AI systems are subtly transforming prediction market trading.

David Minarsch, co-founder of Valory, states that autonomous agents operating on the Olas protocol are providing retail traders with a continuous, strategy-oriented advantage on platforms such as Polymarket.

AI agents are subtly transforming prediction market trading. (Unsplash, modified by CoinDesk)

Key points:

• AI agents are becoming more prominent in prediction markets, enabling retail traders to compete with automated strategies by operating continuously and adhering to disciplined, data-centric methods.
• Olas and its Polystrat agent have shown initial progress, completing over 4,200 trades on Polymarket in a month and achieving returns as high as 376% on individual transactions.
• Olas aspires to create “agent economies” where user-owned AI agents generate value.

Prediction markets have historically aimed to consolidate insights about future events. Increasingly, these signals are being generated not only by individuals but also by machines.

David Minarsch, CEO and co-founder of Valory AG, the entity behind the crypto-AI protocol Olas, asserts that autonomous AI agents are emerging as effective instruments for trading prediction markets, specifically for retail users aiming to compete in a more automated landscape.

Valory develops products at the intersection of blockchain and multi-agent systems (MAS), with its current focus on Olas, previously referred to as Autonolas. This protocol serves as infrastructure for autonomous software agents capable of running services on blockchains, interacting with smart contracts, and collaborating with one another while earning cryptocurrency rewards.

The overarching vision, as described by Minarsch, is what he terms an “agent economy,” a decentralized ecosystem where autonomous AI agents execute valuable tasks and generate benefits for their users.

One of the most visible initiatives in this vision is Polystrat, an AI agent introduced on the prediction-market platform Polymarket in February 2026. This agent trades on behalf of users who self-custody and possess it, implementing strategies around the clock.

“In summary, Polystrat acts as an autonomous AI agent that trades on Polymarket continuously for its human user,” Minarsch explained. The concept is straightforward: while individuals sleep, work, or lose concentration, the agent continues trading.

Prediction markets, platforms where users trade contracts linked to real-world results, have evolved from specialized forecasting tools to a rapidly growing segment of fintech in recent years. The industry experienced a significant breakthrough during the 2024 U.S. presidential election, which saw a surge in trading volumes and mainstream attention, subsequently expanding into sports, economics, and cryptocurrency-related wagers. By 2025, total notional trading volume across major platforms surpassed $44 billion, with monthly activity reaching as high as $13 billion during peak times.

Currently, the market is heavily concentrated around two leading entities: Kalshi, a U.S.-regulated event-contracts exchange supervised by the Commodity Futures Trading Commission, and Polymarket, a crypto-native platform that operates internationally and provides a wide array of prediction markets. Collectively, they represent approximately 85–97% of trading volume in the sector, processing tens of billions of dollars in annual wagers on a variety of topics including elections, central-bank policy, sports, and cultural occurrences.

Why machines may excel beyond humans

The movement toward AI-driven trading arises from a fundamental observation: much of the intelligence embedded in contemporary AI models has yet to be realized in financial markets.

This realization prompted Valory’s team to initiate the development of what they refer to as a “prediction market economy” on Olas in 2023, an ecosystem where AI agents utilize prediction tools and data pipelines to anticipate outcomes and trade accordingly.

Prediction markets themselves are founded on probabilistic forecasting. A basic prediction regarding an event, whether related to a political outcome, economic indicator, or sports result, might be no better than a random guess. However, structured data analysis and disciplined trading methodologies can significantly alter that scenario.

“Merely prompting standard models with markets typically yields results no better than a guess,” Minarsch noted. “However, cutting-edge AI models embedded in custom workflows, known as prediction tools, have historically demonstrated predictive accuracy reaching up to 70% or higher.”

Preliminary results indicate that machines may possess a competitive edge. Third-party data suggests that only about 7% to 13% of human traders achieve positive performance in prediction markets, while the majority incur losses.

Simultaneously, the participation of machines is rapidly increasing. More than 30% of wallets on Polymarket are reportedly utilizing AI agents, according to analytics platform LayerHub.

Minarsch believes this trend signifies a broader transition: individuals are already competing with machines, often without their awareness. “Human participants in prediction markets are alongside numerous machines,” he stated. “Thus, humans are currently engaged in competition with machines.”

The primary distinction is that machines operate with less emotion and are more adept at adhering to consistent strategies.

By making AI agents accessible to everyday users, Olas aims to equalize that competitive landscape.

Initial success for autonomous traders

The early results of Polystrat have been promising.

Within approximately a month of its launch, the agent completed over 4,200 trades on Polymarket, achieving single-trade returns as high as 376%, based on data shared by the team.

“Agents typically outperform humans,” he remarked. “Polystrat AI agents have already surpassed human participants in Polymarket, with over 37% displaying a positive P&L compared to less than half that proportion for human participants.”

Users can customize their agents based on their strategic preferences, data sources, or risk appetite.

The long tail of prediction markets

Beyond performance, Minarsch believes AI agents could unveil an underappreciated opportunity in prediction markets: the “long tail” of niche or localized inquiries.

Many prediction markets center around significant global events, elections, macroeconomic data, or high-stakes sports competitions. However, countless smaller inquiries remain largely uncharted.

“Humans often lack the motivation to seek out this information,” Minarsch explained. “They may not be inclined to invest the effort.” Conversely, AI agents can analyze numerous smaller markets simultaneously.

“The long tail of prediction markets is quite intriguing for AI agents,” he remarked. “You simply direct the agent at the issue, and it handles the work.”

This capability could enhance prediction markets as a data-collection tool for businesses, policymakers, and decision-makers. Forecast markets have long been examined as methods to consolidate dispersed knowledge and reveal insights that traditional surveys or models may overlook.

In this context, prediction markets could evolve into a form of upstream technology for decision-making across various sectors.

Collaboration between humans and AI

Despite the rise of automation, Minarsch does not envision AI agents completely replacing humans.

Instead, he presents them as complementary tools.

“Humans often make decisions in a hurried manner, which can be detrimental,” he remarked. “AI agents can serve as a resource for humans.”

A potential future direction includes enabling users to enhance their agents with proprietary knowledge or specialized datasets. “We observe demand from users who want their agent to access their own knowledge base or exclusive information,” Minarsch noted. “This would enable agents to trade in a more principled manner than a human could.”

Over time, the team asserts that the prediction models and data pipelines powering these agents have significantly improved, yielding consistent alpha when integrated with general-purpose large language models.

Challenges and regulation

The expansion of prediction markets also brings forth ethical and regulatory concerns.

Some critics argue that markets predicting wars, fatalities, or disasters could create incentives to manipulate outcomes or profit from adverse situations.

Minarsch acknowledged the necessity for stringent safeguards.

“There must be regulation concerning the types of prediction markets that should exist,” he stated.

Simultaneously, he believes AI agents could assist in identifying problematic markets or attempts at manipulation by recognizing suspicious patterns.

“Agents could detect patterns and help shut down problematic markets,” he said.

Creating a user-owned AI economy

For Minarsch, the ultimate aim is not merely improved trading strategies.

It is to ensure that everyday users maintain a stake in an increasingly automated digital economy.

A future where AI systems dominate most economic activities could risk disenfranchising individuals if centralized platforms control the technology. “Olas seeks to establish a world where human users can be empowered through their AI agents rather than marginalized by them.”

To counteract this trend, the project prioritizes user ownership of AI systems. “We aim to foster more user-owned agents,” Minarsch stated.

If successful, this model could enable individuals to deploy autonomous software that generates value on their behalf across markets and services. Prediction markets are just the beginning.

Read more: AI rout hits software stocks, but Grayscale says blockchains stand to benefit