Storm Hunter

YES is one of two contracts in any prediction market. If the bot buys YES at 24¢ and the answer to the question turns out to be YES, that contract pays $1 — a $0.76 profit on each one. If the answer is NO, the YES contract is worth $0 and the bot loses what it paid.

Kalshi runs hundreds of contracts on real-world weather questions every day: "Will Manhattan's high temperature exceed 78°F on Friday?", "Will Denver get 4+ inches of snow this weekend?", "Will it rain in Boston on Monday morning?"

The bot only watches these — it ignores politics, sports, crypto markets entirely. That focus is the edge.

A contract under 28¢ means the market thinks there's less than a 28% chance of the event happening. The bot only bets when the market thinks something is unlikely. Why? Because retail traders systematically underprice tail probabilities — they think long-shots are even more unlikely than they really are.

If the bot's research is right, the true probability is higher than 28%, so buying at that price is a positive expected-value bet.

This is the key window. The bot has noticed that prices on weather markets are least efficient in the last 8 hours before they resolve. Professional traders have gone home; retail traders dominate the order book and overreact to forecasts. That's when the mispricings open up.

Outside that window — say, 3 days before resolution — there's plenty of time for the price to find its fair value, and the edge disappears.

Technically, the daily price range needs to be at least 6¢. In plain English: the market needs to have been active enough today that there's a reasonable chance the price will keep moving.

Stagnant markets that haven't traded much are dangerous — usually because the resolution is essentially known already, and the bot's "cheap" entry isn't actually cheap.