Kalshi settlement timeline
When each desk buys, when its markets settle, and when its cash is free again. Times are US Eastern. Measured from Kalshi's public data on settled markets, Aug 1 – Oct 16 2026; the desk lanes replay the weather rule's real historical picks at 1 contract each. Research only: no orders. Built Oct 16 2026.
A day on the timeline
live deskpaper desk open slot for a new edgesettles per hour, rest of Kalshi (darker = more)
Cash out at each hour (three desks, $ at 1 contract per pick)
Average over the replay. Cash leaves at 6 PM and returns from 7 AM (gas) to 10 AM (weather, bought two evenings before); yields money returns within about 2 hours.
Settles per day of week (three desks, replay)
Net dollars per hour
Definition: profit and loss after Kalshi's fees, for every pick that settled, divided by every clock hour in the period (168 a week, nights and weekends included). Alongside it: return on the cash actually tied up, and the float (most dollars out at any hour), so a bigger bankroll can't hide. Replay of the weather rule's real picks.
weather (live)gas (paper) yields (paper, 2 days of picks)
Capacity = contracts other traders bought at or within 1¢ of our ask in the hour after the snapshot (public trade prints). Each pick buys the smaller of the cap and that amount, at least 1. Weather stops growing past about 100; yields is capacity-bound from the start (88% of its picks saw no such trade). The per-contract edge is not yet proven (weather t = 1.25, gas t = 1.04), and losses scale with size too.
What a new edge in each slot would need, to add one subscription line ($0.55/hour)
At 5¢ net per contract, one subscription line is about 1,840 contracts a week; at 10¢, about 920.
Every family: lists, closes, settles
Rule evidence comes from the favourite–longshot study (PROPOSAL.md). "No edge" means the weather rule lost there after fees, not that no edge exists.