Weather Board
Signal queue
--| Signal | Market | Ours | Market | Edge % | 1h flow % | Depth | Closes |
|---|
Edge is our probability minus the market's on the best-priced bucket. BUY needs a double-digit edge with real depth behind it; LEAN is smaller; SPLIT means we disagree on the winning bucket without a priced way in. We never signal a bucket the market prices under 5¢, or one more than a degree from our own pick: a huge gap against a near-zero price is a distribution disagreement, not an entry. Model output, not investment advice.
--
-- --Why this call · --
the evidence behind the selected marketMethodology and data
every number on this page, sourcedEach market pays on one station's posted daily history: the highest reading, in whole degrees Celsius, over that city's local calendar day, ending at the city's own midnight. The exact settlement source is linked on every market.
Prices trade per bucket: one whole degree each, so 28° covers 27.5 to 28.5, and the end buckets are open-ended. An aggregate like "30° or below" is worth exactly the sum of the degrees inside it, nothing extra.
Observations from the settlement station and seven nearby stations, refreshed every minute. The chart's solid line is the running daily maximum, which is the number the market settles on.
The provider's daily card and its hourly curve, read twice: integer Celsius and integer Fahrenheit round into different buckets, so both feed the blend as separate inputs. Their weights are printed on every market in "Why this call".
Hours the forecast calls wet, at a forty percent chance or more, are shaded on the day chart. Rain is the one thing that reliably stops a climb, so a shaded band across the forecast peak is the first risk to price.
The same station's last fifteen days at the same local hour: how much upside was typically left from this position. This is what keeps the model honest when a forecast runs hot against the tape.
ECMWF, ICON and GFS daily maxima ride alongside every market as reference. They carry no weight yet: a new input earns its way into the blend from the settled record, not from looking plausible on day one.
Bucket prices and book depth from the SmartX order book, refreshed every minute. Edge is our probability minus the market's price on the same bucket.
BUY needs a ten point edge, fifty percent model conviction and under eighteen hours to the close; LEAN needs four points and thirty five percent. We never signal a bucket the market prices under five cents, or one more than a degree from our own pick, and markets further than eighteen hours out cap at WATCH.
Trading ends at the city's midnight and the outcome is fixed there: the day is over and its maximum cannot move. The official station row lands within an hour or two, pays the winning bucket one dollar per share, and the market stays on the board for six hours showing the result.
Everything on this page refreshes every sixty seconds. Each market rolls to the next local day at its own midnight, automatically: the queue, the map, the chart and the ladder follow without intervention.
Signals are the model's opinion, not advice. When the model and the market disagree, that is a disagreement, not proof of an edge: the record that settles the argument is collected daily and will be published once it is long enough to mean something.
Temperature field by Ventusky. Station history from the settlement source linked on each market. Built by SmartX.