Decentralized Prediction Markets (DePMs)

• Decentralized Prediction Markets (DePMs) are permissionless blockchain-based platforms enabling wagering on future events via decentralized price signals and game-theoretic protocols.
• They employ diverse mechanisms like AMMs, orderbooks, and splitting gadgets to ensure liquidity, dynamic pricing, and secure market resolution, with platforms like Polymarket and Augur leading the field.
• Key challenges include optimizing market microstructure, ensuring incentive-compatible event resolution, and balancing full permissionlessness with regulatory requirements.

Decentralized Prediction Markets (DePMs) are open, permissionless mechanisms enabling participants to wager on the outcomes of future events, with market operation, price discovery, trade execution, and resolution occurring without reliance on centralized intermediaries. These systems leverage blockchain infrastructure, automated financial mechanisms, decentralized oracles, and cryptoeconomic incentives to generate event-linked price signals, which are often interpreted as probability forecasts. Modern DePMs include platforms such as Polymarket, Augur, and specialized frameworks built on platforms like Ethereum, featuring composable modules and support for various liquidity and resolution strategies.

1. Historical Development and Architectural Taxonomy

The first proposals for DePMs emerged around 2011, building on concepts from earlier centralized markets but integrated with blockchain primitives and distributed consensus (Rahman et al., 17 Oct 2025). Early systems operated via sidechains or limited scriptable extensions to Bitcoin; the Ethereum vision paper (2013) shifted the locus of innovation toward smart contracts, enabling full-stack decentralized platforms.

Influential early designs include Truthcoin, which utilized automated bookmaking in which an operator sets initial prices and bears risk, and the Princeton DePM design, which introduced a passive market creator through combinatorial splitting gadgets. Truthcoin and its derivatives (e.g., Augur) inspired markets where a designated reporting process determines event outcomes. Over time, DePMs migrated from operator-exposed architectures toward systems that minimize trust by using cryptographic tokens, splitting/merging gadgets, decentralized oracles, and composable modules.

By 2025, DePMs such as Polymarket employ modular architectures spanning seven workflow stages:

Stage	Typical Mechanisms	Notable Variants
Infrastructure	Ethereum, L2s, custom app-chains	Fees, composability trade-offs
Market Topic	Freeform, user-submitted predicates	Centralized curation vs. permissionless
Share Structure/Pricing	Winner-take-all (WTA), Yes/No bundles	Scalar, negative risk (YNB-NR)
Trading	AMMs, CLOBs, splitting, bookmaking	Matching vs. splitting
Resolution	Oracles, token-weighted voting, forks	Self-resolving, third-party arbitration
Settlement	Smart contract-pull/push, batch claims	Reentrancy protection
Archiving	On-chain logs, IPFS, subgraph indexing	MEV/cheap talk logging

(Rahman et al., 17 Oct 2025) documents these modules and their associated trade-offs extensively.

2. Mechanisms for Price Discovery and Liquidity

Several pricing and trading mechanisms exist within the DePM literature:

• Automated Market Makers (AMMs): Continuous pools such as those governed by constant product or utility-indifference pricing rules guarantee on-chain liquidity and allow for dynamic, path-dependent price formation (Amini et al., 2023, Pillay, 28 Feb 2025). The AMM’s cost function $C(x; \Pi)$ satisfies convexity, no arbitrage, liquidity-bounded loss, and path independence (except in constant-product AMMs, where path dependence is mathematically demonstrated). The marginal price or "oracle" is defined by the derivative of the cost function:

$$P^a(x; \Pi) = \lim_{t \downarrow 0} \frac{1}{t}C(tx; \Pi)$$

Pooling or withdrawing liquidity is performed via proportional scaling trades that preserve price invariance (Amini et al., 2023).

• Orderbooks and Matching Engines: Centralized orderbook architectures, or off-chain matching with on-chain settlement, can be used (e.g., Polymarket), but typically incur more complex slippage and latency risks. Bookmaking mechanisms (e.g., initial operator risk via a scoring rule) are now less common but have formal ties to proper scoring rule–based models (Conitzer, 2012).
• Splitting Gadgets: Users commit collateral to create a complete set of payoff shares (e.g., one for each possible outcome). This automatically ensures completeness and mutual exclusivity without exposing the operator to risk during initial issuance (Rahman et al., 17 Oct 2025).

Cryptocurrencies—especially stablecoins and, more recently, BTC—are the default numeraire, with settlement structures varying according to liquidity provisioning model (AMMs, cross-market making, DeFi-based redirection) (Shabashev, 15 Sep 2025).

3. Market Resolution: Oracles, Voting, and Game-Theoretic Protocols

In DePMs, market outcome resolution can be classified as follows:

• Token-Staked Oracle Voting: Protocols such as Augur rely on outcome reporting by reputation-token holders, who stake tokens and engage in successive dispute rounds. If repeated disputes occur, the system may fork, splitting the token into distinct universes—only the version corresponding to the objectively correct outcome retains value, thereby penalizing dishonest reporters with absolute loss (Peterson et al., 2015).

$B(\omega, n) = 2A_n - 3S(\omega, n)$

defines the dispute bond size, scaling with dispute escalation.

• Crowdsourced Arbiters and Peer Prediction: Some mechanisms aggregate the outcome votes of a randomly selected group of arbiters (possibly with positions in the market), using peer prediction schemes such as the 1/prior with midpoint mechanism to incentivize truthful reporting. Payments are calibrated by the informativeness of private signals ($\delta = \mu_1 - \mu_0$) and the arbiter’s maximum possible market stake (Freeman et al., 2016).
• Mechanism-Design Frameworks: Proper scoring rule–based models formalize incentives such that each agent’s best response is truthful information reporting. Individual, marginal, group, and pivotal information mechanisms have been shown (under consistency assumptions) to be incentive compatible, and connections to cooperative game theory (e.g., the Shapley value for marginal contributions) enable fair payment distribution. Practical variants separate roles for information and prediction agents, leveraging scoring rules and market scoring rules (Conitzer, 2012).
• Autonomous, Arbiter-Free Protocols: Colored Coins–based architectures directly encode outcome claims in the blockchain. Upon resolution, only tokens corresponding to the correct outcome retain value; enforcement is through chain protocol rules and game-theoretic incentives ensuring that deviation from honest behavior is strictly suboptimal (Bentov et al., 2017).
• DEX Oracles and TWAP: Some systems derive outcomes from time-weighted average prices reported by decentralized exchanges, embedding price oracle logic within the DePM smart contracts (Corn et al., 2021).

4. Information Aggregation, Inference, and Utility Theory

DePMs function as decentralized information aggregation engines. From a machine learning perspective, agents (traders, or programmatic entities) are characterized by probabilistic beliefs $P_{(i)}$ over event spaces and utility functions $U(W)$ on wealth (Storkey, 2011). By varying agent utility, market equilibrium replicates different model combination schemes:

• Product of Experts: Exponentially decaying negative utility $U_E(W) = -\exp(-W)$ yields geometric-mean aggregation.
• Mixture of Experts: Logarithmic utility $U_L(W) = \log W$ results in wealth-weighted averaging.
• Median/Weighted Voting: Linear utility $U_S(W) = W$ for $W>0$ selects the (weighted) median.

Parallelized, distributed inference emerges naturally, as market messages (price updates) can be structured to mimic message-passing algorithms (e.g., for factor graphs and graphical models). This setup allows for both large-scale parallel model building and efficient inference across heterogeneous agents, supporting applications in distributed sensor fusion, financial forecasting, and probabilistic modeling (Storkey, 2011).

5. Arbitrage, Market Efficiency, and Path Dependence

Arbitrage and market efficiency are central to DePM performance:

• Market Rebalancing Arbitrage: When the sum of mutually exclusive "YES" token prices is unequal to 1, arbitrageurs can realize risk-free profit by buying/selling bundles of shares:

$$\text{Long arbitrage if} \sum_{i=1}^n v_{Y_i}(t) < 1$$

$$\text{Short arbitrage if} \sum_{i=1}^n v_{Y_i}(t) > 1$$

Empirical analysis on Polymarket data identified more than \$40 million in realized arbitrage profits, primarily due to temporary liquidity gaps and combinatorial mispricing across related markets (Saguillo et al., 5 Aug 2025).

• Combinatorial Arbitrage: Logical dependencies among outcome conditions spanning different markets can admit riskless profit strategies when joint pricing is inconsistent. Heuristic-driven reductions, leveraging LLMs for semantic dependency extraction, are employed to identify tractable arbitrage opportunities amid combinatorial explosion (Saguillo et al., 5 Aug 2025).
• Path Dependence in AMM-Based Markets: Mathematical proofs demonstrate that constant product AMMs exhibit path dependence—the final market state depends on the sequence of liquidity provision and swaps:

$x \times y = k$

Non-commutativity implies that the resulting prices may encode not only information but also operational artifacts. Simulation and empirical results confirm price deviations (e.g., up to 0.68%) caused solely by differing event sequences, challenging the interpretation of AMM prices as pure "truth machines" (Pillay, 28 Feb 2025).

6. Liquidity Provisioning and Settlement Assets

Liquidity provisioning in DePMs is realized through various mechanisms, each with distinct capital efficiency and risk characteristics (Shabashev, 15 Sep 2025):

• Cross-Market Making: Professional makers mirror USD-based market liquidity into BTC-denominated venues, hedging FX risk via derivatives.
• Automated Market Making (AMM): Guarantees continuous on-chain liquidity but exposes providers to permanent loss and is generally capital-inefficient.
• DeFi-Based Redirection: Uses synthetic borrowing/lending (e.g., by collateralizing wBTC to borrow USDC) to convert BTC liquidity into stablecoin market participation, at the cost of potential liquidation and exchange-rate risk.

BTC has been proposed as a settlement asset in analogy to historical gold standard practice, enabling participants to avoid opportunity costs associated with stablecoins and preserve Bitcoin's store-of-value advantage (Shabashev, 15 Sep 2025). Trade-offs among these methods affect user safety, capital efficiency, and risk profiles.

7. Open Problems and Research Directions

Significant challenges and research frontiers remain:

• Market Microstructure Optimization: Adapting AMM models to outcome share tokens, which are bounded and exhibit discrete jumps at resolution, remains unresolved (Rahman et al., 17 Oct 2025).
• Market Topic Formalization: Avoiding ambiguous or manipulable event definitions is a major issue; the use of formal predicate specification and verification is under-explored (Rahman et al., 17 Oct 2025).
• Incentive-Compatible Resolution: Mechanisms resilient to manipulation, collusion, or sybil attacks—especially in token-weighted voting or peer-prediction frameworks—demand further formal analysis and innovation (Conitzer, 2012, Freeman et al., 2016).
• Archival and Data Reproducibility: Making large volumes of trade, order, and resolution data accessible and replayable (for both ex post analysis and governance accountability) is an unsolved infrastructural problem (Rahman et al., 17 Oct 2025).
• Permissionlessness vs. Regulatory Compliance: The challenge of maintaining fully permissionless markets while managing regulatory risk associated with unlawful or controversial market topics is ongoing.

Further, improving composability, modular swapping of protocol components, and path-aware mechanisms to address non-informational price artifacts are key areas of active investigation (Pillay, 28 Feb 2025, Rahman et al., 17 Oct 2025).

---

Decentralized prediction markets have become a locus of interdisciplinary innovation, combining foundational economic theory, distributed systems engineering, mechanism design, cryptoeconomic incentives, and formal methods. The current ecosystem supports a variety of design choices, each with unique trade-offs in decentralization, expressiveness, manipulation resistance, and user safety. The evolving research agenda addresses both foundational aspects (e.g., the economic theory of agent aggregation and prediction via markets) and practical issues concerning implementation, scalability, and regulatory accommodation.