IOSG Weekly Brief | Make Probability an Asset: Predictive Market Agents Preview #315

Original author: Jacob Zhao, IOSG Ventures

In our previous Crypto AI research reports, we have consistently emphasized that the most practically valuable use cases in the current crypto space are primarily centered around stablecoin payments and DeFi, while agents serve as the key user interface for the AI industry. Therefore, within the trend of crypto-AI convergence, the two most valuable pathways are: short-term AgentFi built on existing mature DeFi protocols (such as lending, yield farming, and advanced strategies like Swap, Pendle PT, and funding rate arbitrage), and medium-to-long-term Agent Payment centered on stablecoin settlement, leveraging protocols such as ACP/AP2/x402/ERC-8004.

Prediction markets have emerged as an undeniable new industry trend in 2025, with annual total trading volume surging from approximately $9 billion in 2024 to over $40 billion in 2025, achieving more than 400% year-over-year growth. This significant expansion has been driven by multiple factors: increased demand fueled by macro political uncertainty, maturing infrastructure and trading models, and a breakthrough in regulatory environments (Kalshi’s legal victory and Polymarket’s return to the U.S.). Prediction Market Agents began taking early shape in early 2026 and are poised to become a new product form in the agent landscape over the coming year.

Prediction markets: From betting tools to the 'global layer of truth'

Prediction markets are a financial mechanism for trading on the outcomes of future events, where contract prices inherently reflect the collective judgment of the market regarding the probability of those events occurring. Their effectiveness stems from the combination of collective wisdom and economic incentives: in an anonymous environment where participants risk real money, dispersed information is rapidly aggregated into price signals weighted by capital commitment, significantly reducing noise and false judgments.

▲ Notional Trading Volume Trend for Prediction Markets | Data Source: Dune Analytics (Query ID: 5753743)

By the end of 2025, the prediction market had largely evolved into a duopoly dominated by Polymarket and Kalshi. According to Forbes, total trading volume in 2025 reached approximately $44 billion, with Polymarket contributing about $21.5 billion and Kalshi around $17.1 billion. Weekly data from February 2026 showed Kalshi’s trading volume ($25.9B) surpassing Polymarket’s ($18.3B), nearing 50% market share. Kalshi achieved rapid expansion through its legal victory in election contract cases, first-mover advantage in regulatory compliance within the U.S. sports prediction market, and clearer regulatory expectations. Currently, the development paths of both platforms have become clearly differentiated:

Polymarket employs a hybrid CLOB architecture with off-chain matching and on-chain settlement, along with a decentralized settlement mechanism, to build a global, non-custodial, high-liquidity market; after regaining compliance in the U.S., it has established a dual-track operational structure of "onshore and offshore."

· Kalshi integrates with traditional financial systems by connecting via API to major retail brokerages, attracting Wall Street market makers to deeply participate in macro and data-driven contract trading; however, its products are constrained by traditional regulatory processes, leading to slower responses to long-tail demands and unexpected events.

Besides Polymarket and Kalshi, other competitive players in the prediction market space primarily follow two paths:

· First, a compliant distribution pathway: embedding event contracts into existing accounts and clearing systems of brokers or large platforms, leveraging channel reach, regulatory qualifications, and institutional trust to gain an advantage (e.g., Interactive Brokers × ForecastEx’s ForecastTrader, FanDuel × CME Group’s FanDuel Predicts), with significant compliance and resource advantages, but still in early stages of product and user adoption.

Second, the crypto-native on-chain path, represented by Opinion.trade, Limitless, and Myriad, achieves rapid scaling through reward mining, short-term contracts, and media distribution, emphasizing performance and capital efficiency; however, its long-term sustainability and risk control robustness remain to be validated.

The two pathways—traditional financial compliance entry points and crypto-native performance advantages—together form a diverse competitive landscape for the prediction market ecosystem.

Prediction markets may appear similar to gambling and are fundamentally zero-sum, but their key distinction lies in whether they generate positive externalities: by aggregating dispersed information through real-money transactions, they publicly price real-world events, creating a valuable layer of signals. The trend is shifting from gambling toward a “global truth layer”—as institutions like CME and Bloomberg integrate, event probabilities are becoming decision-making metadata directly accessible to financial and corporate systems, delivering more timely and quantifiable market-based truths.

From the perspective of global regulatory trends, the compliance pathways for prediction markets are highly divergent. The United States is the only major economy that has explicitly classified prediction markets within its financial derivatives regulatory framework, while markets such as Europe, the UK, Australia, and Singapore generally treat them as gambling and are moving toward stricter regulation; China, India, and others have imposed complete bans. The future global expansion of prediction markets will continue to depend on individual countries’ regulatory frameworks.

Architecture Design of Prediction Market Agents

Prediction Market Agents are currently entering the early adoption phase; their value lies not in “AI making more accurate predictions,” but in amplifying information processing and execution efficiency within prediction markets. Prediction markets are fundamentally information aggregation mechanisms, where prices reflect collective judgments about event probabilities. Real-world market inefficiencies arise from information asymmetry, liquidity constraints, and attention limits. The appropriate role of Prediction Market Agents is Executable Probabilistic Portfolio Management: transforming news, rule-based texts, and on-chain data into verifiable pricing discrepancies, enabling faster, more disciplined, and lower-cost strategy execution, while capturing structural opportunities through cross-platform arbitrage and portfolio risk management.

An ideal prediction market agent can be abstracted into a four-layer architecture:

· The information layer aggregates news, social, on-chain, and official data;

The analysis layer uses LLMs and ML to identify mispricings and calculate edge;

The strategy layer converts Edge into positions using the Kelly Criterion, tiered position building, and risk management.

The execution layer completes multi-market order placement, slippage and gas optimization, and arbitrage execution, forming an efficient automated闭环.

Strategy Framework for Prediction Market Agents

Unlike traditional trading environments, prediction markets differ significantly in settlement mechanisms, liquidity, and information distribution; not all markets or strategies are suitable for automated execution. The core of a prediction market agent lies in whether it is deployed in scenarios with clear, codifiable rules that align with its structural advantages. The following analysis will examine three aspects: selection of underlying assets, position management, and strategy structure.

Not all prediction markets have tradable value; their participation value depends on: clarity of settlement (whether rules are clear and data sources are unique), liquidity quality (market depth, spread, and trading volume), insider risk (degree of information asymmetry), time structure (expiration time and event timeline), and the trader’s own information advantage and expertise. Only when most dimensions meet basic requirements does a prediction market become viable for participation; participants should align their strengths with the market’s characteristics:

· Human core advantages: Markets where decisions rely on expertise, judgment, and integration of ambiguous information, with relatively relaxed time horizons (measured in days/weeks). Typical examples include political elections, macroeconomic trends, and corporate milestones.

· AI Agent Core Advantages: Markets that rely on data processing, pattern recognition, and rapid execution, with extremely short decision windows (measured in seconds or minutes). Typical examples include high-frequency cryptocurrency pricing, cross-market arbitrage, and automated market making.

· Not applicable: Markets dominated by insider information or purely random/highly manipulated markets, which confer no advantage to any participant.

The Kelly Criterion is the most representative money management theory for repeated betting scenarios, aiming not to maximize single-round returns, but to maximize the long-term compounded growth rate of capital. Based on estimates of win probability and odds, it calculates the theoretically optimal position size to enhance capital growth efficiency under positive expected value, and is widely applied in quantitative investing, professional gambling, poker, and asset management.

· The classic form is: f^* = (bp - q) / b

· Here, f∗ is the optimal bet size, b is the net odds, p is the win probability, and q=1−p

· The prediction market can be simplified to: f^* = (p - market_price) / (1 - market_price)

Here, p is the subjective true probability, and market_price is the implied market probability.

The theoretical validity of the Kelly Criterion heavily depends on accurate estimates of true probabilities and odds; in practice, traders find it difficult to consistently and accurately determine true probabilities, so professional gamblers and prediction market participants tend to favor more executable, rule-based strategies that rely less on probability estimates:

· Unit System: Divide your capital into fixed units (e.g., 1%), and allocate a varying number of units based on confidence level; this method automatically limits per-trade risk through unit caps and is the most common practical approach.

· Fixed Ratio Betting: Use a fixed stake amount for each bet, emphasizing discipline and stability, ideal for risk-averse traders or low-confidence environments.

· Confidence Tiers: Predefined discrete position levels with absolute upper limits to reduce decision complexity and avoid the pseudo-precision issues of the Kelly criterion.

· Inverted Risk Approach: Start with the maximum acceptable loss to back-calculate position size, establishing a stable risk boundary based on risk constraints rather than return expectations.

For prediction market agents, strategy design should prioritize executability and stability over theoretical optimality. The key lies in clear rules, simple parameters, and tolerance for judgment errors. Under these constraints, the stepped confidence method combined with a fixed position cap is the most suitable general position management solution for PM agents. This approach does not rely on precise probability estimates; instead, it categorizes opportunities into a limited number of tiers based on signal strength, each corresponding to a fixed position size—even in high-confidence scenarios, a clear upper limit is set to control risk.

From a strategic perspective, prediction markets can be primarily divided into two categories: deterministic arbitrage strategies, characterized by clear, codifiable rules; and speculative directional strategies, which rely on information interpretation and directional judgment. Additionally, there are market-making and hedging strategies, typically led by professional institutions and requiring significant capital and infrastructure.

· Settlement Arbitrage: Settlement arbitrage occurs when an event’s outcome is largely determined but the market has not yet fully priced it in; profits primarily stem from information synchronization and execution speed. This strategy has clear rules, low risk, and can be fully encoded, making it the core strategy most suitable for agents to execute in prediction markets.

· Probability Conservation Arbitrage (Dutch Book Arbitrage): Dutch Book arbitrage exploits structural imbalances arising when the sum of prices for a mutually exclusive and exhaustive set of events deviates from the probability conservation constraint (∑P≠1), locking in a directionless risk-free profit through a combined position. This strategy relies solely on rules and price relationships, carries low risk, and is highly rule-based, making it a classic example of deterministic arbitrage suitable for automated execution by agents.

· Cross-platform arbitrage: Cross-platform arbitrage profits from pricing discrepancies of the same event across different markets, offering lower risk but requiring high precision in latency control and parallel monitoring. This strategy is well-suited for agents with infrastructure advantages, but increasing competition continues to erode marginal returns.

· Bundle Arbitrage: Bundle arbitrage exploits pricing inconsistencies between related contracts; the logic is straightforward but opportunities are limited. This strategy can be executed by an Agent, though it requires moderate engineering effort for rule parsing and portfolio constraints.

· Structured Information-Driven Strategy (Information Trading): This strategy revolves around well-defined events or structured information, such as official data releases, announcements, or ruling windows. As long as the information source is clear and the triggering conditions are definable, the Agent can leverage its speed and discipline advantages in monitoring and execution; however, human intervention is still required when the information shifts to semantic judgment or contextual interpretation.

· Signal Following Strategy: This strategy generates returns by following the actions of accounts or funds with historically strong performance; its rules are relatively simple and can be automated. Its primary risks are signal degradation and reverse exploitation, requiring filtering mechanisms and strict position management. Suitable as an auxiliary strategy for an Agent.

· Unstructured / Noise-driven strategies: These strategies heavily rely on sentiment, randomness, or participation behavior, lack a stable and replicable edge, and have unstable long-term expected value. Due to difficulty in modeling and extremely high risk, they are unsuitable for systematic execution by agents and are not recommended as long-term strategies.

· High-frequency price and liquidity strategies (Market Microstructure): These strategies rely on extremely short decision windows, continuous quoting, or high-frequency trading, demanding very low latency, sophisticated models, and substantial capital. Although theoretically suitable for Agents, they are often constrained by liquidity and competition in prediction markets, making them viable only for a select few participants with significant infrastructure advantages.

· Risk Management & Hedging: These strategies do not directly seek returns but are designed to reduce overall risk exposure. They operate continuously as a foundational risk control module with clear rules and well-defined objectives.

Overall, strategies suitable for agents in prediction markets focus on scenarios with clear, codifiable rules and minimal subjective judgment, where deterministic arbitrage should serve as the primary source of returns, structured information and signal-following strategies as supplements, and high-noise, sentiment-driven trading systematically excluded. The agent’s long-term advantage lies in its high discipline, speed of execution, and risk management capabilities.

Predictive market agent business model and product form

The ideal business model design for prediction market agents offers different avenues for exploration at various levels:

· Infrastructure layer, providing multi-source real-time data aggregation, Smart Money address database, unified prediction market execution engine, and backtesting tools, generating stable revenue through B2B fees unrelated to prediction accuracy;

· Strategy layer: Integrate community and third-party strategies to build a reusable and evaluable strategy ecosystem, capturing value through invocation, weighting, or execution splits, thereby reducing dependence on a single alpha source.

· Agent/Vault layer: Agents directly participate in live trading on a fiduciary basis, leveraging on-chain transparent records and a robust risk control system to collect management and performance fees.

Product forms corresponding to different business models can also be categorized as:

· Gamified/Entertainment Mode: Lowers the barrier to entry through Tinder-like intuitive interactions, offering the strongest user growth and market education capabilities, making it an ideal entry point for breaking out of niche circles, but must be monetized through subscription or execution-based products.

· Strategy Subscription / Signal Mode: No custody of funds, regulatory-friendly, clear allocation of rights and responsibilities, and a relatively stable SaaS revenue model—making it the most viable commercialization path at this stage. Its limitations include ease of strategy replication and execution losses, resulting in a limited long-term revenue ceiling. These can be significantly improved through a semi-automated “signal + one-click execution” format, enhancing user experience and retention.

· Vault custody model: Offers advantages in scale and execution efficiency, resembling asset management products, but faces multiple structural constraints including asset management licensing requirements, trust barriers, and centralized technology risks. Its business model is highly dependent on market conditions and sustained profitability. Unless it has a long-term performance track record and institutional backing, it is not recommended as the primary pathway.

Overall, a diversified revenue model combining “infrastructure monetization,” “strategy ecosystem expansion,” and “performance participation” helps reduce reliance on the single assumption that AI will consistently outperform the market. Even as alpha narrows with market maturation, foundational capabilities such as execution, risk management, and settlement retain long-term value, enabling a more sustainable business闭环.

Project examples of prediction market agents

Currently, prediction market agents are still in the early stages of exploration. Although the market has seen diverse attempts ranging from foundational frameworks to upper-layer tools, no standardized product has yet emerged that is mature in strategy generation, execution efficiency, risk control systems, and business闭环.

We have divided our current ecosystem into three tiers: Infrastructure, Autonomous Agents, and Prediction Market Tools.

· Polymarket Agents framework

Polymarket Agents is an official developer framework launched by Polymarket, designed to address engineering standardization issues around connectivity and interaction. The framework encapsulates interfaces for fetching market data, constructing orders, and basic LLM calls. It solves the problem of "how to place orders via code," but leaves core trading capabilities—such as strategy generation, probability calibration, dynamic position management, and backtesting systems—largely unaddressed. It functions more as an officially endorsed "integration specification" than a ready-to-use product with alpha-generating capabilities. Commercial-grade agents still require building a complete research, development, and risk management core on top of it.

· Gnosis Prediction Market Tools

The Gnosis Prediction Market Agent Tooling (PMAT) provides full read-write support for Omen/AIOmen and Manifold, but offers only read-only access to Polymarket, resulting in clear ecosystem barriers. It serves as an excellent foundation for developing agents within the Gnosis ecosystem, but has limited practicality for developers primarily focused on Polymarket.

Polymarket and Gnosis are currently the only prediction market ecosystems that have clearly productized "Agent development" into an official framework. Other prediction markets, such as Kalshi, still primarily operate at the API and Python SDK level, requiring developers to independently build critical system capabilities such as strategy, risk management, execution, and monitoring.

The "predictive market AI agents" currently available in the market are still in early stages; despite being labeled as "agents," their actual capabilities fall significantly short of enabling fully autonomous, closed-loop trading. Most lack independent, systematic risk management layers and have not integrated position management, stop-losses, hedging, or expected value constraints into their decision-making processes. Overall, productization remains underdeveloped, and no mature, long-term operational systems have yet emerged.

· Olas Predict

Olas Predict is the most productized predictive market agent ecosystem to date. Its core product, Omenstrat, is built on Omen within the Gnosis ecosystem, leveraging FPMM and a decentralized arbitration mechanism to support low-value, high-frequency interactions—but is constrained by limited liquidity in individual Omen markets. Its "AI predictions" primarily rely on general-purpose LLMs, lacking real-time data and systematic risk controls, with historical win rates varying significantly across categories. In February 2026, Olas launched Polystrat, extending agent capabilities to Polymarket—users can define strategies in natural language, and the agent automatically identifies probability discrepancies in markets settling within four days and executes trades. The system manages risk through local Pearl execution, self-hosted Safe accounts, and hardcoded constraints, making it the first consumer-grade autonomous trading agent designed for Polymarket.

· UnifAI Network Polymarket Strategy

Offer a Polymarket automated trading agent with a core strategy of tail risk exposure: scan for contracts nearing settlement with implied probabilities >95% and buy them, targeting a 3–5% spread. On-chain data shows a win rate close to 95%, but returns vary significantly across categories, making the strategy highly dependent on execution frequency and category selection.

· NOYA.ai

NOYA.ai aims to integrate the "research—judge—execute—monitor" cycle into an Agent闭环, with an architecture encompassing the intelligence layer, abstraction layer, and execution layer. Omnichain Vaults have already been delivered; the Prediction Market Agent is still under development and has not yet formed a complete mainnet闭环, with the overall project currently in the vision validation phase.

Current predictive market analysis tools are not yet sufficient to constitute a complete "predictive market agent"; their value primarily lies in the information and analysis layers of the agent architecture, while trade execution, position management, and risk control still require manual oversight by traders. In terms of product form, they better align with the positioning of "strategy subscription / signal assistance / research enhancement" and can be regarded as early prototypes of predictive market agents.

Through a systematic review and empirical screening of projects listed in Awesome-Prediction-Market-Tools, this report selects representative projects that have already developed initial product forms and use cases as case studies. The focus is primarily on four areas: analysis and signals layer, alerts and whale tracking systems, arbitrage discovery tools, and trading terminals with aggregation execution.

Market analysis tools

Polyseer: A research-oriented prediction market tool that employs a multi-agent architecture (Planner / Researcher / Critic / Analyst / Reporter) to gather bidirectional evidence and aggregate Bayesian probabilities, delivering structured research reports. Its advantages lie in methodological transparency, engineered workflows, and full open-source audibility.

Oddpool: Positioned as the "Bloomberg Terminal for prediction markets," offering cross-platform aggregation, arbitrage scanning, and real-time data dashboards for Polymarket, Kalshi, CME, and more.

Polymarket Analytics: A global data analysis platform for Polymarket, systematically presenting trader, market, position, and trading volume data with clear insights and intuitive visuals, ideal for foundational data queries and research reference.

Hashdive: A data tool for traders that quantitatively filters traders and markets through Smart Score and a multidimensional screener, offering practical utility in identifying smart money and making copy-trading decisions.

Polyfactual: Focuses on AI-driven market intelligence and sentiment/risk analysis, embedding insights directly into trading interfaces via a Chrome extension, with a focus on B2B and institutional user scenarios.

Predly: An AI mispricing detection platform that identifies pricing discrepancies on Polymarket and Kalshi by comparing market prices with AI-calculated probabilities. The official claim is an alert accuracy rate of 89%, positioned for signal discovery and opportunity screening.

Polysights: Covers 30+ market and on-chain metrics, and uses Insider Finder to track unusual activities such as new wallets and large single bets, ideal for daily monitoring and signal discovery.

PolyRadar: A parallel multi-model analysis platform that provides real-time interpretation, timeline evolution, confidence scoring, and source transparency for single events, emphasizing cross-validated AI analysis as a positioning tool.

Alphascope: An AI-driven predictive market intelligence engine providing real-time signals, research summaries, and probability change monitoring; still in early stages, focused on research and signal support.

· Alerts / Whale Tracking

Stand: Clearly define the positioning of whale tracking and high-confidence action alerts.

Whale Tracker Livid: Productizing whale position changes

Arbitrage Discovery Tool

ArbBets: An AI-driven arbitrage discovery tool focused on Polymarket, Kalshi, and sports betting markets, identifying cross-platform arbitrage and positive expected value (+EV) opportunities, positioned at the high-frequency opportunity scanning layer.

PolyScalping: A real-time arbitrage and scalping analysis platform for Polymarket, featuring full-market scans every 60 seconds, ROI calculations, Telegram alerts, and filtering by liquidity, spread, volume, and other metrics—ideal for active traders.

Eventarb: A lightweight, cross-platform arbitrage calculation and alert tool covering Polymarket, Kalshi, and Robinhood—focused features, free to use, ideal as a basic arbitrage aid.

Prediction Hunt: A cross-platform prediction market aggregator and comparison tool that provides real-time price comparisons and arbitrage identification for Polymarket, Kalshi, and PredictIt (refreshes approximately every 5 minutes), designed to uncover information symmetry and market inefficiencies.

· Trading Terminal / Aggregated Execution

A institutional-grade prediction market trading terminal backed by YC Fall 2024, offering a Bloomberg-style interface with real-time tracking of 15,000+ contracts from Polymarket and Kalshi, advanced data analytics, and AI-powered news intelligence, designed for professional and institutional traders.

Matchr: A cross-platform prediction market aggregation and execution tool covering 1,500+ markets, enabling optimal price matching through intelligent routing and automating yield strategies based on high-probability events, cross-market arbitrage, and event-driven opportunities, positioned at the execution and capital efficiency layer.

TradeFox: A professional prediction market aggregation and prime brokerage platform backed by Alliance DAO and CMT Digital, offering advanced order execution (limit orders, take-profit/stop-loss, TWAP), self-custody trading, and multi-platform smart routing—targeting institutional traders, with plans to expand to platforms such as Kalshi, Limitless, and SxBet.

Summary and Outlook

Currently, prediction market agents are in the early stages of development.

1. Market Foundation and Evolution: Polymarket and Kalshi have established a duopoly, providing sufficient liquidity and scenario foundations for agents built around them. The core distinction between prediction markets and gambling lies in positive externalities—aggregating dispersed information through genuine trading to publicly price real-world events, gradually evolving into a “global truth layer.”

2. Core positioning: The prediction market agent should be positioned as a probabilistic asset management tool capable of converting news, rule-based texts, and on-chain data into verifiable pricing discrepancies, and executing strategies with greater discipline, lower cost, and cross-market capability. An ideal architecture can be abstracted into four layers—information, analysis, strategy, and execution—but its actual tradability heavily depends on the clarity of settlement, the quality of liquidity, and the degree of information structuring.

3. Strategy Selection and Risk Control Logic: From a strategy perspective, arbitrage with high certainty (including settlement arbitrage, probability conservation arbitrage, and cross-platform spread trading) is best suited for automated execution by agents, while directional speculation should only serve as a supplement. In position management, priority should be given to executability and fault tolerance; a tiered approach combined with a fixed position cap is most suitable.

4. Business Model and Outlook: Commercialization is structured into three layers: the infrastructure layer generates stable B2B revenue through data execution infrastructure; the strategy layer monetizes via third-party strategy calls or revenue sharing; and the Agent/Vault layer participates in live trading under on-chain transparent risk controls, earning management and performance fees. Corresponding formats include entertainment-focused entry points, strategy subscriptions/signals (currently the most viable), and high-barrier Vault custody. The sustainable path is “infrastructure + strategy ecosystem + performance participation.”

Although the prediction market agents ecosystem has seen diverse efforts ranging from foundational frameworks to upper-layer tools, there is still no mature, replicable standardized product yet in key areas such as strategy generation, execution efficiency, risk control, and business closure. We look forward to the future iteration and evolution of prediction market agents.

Original link

Click to learn about the open positions at BlockBeats

Welcome to the official BlockBeats community:

Telegram subscription group: https://t.me/theblockbeats

Telegram group: https://t.me/BlockBeats_App

Official Twitter account: https://twitter.com/BlockBeatsAsia