Starship's recent launch campaigns suggest Flight 14 is close. The question is whether the September 18 cutoff leaves enough room for the delays that usually happen at the end.
PastBehavior expects Flight 14 only a few days later than the market does. The September 18 cutoff happens to sit between those estimates, creating a larger probability gap than the underlying disagreement suggests.
Kalshi is running a market on whether SpaceX launches Starship Flight 14 before September 18, 2026. It resolves off FAA data and requires an actual liftoff, so a scrub or an aborted countdown does not count. Flight 13 burned two attempts before it flew, which makes that distinction a practical one rather than a technicality.
Study date is August 25. A qualifying launch has to happen on or before September 17. That leaves 23 days.
One number carries most of this forecast.
Every Starship campaign ends the same way. Static fire the ship, static fire the booster, stack, fly. The final static fire is the gate that unlocks everything after it. Across the four most recent campaigns, the time from that last major ground test to liftoff looked like this.
A late-August Booster 21 static fire implies a historical launch band of roughly September 10–19. The qualifying cutoff is September 17.
Ship 41 completed its full-duration six-engine static fire on August 21. Booster 21 had not fired at all as of the study date.
The cutoff falls inside that band. It does not sit past it.
Launch-to-launch cadence is the more familiar statistic, but those gaps run from 37 days to 221 and mostly reflect whatever had grounded the program at the time. The ground-test interval reflects where the hardware actually is.
Better than the "SpaceX scales back" coverage suggests.
Ship 41 finished both static fires and rolled back to the production site under its own power, a first, for final checks and flight termination system installation. Booster 21 went out to Pad 2 on August 24, one day before the study date, with its static fire expected within the week.
The regulator is not the obstacle people assume. The FAA declined to require a mishap investigation after Flight 13. Environmental review covering orbital trajectories closed in February, and the annual launch cap at Starbase had already been raised to 25. SpaceX also dropped the tower catch of the upper stage, so both stages now splash down, which takes the slowest approval off the critical path.
One real open item remains. Flight 14 is the first orbital insertion, and a licence modification for that specific profile has not been publicly confirmed. Every interval above came from a flight repeating a known profile under an existing licence. That is genuine uncertainty, though a smaller version of it than two years ago.
Four dated signals pointed at Flight 14 by the study date.
Every dated Flight 14 signal had moved later by the study date. None had moved earlier.
That is the program's normal behaviour rather than a Flight 14 problem. A no-earlier-than date is a floor, not a midpoint.
Ordinary friction does the rest. Flight 13 needed three attempts and lost eight days. Booster 19 needed an unplanned second static fire three weeks after its first. Mid-September is peak hurricane season on the Texas coast. Inside 23 days, one of those is survivable. Two are not.
The 16-point probability difference looks larger than the underlying disagreement. The two estimates are only a few days apart, with the contract cutoff sitting between them.
That is why this forecast carries an action of BID NO alongside a signal quality of LOW. The action follows the experiment's pre-registered rule, which treats a probability gap of 15 points or more as a signal. Signal quality describes something different: how robust the underlying disagreement is. Here it is a disagreement about timing rather than about mechanism, so it is less robust than PB-002 or other forecasts where PastBehavior and the market read the situation itself differently.
The strongest case against the forecast is straightforward.
Booster 21 was already on Pad 2 on August 24. The two previous Pad 2 campaigns went from an equivalent state to launch in about 15 days, pointing to roughly September 8–9.
That would clear the cutoff comfortably, and it is why PastBehavior sits at 45% rather than something much lower. The hardware is close enough that the vehicle is very likely capable of flying before the deadline. What the historical record questions is whether the last few gates, the booster static fire, the stack, the notices and the first two or three countdown attempts, all fit inside 23 days.
This is a paper signal recorded for the PastBehavior experiment. It is not advice to trade real money.
Moderate confidence is doing real work here. The closest reference class has four observations, and the two most relevant have two.
PastBehavior is testing whether historical patterns can produce forecasts that can be written down, compared against a market, and scored after resolution.
This is Prediction Markets No. 004.