A few days before Christmas, Howard Qin was scrutinizing weather forecasts on two laptops at home in Shanghai when he noticed prices of contracts for New York City snowfall creeping higher on the prediction market Kalshi. He checked the Times Square webcam for specks of white — affirmative.
The recent Stanford math graduate and lifelong weather buff had spent about $200 on predictions that total NYC snowfall that month would top various thresholds — for example, more than 2 inches. Now the value of his shares was rising.
There was just one wrinkle: Qin had a classical guitar recital to attend that night. “I’m not going to bail on my concert just to trade on this. That would be a little extreme,” he recalls thinking. He sold and netted $327.79 — a 57% gain and his biggest win yet.
That was 2024. Two years later, Qin, now 24, is still making small bets “just for fun,” including on Central Park snowfall during January’s megastorm (he won a few dollars). What’s changed is the sheer number of people placing bets on weather markets — and the amount of money flowing through them. Trading volume for the January snowstorm topped $6 million on Kalshi, one of the largest weather contracts ever traded on the rapidly growing platform.
That’s still tiny compared with Kalshi bets on sports and elections. But weather trades are gaining traction, drawing in casual participants, weather experts and AI-driven weather-tech firms testing their wares. As these markets grow, weather nerds and climate researchers are debating whether prediction markets can improve forecasts by aggregating knowledge — and, in turn, inform investment and policy — or whether they’re simply zero-sum games in which uninformed gamblers make (or lose) a quick buck.
Testing Weather Models
Qin got into weather betting to test his meteorological knowledge. So he was excited when his managers at Windborne Systems — a weather-tech startup that launches balloons and uses the data to power an AI forecasting tool — encouraged him to use the company’s models to make trades during his internship.
“It’s a great market to play around with to test out your predict accuracy,” said Chief Executive Officer John Dean. He described the effort as “dogfooding,” a Silicon Valley term for using a company’s own product to identify bugs and make improvements.
The approach has had concrete payoffs. Through trading, Dean said he realized data from official weather stations — used by Kalshi and Polymarket to settle contracts — can be noisy. A sensor’s temperature readings may spike in direct sun, making conditions seem hotter than they are. Windborne subsequently adjusted how it preprocesses its model-training data.
There is also the straightforward motivation of money to be made.
Established weather vendors sell their data — such as temperature, wind speed and cloud cover — to energy traders and hedge funds, who trade using that information. Marvin Gabler, co-founder and CEO of Swiss-based weather forecasting startup Jua, sees a way to unlock even more value: by trading on prediction markets with insights generated by his company’s models.
Last year, Gabler set up a separate investment vehicle, pooling funds from friends and family, and is trading on maximum-temperature contracts on Polymarket using Jua’s forecasts. So far, “relative returns are high,” he said, though market volume and liquidity “are still too low for a well-sized fund.” He declined to share further specifics.
Better Forecasts?
Not every person trading these markets is an expert in meteorology. In fact, one of the best performers on crypto-powered platform Polymarket is a total novice: A 23-year-old law student in Germany, who goes by the username Hans323 on the website, is currently the sixth-highest profit earner of all time on Polymarket’s weather markets. His bets tracking daily temperatures across cities like London and New York have become a touchstone in prediction market circles, followed by eager eyes wanting to copy his strategies and turn a profit themselves.
Prediction markets may already be outperforming traditional weather forecasts, according to an analysis published last month by Patrick Brown, head of climate analytics at Interactive Brokers. He compared the implied forecasts of his firm’s prediction markets to ones from the US National Weather Service and found the former were more accurate. The key difference, Brown contended, was that prediction markets better “incentivized human judgment.”
“There is a direct financial reward for being accurate and a direct financial penalty for being inaccurate. This creates a dual effect of attracting accurate individuals and systems into the market while deterring those who are inaccurate,” Brown wrote.
Still, some weather bettors are skeptical of the idea that their aggregated wagers produce any useful information. While Dean says weather prediction markets can reflect human intuition — unlike AI weather forecasts — Atte, a 40-something Finnish freelance software developer who is also among the top weather traders on Polymarket, describes his weather bets as “completely worthless.” That’s despite his raking in some $33,000 on them since October.
“I would prefer to be societally useful,” said Atte, who goes by 1-800-LIQUIDITY on Polymarket and declined to share his full name because of safety concerns. He has sought jobs at AI startups that would put his coding chops to use, but for now he’s applying them to weather markets, including writing software to automate trades.
As the climate veers further into uncharted territory — the last 11 years were the hottest on record and the warming may be accelerating — critics also worry that the gamification of weather could bring perverse consequences, including data manipulation and the sabotaging of weather stations.
Other prediction markets may have already seen foul play. Late last year, a live map of the Russia-Ukraine war produced by a Washington, DC, think tank was mysteriously altered — long enough for Polymarket to resolve a bet that Russians had captured a city they hadn’t. And in March, an Israeli reporter claimed Polymarket users tried to pressure him to change a story involving a missile strike outside Jerusalem.
Polymarket did not respond to emailed questions.
Science Markets
Beyond the open-access platforms of Kalshi and Polymarket, scientists are building bespoke trading platforms that they hope will draw out societally useful information. There, weather experts with little or no betting or professional investing experience are stepping into betting pools designed to discover insurance-relevant risk — and infusing their slow-moving research with adrenaline.
Mark Roulston is a Ph.D. planetary scientist who spent a decade at investment firm Winton Group, incorporating weather and climate information into quantitative trading strategies. He developed a passion for prediction markets as a way to help institutions extract hidden information from diverse groups of researchers, who are invited to bet on their expertise using other people’s money.
That’s the first and biggest difference between Roulston’s effort and conventional markets: A sponsor supplies the betting money, rather than the losers funding the market. The sponsor always “loses,” in the sense that it’s giving away money without placing a bet. But it “wins” by cultivating the intelligence expressed by the participants, who keep whatever money they win.
“We create markets not because we think, ‘Oh, this is something lots of people will want to bet on,’” Roulston said. “We create a market because there’s an end user — like a reinsurance company — that says ‘We would like better predictions of this.’” He says this approach is closer to the vision for prediction markets advocated by economic luminaries such as Nobel Prize winner Kenneth Arrow.
Winton Group backed Roulston’s first formal prediction effort in 2018, which aimed to forecast UK spring and summer temperature and precipitation.
“It was very addictive, actually,” said Kristian Strommen, a climate science and weather forecasting researcher at the University of Oxford. Making live forecasts is “very different from what you do in academia — writing a paper, and there’s no immediate application of anything.”
That experience also taught participants about the mutable shape of expertise in a betting market. Strommen’s team of half a dozen or so ended up placing second in the competition of 24 teams, behind a one-man operation.
After registering solo, “I actually ended up winning the competition, much to the annoyance of those guys,” said Andrew McRae, at the time a fellow Oxford researcher.
McRae, who still participates in Roulston’s markets, now on a team with Strommen, said he won in part because his solitary position made him hyperconscious of how he was spending his time. He knew his colleagues’ strength and combined size meant their model was probably more nuanced. So he looked for small bits of leverage, making improvements on the margins by writing programs that allowed him to trade faster than other teams and by betting on underpriced outcomes, rather than on likeliest outcomes.
“It felt a lot more like a real financial market where there’s lots of other participants on there, and the market is sort of correct,” he said. “And it’s just like, I’ve got a little bit of information on top of that.”
The Scor Foundation for Science, the philanthropic arm of a French reinsurer, announced in late 2024 it would support Roulston’s group, known by its acronym, CRUCIAL, and housed at Lancaster University. Scor is interested in the approach because it can draw diverse insights out of researchers more efficiently than if it were to hire dozens of them at once, said Philippe Trainar, CEO of the Scor Foundation and chief economist of Scor SE.
Roulston runs annual competitions on the number of Atlantic and Pacific cyclones and has four live markets to gauge when the next El Niño will hit — likely bringing record global temperatures with it.
The complexity of climate change may make the topic unsuited for popular online prediction markets, said Madison Condon, an associate professor of law at Boston University, who writes about climate models and projections. There’s definitely space for scientists to pool knowledge outside of complex Earth system models, she said, but markets may not be “the best way for synthesizing that domain knowledge or expert judgment.”
“It’s not like a basketball game,” she said. Projecting what climate change may do to hurricanes, or if the next El Niño is on its way, “is just a different type of knowledge than polling the population would give you.”
Betting on the weather isn’t new. The first weather derivatives began trading in the 1990s, and insurers have long scrutinized the costs of weather and climate risks to underwrite policies. But the market for offloading weather risk remains relatively small. Meanwhile, intensifying climate threats are exposing cracks in the insurance industry as extreme weather makes some places uninsurable.
Some industry veterans see prediction markets as a potential solution to the problem — or at least a disruption of the status quo.
While weather derivatives are constrained by low liquidity and parametric insurance is costly, prediction markets open the floodgates to more traders, said Jim Huang, who spent a decade at the Chicago Mercantile Exchange working on product strategy and is now building a weather-focused prediction market platform called WeatherBook.
That can help boost liquidity and improve the pricing of weather risks — “something that I don’t think insurance or derivatives will be able to achieve, even if you give it more time,” he said.
‘Put Your Money Where Your Mouth Is’
One of the earlier generation of prediction markets can also trace its history back to weather predictions, though in this case the bets were made in pursuit of a childhood fantasy: a snow day.
Like all other 9- and 10-year-olds, Dean and John Aristotle Phillips spent the night before snowstorms in their North Haven, Connecticut, home praying that school would be canceled the next day. It was the mid-1960s. Then one day they realized they could do something more than pray: They could bet.
“It was money,” John Phillips said. The two sides were snow day and no snow day. “You had to have the money to bet with. There was no credit, no margin.”
The brothers’ snow day prediction market was the first partnership in entrepreneurial careers that soon led them to a snow-shoveling business, John’s design of a nuclear bomb from public sources as a Princeton junior, work in election data and technology, and, in 2014, the launch of one of the first major online prediction markets, PredictIt.
Once the boys’ wintertime taunt, “Put your money where your mouth is” is now “kind of the motto of PredictIt,” Phillips said.
The betting platform started as a nonprofit collaboration between Victoria University of Wellington, New Zealand, and the Phillips’ data and consulting company, Aristotle International Inc. A US nonprofit called the Prediction Market Research Consortium Inc. took over the university’s role last year, with Aristotle still running it day to day.
For investors, betting itself can be an efficient way to deliver actionable policy information that complements normal due diligence, Phillips said. An investor considering putting money into a wind turbine firm, for example, must weigh factors spanning business, policy and geopolitics.
“The real thing is, what’s the government going to do in this regard?” Phillips said. “To me, that’s a really fertile area for these markets to contribute.”
Source: https://www.bloomberg.com/news/articles/2026-04-10/what-are-weather-prediction-markets-and-do-they-work
