AI Agents Are Quietly Leading a Revolution in Prediction Markets

Prediction markets have rapidly evolved into a booming fintech sector, diverging from traditional investment tools. But the most notable aspect of this growth is the shifting power from human traders to AI agents. Recent months have shown that machines executing automated strategies 24/7 are achieving far better results than retail users.

The Edge of Machines in Prediction Markets: Data Speaks

David Minarsch, founder of Valory AG, explained that the vision behind the crypto-AI protocol Olas is to bridge the performance gap between humans and machines in prediction markets. Olas is designed as an infrastructure for autonomous software agents that can run services on blockchains, interact with smart contracts, and coordinate to earn crypto rewards.

Data shows that machines have a clear performance advantage in prediction markets. According to third-party sources, only 7% to 13% of human traders achieve positive results, while the rest incur losses. This underscores why machine learning, structured data analysis, and disciplined trading strategies are crucial, making machines inherently superior.

“General-purpose large language models often don’t perform well in prediction markets,” Minarsch said. “But cutting-edge AI models integrated into specific workflows have historically shown prediction accuracy above 70%.” This difference signals not just a tool shift but a qualitative transformation in prediction markets.

Olas’s Agent Economy Vision and Polystrat’s Early Wins

Olas aims to realize a broader vision through its AI agents, which Minarsch calls an “agent economy”: a fully decentralized ecosystem where autonomous AI agents perform useful tasks and create value for users.

The most tangible example of this vision is Polystrat, an AI agent launched about a month ago on Polymarket. Polystrat is designed as an autonomous system that provides self-protection and trades continuously on behalf of its owner, 24/7. While users sleep, work, or lose focus, Polystrat seamlessly executes its strategies.

Early performance of Polystrat has demonstrated the viability of this model. In its first month, the agent completed over 4,200 trades on Polymarket. Even more impressive, data shared by the team shows some trades yielded returns of up to 376%. Compared to human participants, over 37% of Polystrat’s and similar agents’ trades are profitable, whereas the rate for human traders is below 18%.

Why Are Autonomous Agents More Efficient Than Human Traders?

The reasons for machines’ success in prediction markets are simple but fundamental: first, the complete absence of emotional decision-making. Humans are often driven by fear, greed, or frustration. In contrast, AI agents mechanically adhere to consistent strategies.

Second, these agents are free from human biological limits. Humans face sleep, fatigue, and distraction. Polystrat and similar systems can scan and evaluate opportunities every second the markets are open. According to LayerHub analytics, over 30% of wallets on Polymarket already use AI agents—a rapidly growing trend.

Minarsch views this trend as a broader sign of change: “There are many machine and human participants in prediction markets. So, humans are already in a contest with machines.” Whether aware or not, human investors are competing against automated systems. Olas aims to level the playing field and even give humans an advantage by offering AI agents for everyday users.

The Long Tail of Prediction Markets: The Goldmine of Niche Questions

Beyond Polystrat’s quick success, Minarsch believes AI agents can unlock an often-overlooked opportunity in prediction markets: the long tail of niche and localized questions.

Most prediction markets focus on major global events, elections, macroeconomic indicators, or high-profile sports. But many smaller, local, or specialized questions largely escape the attention of human traders.

“People usually don’t bother to seek out information,” Minarsch said. “They don’t have the time or effort.” But AI agents can do this. They can analyze hundreds of niche markets simultaneously. “The long tail of prediction markets is a very attractive area for AI agents. You just direct the agent to the problem, and it handles the rest.”

This dynamic can help prediction markets expand as a data collection tool for businesses, policymakers, and decision-makers. Prediction markets have long been studied as a way to aggregate dispersed information and uncover insights that traditional surveys might miss. With the catalytic power of AI agents, this activity could evolve into a form of upstream technology influencing decision-making across sectors.

Human-AI Collaboration: Complementary Powers

Despite the rise of automation, Minarsch does not see AI agents fully replacing humans. Instead, he describes them as complementary tools.

“Humans tend to make decisions more hastily, which can be harmful,” he said. “AI agents can serve as a reliable system that humans can trust.” In the future, users may enhance their agents with private information or specialized data sets.

“We see demand from users to access their own knowledge bases or proprietary data,” Minarsch added. “This will enable agents to operate more principled and aligned with human preferences.” Over time, prediction models and data pipelines are expected to improve significantly, especially as they integrate with general-purpose large language models, creating continuous new opportunities.

Regulation, Ethics, and the Future of the Agent Economy

The rapid growth of prediction markets also raises ethical and regulatory questions. Critics warn that predicting harmful events like wars, deaths, or natural disasters could create incentives to manipulate outcomes or profit from catastrophes.

Minarsch acknowledges the need for careful safeguards. “Regulation is needed on what kinds of prediction markets should exist.” He also believes AI agents can help detect suspicious patterns, identify problematic markets, or uncover manipulation attempts. “Agents can spot patterns and help shut down problematic markets.”

Balancing these two forces—automation’s speed and regulation’s caution—is a critical challenge.

A User-Owned AI Economy

Minarsch’s ultimate goal is not just better trading strategies but ensuring everyday users retain ownership and control in an increasingly automated digital economy.

As AI systems conduct most economic activities, if control remains centralized on platforms, individuals’ rights to protect their interests could be seriously compromised. Olas emphasizes user ownership in AI systems to reverse this dynamic.

“We want to create more user-owned agents,” Minarsch said. “Users should be empowered, not disempowered, by AI agents.” If successful, this model could enable people to deploy autonomous software that creates value on their behalf across markets and services.

Prediction markets are just the beginning of this broader agent economy vision. Combining technological advances with economic incentives, AI agents and prediction markets could play a foundational role in fintech and decision-making systems over the next five years.