Manipulation in Thin Prediction Markets: What the Evidence Shows and How Traders Can Respond

Thin prediction markets are easier to move than deep ones. That is arithmetic. When the order book has a few hundred shares on each side, a modest trade can change the last price by several cents, and a change in the last price can be worth far more than the cost of the trade to someone who already holds a large position. That gap is a common incentive behind manipulation concerns.

But an odd price is not proof of manipulation. Thin markets also produce noisy prices, stale quotes and wide spreads for innocent reasons. The useful question is not "is this market manipulated?" It is "what would it cost someone to move this price, and who benefits?" The evidence below is a mix of controlled research, statistical detection and enforcement cases. None of it justifies accusing a particular trader without proof.

Why thin books are vulnerable

Imagine a market with 10,000 Yes shares held by one trader and a book with 100 shares at 20 cents and 100 at 25 cents. A 200-share buy costs $45 and leaves the last price at 25 cents. The holder's mark rises by $500. Even if the price fades, the higher mark can attract buyers or flatter the position's value.

The smaller the depth, the larger the ratio between the cost of moving the price and the change in the value of an existing position. The same logic applies to a trader who wants to influence a related market or a resolution source.

Several distinct behaviors get grouped under "manipulation":

• Spoofing: placing orders with no intention of executing them, to create a false impression of supply or demand.
• Wash trading: buying and selling to yourself or a colluding counterparty to inflate volume without taking a net position.
• Painting the tape: executing small trades at a favorable price to make the last trade look better than the real market.
• Insider trading: trading on material nonpublic information, which is an integrity problem even when it is not price manipulation.
• Settlement manipulation: trading the underlying asset or source that determines the contract's payout.

Not all are illegal everywhere, and not all are easy to detect. That is why the evidence matters more than the label.

What the research and reporting show

What the research shows

The cleanest evidence on manipulability comes from a field experiment across 817 prediction markets. Researchers randomly shocked prices and then collected hourly data to see whether the effects persisted. They found that prediction markets can be manipulated: the effects of the trades were visible even 60 days later. But the effects faded over time, and markets with more traders, greater volume and an external source of probability estimates were harder to manipulate.Rasooly and Rozzi

The result cuts both ways: manipulation is possible, but depth and independent information are the defenses. A thin market with no external benchmark is the easiest target.

The second line of evidence is about fake volume. A Columbia Business School working paper by Allen Sirolly, Hongyao Ma, Yash Kanoria and Rajiv Sethi uses a network-based method to detect clusters of accounts that trade mostly with each other. Applied to Polymarket, the authors estimate that transaction patterns indicative of wash trading began to rise in July 2024, peaked at nearly 60% of volume in December 2024, persisted through late April 2025, and rose again to about 20% of volume in early October 2025.Columbia Business School

The paper identifies patterns, not proof of fraud: some wash trading resembles legitimate automated market making, and identities are shielded on crypto-native venues.

What the reporting adds

In September 2026, CNBC reported that trading volumes on certain Kalshi and Polymarket products were raising questions. On Polymarket's international exchange, markets with multiple contracts showed more activity on low-probability outcomes than on likely winners, including in sports and elections. On Kalshi's ether perpetual futures, nearly half of the dollar volume on September 20 came from trades sized between $5,495 and $5,505, a pattern that observers said looked unusual relative to resting liquidity.CNBC

Both companies denied wash trading. Polymarket attributed the low-odds activity to sophisticated traders finding mispricings; Kalshi denied wash trading and said it had tracked hundreds of users. Experts CNBC spoke to were divided on whether the patterns established misconduct.

A pattern is a reason to look, not a verdict. The article also noted the Columbia lead author said the activity was negligible by April 2026; Polymarket credited surveillance and fees.

Documented disciplinary cases

The CFTC's enforcement advisory from February 2026 is more concrete because it describes two cases that Kalshi resolved through its own compliance process.

In May 2025, a political candidate appeared in videos that seemed to show him trading on his own candidacy. Kalshi's compliance team contacted him the same day. According to the CFTC, the trader acknowledged that he knew the trades were improper and violated exchange rules against trading a contract over which he had influence. Kalshi imposed a $2,246.36 financial penalty, including $246.36 of disgorgement and a $2,000 penalty, plus a five-year suspension from the exchange. The CFTC said the conduct potentially violated the Commodity Exchange Act's prohibitions on manipulative schemes and fraud.CFTC advisory

In a second case, an individual traded a prediction market tied to a YouTube channel while working as an editor for that channel, and likely had advance knowledge of the videos. Kalshi imposed a $20,397.58 penalty, including $5,397.58 of disgorgement and a $15,000 penalty, plus a two-year suspension. The CFTC described the conduct as potentially violating the prohibition on misappropriation of confidential information, commonly called insider trading.

The advisory also lists wash sales, pre-arranged trading, disruptive trading, fraud and manipulation as practices the Division can pursue. Platform rules and federal rules are separate layers: a trader can be suspended and still face a regulator.

Settlement manipulation is a separate risk

Some prediction markets settle on an asset price that traders can move by trading the underlying. A 2026 working paper by David Dai, Ruizhe Jia and Shihao Yu studies Polymarket's five-minute Bitcoin contract. It found settlement-time spot flow spikes and large post-settlement reversals, with manipulators profiting mostly from retail. The effect was largely absent in fifteen-minute contracts.Dai, Jia and Yu

This is a market-design problem. Short-dated contracts that settle on a manipulable price are structurally exposed; the fifteen-minute version largely removed the effect in the study.

How traders can protect themselves

1. Trade markets with depth. Check the full order book, not the top line. If your order is large relative to the resting size, you are the liquidity, and you are paying for the privilege.

1. Treat volume with skepticism. Volume can be inflated by wash trading, incentive programs or market-maker churn. Compare volume with open interest and with the depth available at the touch. A market with huge volume and tiny depth is not deep.

1. Do not chase sudden moves in thin books. A five-cent jump on small volume may be a real information event or a painted print. Wait for confirmation from a second source or a deeper market.

1. Use limit orders and size to the book. A market order in a thin market invites slippage.

1. Check the resolution rules for manipulable sources. If the contract settles on a price or a statistic that a trader can influence, demand a larger edge or avoid it.

1. Diversify across independent events. Correlated positions behave like one large bet, which is the position a manipulator wants to exploit.

1. Report suspicious activity. The CFTC accepts tips and complaints, and platforms have surveillance teams. A report is not proof, but it is the channel that creates a record.

The limits of the evidence

Wash-trading detection is statistical. A cluster of accounts that trade with each other can be a market-making operation, a group of friends, or a legitimate arbitrage strategy. False positives are possible. Enforcement is slow, identities are often pseudonymous, and the legal line between aggressive trading and manipulation can be genuinely unclear.

That is why the responsible conclusion is narrow: thin markets are easier to manipulate, the research shows measurable effects, and enforcement cases exist. It does not follow that every thin market is rigged or that every large trader is a manipulator. The trader's job is to price the risk, not to solve the case.