for all times on the local calendar day, with revisions counted until the following day's first data point publishes.
The station traps
Airports are the trap. A city usually has more than one, and the market does not always pick the obvious one.
Denver settles at Buckley Space Force Base, not Denver International. Dallas settles at Love Field, not DFW. Houston settles at Hobby, not Intercontinental. New York settles at LaGuardia, not JFK or Newark. Outside the United States the same pattern appears: Paris settles at Le Bourget rather than Charles de Gaulle, Moscow at Vnukovo, Taipei at Songshan rather than Taoyuan, London at London City rather than Heathrow.
These are not interchangeable. Two airports in the same metro can differ by a degree or more on a given afternoon, and one degree is exactly the width of a rung. If you are pulling a forecast for the wrong airport you are forecasting a different market.
The reliable way to check is the resolution link itself. Most rulebooks embed a Wunderground history URL, and the last path segment is the station identifier. That code is the ground truth, not the airport name in the title.
Whole degrees, with one exception
Thirty eight markets state that the source measures to whole degrees and that this is the precision used for resolution. That matters more than it sounds.
It means there is no rounding boundary to be clever about. The underlying observations are already integers, so a market on twenty six degrees is not asking whether the true temperature exceeded twenty five point five. It is asking whether the number twenty six appears in the day's readings. If the day tops out at twenty five, the twenty six rung pays nothing, no matter how close the underlying air actually got.
The exception is Hong Kong, which settles against the Hong Kong Observatory rather than an airport, and whose rulebook specifies one decimal place. That is a genuinely different instrument with a different resolution procedure, and it should not be modelled alongside the airport markets.
Celsius ladders and Fahrenheit ladders are not the same difficulty
The non-US markets use eleven rungs of one degree Celsius each, with the top and bottom rungs open ended. The US markets also use eleven rungs, but each middle rung spans two degrees Fahrenheit.
Two degrees Fahrenheit is about one point one degrees Celsius, so the American ladders are slightly wider and therefore slightly easier to hit. You can see it in the prices. On a Celsius ladder the leading rung typically sits around forty cents. On a Fahrenheit ladder we have seen leaders in the mid fifties with the top two rungs together holding close to ninety percent of the probability.
Wider rungs mean a higher hit rate and a higher entry price. Whether that is better depends entirely on which effect dominates, and it is not obvious in advance.
Two resolution sources, not one
Most markets read Wunderground. Three of the Celsius markets read the United States National Weather Service timeseries instead: Istanbul, Tel Aviv, and Moscow. The rulebook tells you which, and the difference matters if you are cross-checking against a third source, because the two feeds do not always carry identical revision histories.
Both sources take revisions. The rulebook is explicit that corrections are considered until the following day's first data point publishes, after which the number is frozen. That is why a market can sit unresolved for hours after the day is over.
The close is not the end of the day
This is the structural detail that surprises people most. Trading closes at a fixed hour that has nothing to do with the local day being measured.
For a market on a European city the close lands in the local afternoon, after the day's peak has typically occurred, so the last hours of trading happen with the answer partly known. For a market on an American city the same fixed close lands in the local early morning, before the day has warmed at all. Those markets stop trading before the weather they are about has happened.
The consequence is that American temperature markets are pure forecasting exercises with a long lead, while Asian and European ones let you watch observations come in. Same product, completely different game, and the price behaviour reflects it.
How to check a market in sixty seconds
Before trading any temperature ladder, read the resolution text and answer four questions.
Which station code appears in the source link. Whether the unit is Celsius or Fahrenheit. Whether it says whole degrees. And what local time the close corresponds to, which tells you whether you are forecasting or observing.
Everything else in the document is boilerplate. Those four lines determine what you are actually buying.
Frequently asked
Can I use a city forecast? Not reliably. Settlement airports sit ten to forty kilometres outside the centres they are named after, and the offset between a city forecast and an airport reading is stable enough to move you a full rung.
What if the station is missing data? Resolution uses the highest reading for all times on the day, so gaps reduce the sample rather than voiding the market. Special observations count alongside routine hourly ones.
Do revisions ever change a settled market? The window closes when the following day's first data point publishes. Corrections inside that window count, corrections after it do not.
Where to read the ladders
Reading eleven rungs at once is awkward on most interfaces. SmartX carries the same Polymarket books laid out as a ladder with depth on each rung, which makes the shape of the distribution visible at a glance.