What actually moves a prediction market price

The short answer: three things move the price. New information moves it a lot, order flow moves it a little, and the slow arithmetic of time moves it constantly. The skill in reading a market is telling those three apart, because only the first one should change your mind.

The spread is a fog bank, not a number

Every market shows one price, but nobody can trade at it. A contract "at 62" might be 61 bid, 64 ask, with real depth only at 58 and 67. The mid-point is a courtesy fiction. When we score our own forecasts we grade against the executable quote, fees included, because that is the only number anyone could have acted on. Any analysis that grades against the mid-point is quietly flattering itself by a point or two per trade.

Time does most of the daily work

A weather contract that sits at 40 cents for three days is not "stable". If the event needs a hot afternoon and the afternoon keeps not arriving, standing still is drifting. Watch the same contract against a forecast that updates on every model run and the difference becomes visible: the market moves in steps when traders wake up; the physical probability moves smoothly as the horizon shortens.

JMM forecast58%Market price52¢Last 72 hours

Resolution rules are the fine print that settles everything

"NYC high temperature above 88°F" sounds unambiguous until you ask which thermometer. Contracts settle on one station's daily climate report, and stations disagree with the city around them by whole degrees on a still, sunny day. The single highest-value habit in these markets is reading the resolution source before reading the price. When we map a contract we record the exact station, variable, threshold, and timezone, and a human checks it, because a mismapped station poisons every downstream number.

What a price cannot tell you

A thin market's price carries at most a few bits of information. It cannot tell you the shape of the distribution, only where one cut through it lands. Fifty ensemble members can. That asymmetry, publishing the distribution instead of one cut, is the entire reason this site exists: a price and a probability band answer different questions, and the band answers more of them.

Sources

• ECMWF open data, ensemble forecasts, 00z/12z runs (CC BY 4.0), accessed Aug 1, 2026.
• NWS/NOAA daily climate reports for KNYC, public domain, accessed Aug 1, 2026.
• Venue fee schedules and resolution rules as published, versioned in our archive at snapshot time.