Prediction markets have lengthy promised to combination insights about future occasions. More and more, these indicators are coming not simply from folks, however from machines.
In response to David Minarsch, CEO and co-founder of Valory AG, the group behind the crypto-AI protocol Olas, autonomous AI brokers are rising as highly effective instruments for buying and selling prediction markets, significantly for retail customers making an attempt to compete in an more and more automated setting.
Valory builds merchandise on the intersection of blockchain and multi-agent techniques (MAS), and its present focus is Olas, previously often known as Autonolas. The protocol is designed as infrastructure for autonomous software program brokers that may run companies on blockchains, work together with good contracts, and cooperate with each other whereas incomes crypto rewards.
The broader imaginative and prescient is what Minarsch calls an “agent financial system”. A decentralized ecosystem the place autonomous AI brokers carry out helpful duties and generate worth for his or her customers.
Some of the seen experiments in that imaginative and prescient is Polystrat, an AI agent launched on the prediction-market platform Polymarket in February 2026. The agent trades on behalf of customers who self-custody and personal it, executing methods repeatedly across the clock.
“In a nutshell, Polystrat is an autonomous AI agent that trades on Polymarket 24/7 on behalf of its human consumer,” Minarsch stated. The thought is straightforward: whereas people sleep, work or lose focus, the agent retains buying and selling.
Prediction markets, platforms the place customers commerce contracts tied to real-world outcomes, have surged from area of interest forecasting instruments right into a fast-growing nook of fintech over the previous few years. The trade’s breakout second got here in the course of the 2024 U.S. presidential election, when buying and selling volumes spiked and the markets gained mainstream visibility, adopted by speedy enlargement into sports activities, economics, and crypto-related bets. By 2025, complete notional buying and selling quantity throughout main platforms exceeded $44 billion, with month-to-month exercise reaching as a lot as $13 billion throughout peak durations.
At present the market is very concentrated round two dominant gamers: Kalshi, a U.S.-regulated event-contracts trade overseen by the Commodity Futures Buying and selling Fee, and Polymarket, a crypto-native platform that operates globally and gives a broader vary of prediction markets. Collectively they account for roughly 85–97% of buying and selling quantity within the sector, processing tens of billions of {dollars} in annual bets on every thing from elections and central-bank coverage to sports activities and cultural occasions
Why machines could outperform people
The push towards AI-driven buying and selling stems from a easy statement. A lot of the intelligence embedded in trendy AI fashions hasn’t but translated into monetary markets.
That realization prompted Valory’s group to start constructing what they name a “prediction market financial system” on Olas in 2023, an ecosystem the place AI brokers use prediction instruments and knowledge pipelines to forecast outcomes and commerce on them.
Prediction markets themselves are constructed on probabilistic forecasting. A easy guess about an occasion, whether or not a political end result, financial indicator or sports activities consequence, is likely to be no higher than a coin flip. However structured knowledge evaluation and disciplined buying and selling methods can change that equation.
“Merely prompting off-the-shelf fashions with markets normally leads to outcomes no higher than a coin-flip,” Minarsch stated. “However state-of-the-art AI fashions wrapped in customized workflows, so known as prediction instruments, have traditionally proven predictive accuracy as much as 70% and better.”
The outcomes thus far counsel that machines could have a bonus. Third-party knowledge signifies that solely about 7% to 13% of human merchants obtain constructive efficiency on prediction markets, whereas the bulk lose cash.
On the identical time, machine participation is rising shortly. Greater than 30% of wallets on Polymarket are already utilizing AI brokers, in keeping with analytics platform LayerHub.
Minarsch believes this pattern displays a broader shift: people are already competing with machines whether or not they understand it or not. “You’ve got human members in prediction markets alongside many machines,” he stated. “So people are already in a battle with machines.”
The important thing distinction is that machines are much less emotional and higher at sticking to constant methods.
By making AI brokers out there to on a regular basis customers, Olas goals to degree that taking part in subject.
Early traction for autonomous merchants
The early efficiency of Polystrat has been encouraging.
Inside roughly a month of launch, the agent executed greater than 4,200 trades on Polymarket and recorded single-trade returns as excessive as 376%, in keeping with knowledge shared by the group.
“Brokers are likely to do higher than people,” he stated. “Polystrat AI brokers already outperform human members in Polymarket, with over 37% of them exhibiting a constructive P&L versus lower than half that quantity for human members.”
Customers can configure their very own brokers relying on technique preferences, knowledge sources or danger tolerance.
The lengthy tail of prediction markets
Past efficiency, Minarsch believes AI brokers may unlock an neglected alternative in prediction markets: the “lengthy tail” of area of interest or localized questions.
Many prediction markets revolve round main international occasions, elections, macroeconomic knowledge or high-profile sports activities competitions. However numerous smaller questions stay largely unexplored.
“People typically don’t hassle digging for the data,” Minarsch stated. “They’ll’t be bothered to take the time.” AI brokers, in contrast, can analyze giant numbers of smaller markets concurrently.
“The lengthy tail of prediction markets could be very attention-grabbing for AI brokers,” he stated. “You simply level the agent on the downside and it does the work.”
This might assist broaden prediction markets as a data-gathering software for companies, policymakers and decision-makers. Forecast markets have lengthy been studied as methods to combination dispersed data and floor insights that conventional surveys or fashions may miss.
In that sense, prediction markets could turn out to be a form of upstream expertise for decision-making throughout industries.
Human-AI collaboration
Regardless of the rise of automation, Minarsch doesn’t see AI brokers changing people completely.
As an alternative, he frames them as enhances.
“People make decisions in a extra rushed method, which might be detrimental,” he stated. “AI brokers can act as one thing people depend on.”
One future route includes permitting customers to reinforce their brokers with proprietary data or specialised knowledge units. “We see demand from customers who need their agent to faucet into their very own data base or proprietary info,” Minarsch stated. “That might permit brokers to commerce in a extra principled method than a human may.”
Over time, the group says prediction fashions and knowledge pipelines powering these brokers have improved considerably, producing sustained alpha when mixed with general-purpose giant language fashions.
Dangers and regulation
The expansion of prediction markets additionally raises moral and regulatory questions.
Some critics argue that markets forecasting wars, deaths or disasters may create incentives to govern outcomes or revenue from dangerous occasions.
Minarsch acknowledged that cautious guardrails are wanted.
“There must be regulation about what sorts of prediction markets ought to exist,” he stated.
On the identical time, he believes AI brokers may additionally assist detect problematic markets or manipulation makes an attempt by figuring out suspicious patterns.
“Brokers may spot patterns and assist shut down problematic markets,” he stated.
Constructing a user-owned AI financial system
For Minarsch, the last word aim just isn’t merely higher buying and selling methods.
It’s guaranteeing that on a regular basis customers retain a stake in an more and more automated digital financial system.
A future the place AI techniques carry out most financial exercise may danger disenfranchising people if centralized platforms management the expertise. “Olas goals to create a world the place human customers might be empowered via their AI brokers slightly than disenfranchised by them.”
To counter that dynamic, the venture emphasizes consumer possession of AI techniques. “We wish to create extra user-owned brokers,” Minarsch stated.
If profitable, that mannequin may permit folks to deploy autonomous software program that generates worth on their behalf throughout markets and companies. Prediction markets are simply the place to begin.
Learn extra: AI rout hits software program shares, however Grayscale says blockchains stand to profit