AI Agents Quietly Dominate Prediction Markets, Squeezing Retail Traders

AI agents are quietly reshaping prediction-market trading — and retail traders are already feeling the pressure. Valory AG, the team behind the crypto-AI infrastructure Olas (formerly Autonolas), says autonomous agents are becoming a powerful force in forecasting and trading on prediction markets. David Minarsch, Valory’s CEO and co-founder, describes Olas as plumbing for an emerging “agent economy”: a decentralized ecosystem where autonomous software agents run services on blockchains, interact with smart contracts, cooperate, and earn crypto rewards for users. A live test of that vision arrived in February 2026 with Polystrat, an autonomous agent deployed on Polymarket. Polystrat trades 24/7 on behalf of self-custodying users, executing strategies continuously so human owners aren’t sidelined by sleep, distraction or inconsistent decision-making. “In a nutshell, Polystrat is an autonomous AI agent that trades on Polymarket 24/7 on behalf of its human user,” Minarsch says. Why this matters Prediction markets — platforms that let traders buy and sell contracts tied to real-world outcomes — have exploded from niche forecasting tools into a sizable corner of fintech. The sector’s breakout came during the 2024 U.S. presidential election, and by 2025 total notional trading across major platforms topped $44 billion, with monthly peaks near $13 billion. Trading is concentrated: Kalshi (a U.S.-regulated exchange overseen by the CFTC) and crypto-native Polymarket account for roughly 85–97% of activity. Against that backdrop, a simple insight has driven the shift toward automation: the predictive power of modern AI hasn’t been fully applied to markets. Valory began building a “prediction market economy” on Olas in 2023 to change that — connecting agents to data pipelines and prediction tools to forecast outcomes and trade them. Machines, not emotions Minarsch notes off-the-shelf language models prompted on markets typically perform no better than a coin flip. But when state-of-the-art models are embedded in custom workflows — what Valory calls prediction tools — they can achieve substantially higher accuracy. The firm claims these workflows have historically shown predictive accuracy up to 70% or more. Market data point to a machine advantage: only about 7–13% of human traders reportedly generate positive returns on prediction markets, while early signs of machine participation are rising fast. Analytics platform LayerHub finds more than 30% of wallets on Polymarket are already using AI agents. The results from Polystrat are eye-catching. Within roughly a month of launching, it executed over 4,200 trades on Polymarket, recording individual-trade returns as high as 376%. According to Valory, more than 37% of Polystrat agents posted positive P&L — “less than half that number” of human participants reportedly post gains. Leveling the playing field — and expanding markets Olas aims to democratize access to these agents so retail users aren’t outcompeted by institutional automation. Users can configure agents for strategy preferences, data sources and risk tolerance. Minarsch highlights another opportunity: the long tail of niche markets that humans often ignore. “You just point the agent at the problem and it does the work,” he says, which could unlock forecasting on localized or specialized questions and turn prediction markets into richer data sources for businesses and policymakers. Agents could also be augmented with proprietary datasets, allowing users to encode their own knowledge bases into trading strategies — potentially creating stronger, more principled performance than humans can sustain manually. Ethics, regulation and detection The rise of AI-driven markets raises thorny questions. Critics warn that markets on sensitive topics — wars, deaths or disasters — could incentivize manipulation or moral hazards. Minarsch concedes regulation will be needed to define what markets should exist, but he also argues agents could help policing efforts: automated systems can spot suspicious patterns and flag problematic markets. A user-owned future For Valory, the larger mission is political as much as technical: to ensure everyday users keep a stake in an automated digital economy. “Olas aims to create a world where human users can be empowered through their AI agents rather than disenfranchised by them,” Minarsch says. The project emphasizes user-owned agents so people can deploy autonomous software that generates value on their behalf — with prediction markets serving as an early proving ground. If that model succeeds, autonomous agents may become routine tools for forecasting, trading and decision-support across industries — and the next battleground for who controls value in a machine-driven economy. Read more AI-generated news on: undefined/news