Headline: AI agents are quietly taking over prediction-market trading — and retail traders are racing to keep up
Prediction markets — once niche forecasting tools — are rapidly evolving into a battleground between human intuition and autonomous machines. Valory AG, the team behind the crypto-AI protocol Olas, says the next wave of market participants won’t be people at all but autonomous AI agents that trade around the clock on behalf of users.
Valory’s CEO and co-founder David Minarsch frames this shift as the start of an “agent economy”: a decentralized ecosystem in which multifunctional AI agents run services on blockchains, interact with smart contracts, cooperate with each other, and earn crypto rewards for their owners. The company’s current focus, Olas (formerly Autonolas), provides the infrastructure for those agents and the prediction tools and data pipelines they need to forecast outcomes and place trades.
A live test of that idea arrived in February 2026 with Polystrat, an autonomous agent deployed to trade on Polymarket. Polystrat runs continuously for self-custodying users, executing strategies while humans sleep, work or lose focus. “Polystrat is an autonomous AI agent that trades on Polymarket 24/7 on behalf of its human user,” Minarsch said, noting the appeal of machines that stick to disciplined strategies without emotional bias.
Why it matters
- Prediction markets have scaled into a sizable fintech niche. Trading volumes exploded around the 2024 U.S. presidential election and kept growing through 2025: total notional trading across major platforms topped $44 billion, with monthly peaks near $13 billion.
- The sector is concentrated. U.S.-regulated Kalshi and crypto-native Polymarket together account for roughly 85–97% of trading volume, handling bets on elections, central bank policy, sports, culture and more. Kalshi operates under CFTC oversight; Polymarket runs globally and offers a broader array of markets.
- Machine participation is already material. Analytics provider LayerHub reports that over 30% of wallets on Polymarket are using AI agents.
Performance edge for machines
Valory’s push toward AI-driven trading sprang from a clear insight: the predictive power of modern AI models often hasn’t translated directly into market edges without specialized workflows. “Simply prompting off-the-shelf models with markets usually results in outcomes no better than a coin flip,” Minarsch said. But when state-of-the-art models are wrapped into “prediction tools” and combined with disciplined trading processes, Valory claims accuracy levels of 70% or higher in some setups.
Early performance data for Polystrat supports the contention that machines can outperform most humans. Within about a month of launch the agent executed more than 4,200 trades on Polymarket and produced single-trade returns up to 376%, according to Valory. The team reports over 37% of Polystrat agents showing a positive P&L, versus “less than half that number” for human participants — consistent with third-party data that only about 7–13% of human traders are net positive in prediction markets.
Design for retail users
Olas emphasizes user ownership and configurability: people self-custody their agents, pick strategies, choose data sources and set risk tolerances. Minarsch believes that giving retail users access to autonomous agents can level the playing field as markets grow more automated, preventing centralized platforms from hoarding the upside of AI trading.
Beyond headline markets: the long tail
One practical advantage for agents is scale. While humans focus on major global events, AI can sift thousands of smaller, niche markets simultaneously — the “long tail” that humans often ignore. That could broaden prediction markets’ usefulness as a real-time data-gathering tool for businesses and policymakers, surfacing granular forecasts that traditional surveys miss.
Risks and guardrails
The expansion of automated prediction trading raises ethical and regulatory questions. Critics warn that markets forecasting wars, deaths or disasters could create perverse incentives or become targets for manipulation. Minarsch acknowledges a need for rules about what markets should exist and suggests AI could help police the space: agents may be able to detect suspicious patterns and flag or shut down problematic markets.
A complementary future
Valory does not see agents as outright replacements for humans. Rather, Minarsch envisions AI agents as complements that act more consistently than humans and can be augmented with proprietary datasets or domain expertise to trade “in a more principled way.” The longer-term aim is an ecosystem where everyday users retain ownership and benefit directly from agent-driven economic activity.
Prediction markets may only be the starting point. If Olas and similar projects succeed, they could enable ordinary people to deploy autonomous software that generates value across markets and services — a user-owned alternative to opaque, centralized AI systems. For crypto traders and builders, the next phase of prediction markets will be as much about governance and ownership as it is about accuracy and alpha.
Read more AI-generated news on: undefined/news
