Manipulation in Prediction Markets: An Agent-based Modeling Experiment

Bridget Smart, Ebba Mark, Anne Bastian, Josefina Waugh · 2026

Read Manipulation in Prediction Markets: An Agent-based Modeling Experiment by Bridget Smart, Ebba Mark, Anne Bastian, Josefina Waugh (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-20452.txt.

Introduction

Bridget Smart, Ebba Mark, Anne Bastian, and Josefina Waugh model manipulation in prediction markets with agents. Prediction markets pool money on event outcomes and aggregate scattered beliefs into prices. Their accuracy in elections has drawn both praise and worry about manipulation and its effects on democratic processes.

The authors run an agent-based modeling experiment to test how manipulation plays out. Simulated traders let them vary conditions that real markets cannot. Knowing whether a lone wealthy actor can bend prices, and for how long, informs how seriously election forecasts should be taken. The stakes are real, since these prices are read as news.

Catalog record

CREATORBridget Smart, Ebba Mark, Anne Bastian, Josefina Waugh
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE63 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-20452

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Manipulation in Prediction Markets: An Agent-based Modeling Experiment Bridget Smart, Ebba Mark, Anne Bastian, Josefina Waugh Abstract Prediction markets mobilize financial incentives to forecast binary event outcomes through the aggregation of dispersed beliefs and heterogeneous information. Their growing popularity and demonstrated predictive accuracy in political elections have raised speculation and concern regarding their susceptibility to manipulation and the potential consequences for democratic processes. Using agent-based simulations combined with an analytic characterization of price dynamics, we study how high-budget agents can introduce price distortions in prediction markets. We explore the persistence and stability of these distortions in the presence of herding or stubborn agents, and analyze how agent expertise affects market-price variance. Firstly we propose an agent-based model of a prediction market in which bettors with heterogeneous expertise, noisy private information, variable learning rates and budgets observe the evolution of public opinion on a binary election outcome to inform their betting strategies in the market. The model exhibits stability across a broad parameter space, with complex agent behaviors and price interactions producing self-regulatory price discovery. Second, using this simulation framework, we investigate the conditions under which a highly resourced minority, or “whale” agent, with a biased valuation can distort the market price, and for how long. We find that biased whales can temporarily shift prices, with the magnitude and duration of distortion increasing when non-whale bettors exhibit herding behavior and slow learning. Our theoretical analysis corroborates these results, showing that whales can shift prices proportionally to their share of market capital, with distortion duration depending on non-whale learning rates and herding intensity. Keywords: prediction markets, agent-based model, betting markets, market manipulation 1 1 footnotetext: Corresponding authors: Bridget Smart, [email protected] , and Ebba Mark, [email protected] . 1. Introduction Prediction markets, where participants wager on binary outcomes, have been proposed as effective mechanisms for aggregating dispersed beliefs into accurate predictors in environments characterized by uncertainty and heterogeneous access to information [ 44 , 2 ] . Their appeal rests on the idea that traders, motivated by financial incentives, reveal private information through their willingness to buy or sell contracts corresponding to a particular outcome. Prediction market prices have been found to closely approximate the mean belief about an event’s probability in theoretical settings with well-informed, constrained and risk-adverse traders [ 57 , 21 ] . In several real elections, prediction-market forecasts have matched or outperformed traditional polling or expert judgment [ 42 , 5 , 55 , 12 ] , although evidence is mixed [ 28 , 15 ] . Positive examples are sometimes attributed to the “wisdom of crowds”, which relies on assumptions of independence, diversity, and the absence of dominant participants, conditions which are not always satisfied in practice [ 48 ] . Much of the empirical evidence supporting prediction-market accuracy comes from markets with strict limits on individual trading positions. Early studies of prediction market accuracy and dynamics largely relied on platforms with strict per-trader position limits, such as the Iowa Electronic Markets [ 5 ] and PredictIt [ 39 ] , where individual exposure was constrained to a few thousand U.S. dollars per market (e.g., $3,500 on PredictIt as of 2025 [ 51 ] ). These caps strictly limited any single participant’s ability to meaningfully influence prices. In contrast, newer platforms permit substantially larger, contract-specific position limits for eligible participants, in some cases reaching the multi-million-dollar range. Under these conditions, the classic assumption that individual traders are too small to affect prices, and that any attempted manipulation is rapidly arbitraged away [ 23 ] , is no longer guaranteed to hold. Both theoretical and empirical work shows that when traders are sufficiently capitalized or face strong incentives, prediction markets may sustain non-informational price deviations for meaningful periods [ 11 , 2 ] . This institutional shift has coincided with a rapid expansion in prediction market activity. During the 2024 U.S. presidential election, market capitalization reached hundreds of millions of USD [ 26 , 36 ] . For example, weeks prior to the 2024 U.S. presidential election between Donald Trump and Kamala Harris, a consequential ruling by the Commodity Futures Trading Commission legalized KalshiEx [ 52 ] . Within days, the platform announced its ability to accommodate individual trades of up to $100 million [ 20 , 29 ] , with notional volume exceeding $20 billion in December 2025 [ 26 ] , marking a departure from smaller, tightly-capped markets. In this context, a type of market distorter commonly referred to as a ‘‘whale’’ 1 1 1 “Whales” refer to participants in a betting market who control and deploy a volume of capital significantly larger than the average participant, thereby potentially possessing the ability to temporarily influence market prices through their trades or wagers. has been identified across multiple platforms in recent years [ 38 , 18 ] . For example, in 2024, Polymarket identified a $45 million bet placed by a French national to favor Donald Trump, temporarily pushing up his odds in the prediction market [ 38 , 46 ] . At such levels of capital concentration, salient price movements may interact with heterogeneity in how traders weight prior beliefs, market-external information, and market prices, generating dynamics that differ qualitatively from those observed in smaller, tightly capped markets. The potential for such market manipulation invites a suite of questions regarding the potential feedback between prediction markets and real-world political outcomes. Their credibility as predictive tools could earn them a higher rank in the information landscape available to voters when shaping their voting decisions. Research on informational cascades and bandwagon effects suggests that shifts in public beliefs triggered by prominent information signals like polls can influence real political outcomes, making endogeneity between market prices and behavior difficult to rule out [ 31 , 32 , 8 , 1 , 40 ] . Anecdotal and journalistic evidence indicates that members of the public have begun to rely on prediction market values in addition to polls to form their understanding of public opinion [ 16 , 20 ] . Thus, a potential feedback between market price and electorate beliefs raises a logical concern about incentives for strategic manipulation of market price to potentially distort democratic processes. This motivates our central question: Under what conditions do contemporary prediction markets function as self-correcting information processing mechanisms, and when can they instead temporarily sustain price deviations introduced by a highly resourced, biased minority? We address this by examining the prediction market as an information-aggregating mechanism and how trader behavior can introduce volatility and error into the market. In particular, we consider how large trades can lead to sustained or fluctuating price movements through a biased large trader (whale) operating among traders who partially herd toward the market price. 1.1. Methodological Approach and Contributions In this study, we develop an open source Agent-Based Model (ABM) which simulates a prediction market in which betting agents, characterized by heterogeneous learning functions, expertise, biased perceptions, and budgets, bet on the outcome of a binary election. We use this model to consider how wealth distribution, risk aversion, herding and belief dynamics influence prediction market price. Our proposed agent-based model framework exhibits stable performance and accurately predicts the data-informed stylized election outcome across a wide range of parameter values. Furthermore,

Read the complete text