A small stake turned into roughly $14,000 after two improbable temperature spikes were recorded at Paris‑Charles de Gaulle — anomalies now at the centre of a French criminal probe into data used to settle bets on Polymarket. Météo‑France says a sensor at the LFPG station showed unlikely jumps on 6 and 15 April and has lodged a complaint with the air transport gendarmerie for alteration of the functioning of a data‑processing system. The readings flipped resolution of short‑term temperature markets on Polymarket, a US prediction platform, producing at least $34,000 in payouts after one market’s trading volume swelled to $590,527.

How the markets and the measurement link up

Polymarket runs short-term prediction markets that ask users to wager on specific outcomes — in this case, the day’s maximum temperature in Paris. Markets are structured as discrete ranges, for example “13°C or below” up to “23°C or above”. Resolution of those markets depends on third-party weather feeds. In the Paris cases, Polymarket’s settlement data relied on Weather Underground, which in turn pulls readings from the LFPG weather station at Paris-Charles de Gaulle airport.

The dependence on a single physical sensor is what investigators now describe as a weak point. If a thermometer at LFPG reports an erroneous spike at the critical hour, the market settles in favour of positions that match that spike — regardless of the broader meteorological context. And because blockchain-based markets pay out automatically once a data feed settles, gains are immediate and irreversible.

The two incidents and the payouts

Officials say two suspicious readings occurred within ten days. On 6 April 2026, the station briefly recorded a jump above 21°C just after 7pm — a rise of more than three degrees in minutes that clashed with the day’s observed trend. One bettor turned a stake described as "a few dozen dollars" into roughly $14,000, according to market data examined by investigators.

Nine days later, on 15 April, a similar pattern repeated. Recorded maximums for surrounding days sat around 18–19°C, yet LFPG showed 22°C on the 15th.

On Polymarket, probability for that temperature band swung from near-zero to about 95% in half an hour. A winning account collected just over $20,000.

Trading volume on that market reached $590,527 — an unusually high figure that has drawn extra scrutiny.

Together the two events produced more than $34,000 in payouts to bettors who timed positions to the suspect readings. One detail that alarmed investigators was the timing of account creation: at least one winning account had been opened only days before a spike, suggesting premeditation rather than happenstance.

What the meteorologists say

Ruben Hallali, a meteorologist with a PhD in atmospheric sciences, former Météo-France employee and co‑founder of HD Rain, analysed the observation series and called the jumps "highly improbable" for those dates and times. "One can imagine an individual with good knowledge of how the sensors work intervened physically at the sensor level to raise temperatures by two degrees at the right moment," he said.

Hallali’s view is that the pattern is inconsistent with normal thermal variability and points to localised interference rather than a genuine weather event. Such an intervention could be as crude as applying a local heat source near the instrument at a key minute — an approach that would register on the sensor without affecting the actual ambient conditions across the wider area.

Official response and legal action

Météo‑France confirmed that it carried out physical checks on one of its instruments and detected signs of tampering.

The agency lodged a criminal complaint with the air transport gendarmerie, citing alteration of the functioning of a data-processing system. That offence can carry substantial penalties under French law.

The complaint asks investigators to establish whether third parties deliberately interfered with the sensor.

Related Articles

Météo‑France has filed a criminal complaint with the air transport gendarmerie for alteration of the functioning of a data‑processing system.

This article was created with AI assistance.