Trading the observed temperature maximum
Forget forecasting — wait until the day's high has already happened, then buy the winning bucket. The purest information-latency trade we could design, and it still went 1-for-6.
The hypothesisOnce a day's maximum temperature is in and the afternoon is cooling, the winning bucket is determined fact, not forecast. Read the airport station's live observations; when the max is locked, buy the winner before the market finishes converging.
The test
A daemon polled official aviation weather observations for each market's resolution station, declared a "lock" when the running max was hours old with temperatures declining (hardened later with cloud-cover and wind guards against false afternoon dips), and logged the winning bucket's live executable ask. Everything settled against the market's actual resolution — which is the step that turned out to matter.
The result
Two failure modes, both fatal. First, no prey: roughly 90% of the time, by the moment certainty was defensible the bucket already asked 98–99¢ — the fillable entries existed only when certainty was premature, and those went 1W–5L (−78% ROI), mostly afternoons that kept heating after our "lock" (one station went 25°→27° after we called the max). Second, and worse: a verified resolution-source mismatch — these markets settle on a different data provider than the station feed a trader can watch; we documented a day where our source's maximum mapped to a different bucket than the market resolved. The probe was sometimes right about the weather and wrong about the referee.
Conclusion
Even "the outcome has already physically happened" is not an edge if the market converges before you can act and the settlement source differs from your instrument. Information latency is our most-validated edge class in principle — this experiment shows the two additional conditions it needs in practice: a fillable price at certainty time, and the referee's own data feed. Weather markets provide neither.