Share
Polymarket crossed $100 billion in lifetime notional trading volume in June 2026, recording $6.38 billion in a single week during the FIFA World Cup. It has become the defining commercial success story of decentralized finance in 2026, a platform that built genuine price discovery infrastructure around real-world events and turned prediction markets from a niche academic concept into a product that institutional investors, retail gamblers, and Wall Street quants all use simultaneously.
None of that changes what Columbia Business School found when it ran network analysis on two years of Polymarket’s onchain transaction data. Roughly 25% of the platform’s historical volume came from wallets trading back and forth, often with themselves or with coordinated accounts, without changing their net exposure.
Wash trading peaked at nearly 60% of Polymarket’s weekly volume in December 2024, with some weeks in sports and election markets showing more than 90% of trades as likely inauthentic. By July 2026, Polymarket’s new fee structure, rolled out between January and March, was widely cited as the fix. The evidence suggests otherwise. Here’s why.
Columbia Business School researchers led by professor Yash Kanoria analyzed more than two years of onchain Polygon data across Polymarket’s 1.26 million wallets. Their paper, titled “Network-Based Detection of Wash Trading,” has been circulating on SSRN and has not yet been peer-reviewed, a caveat that matters when the 25% headline gets cited as a settled fact rather than a modeled estimate. The methodology is specific enough to warrant its own explanation because it determines what the study can and cannot prove.
The team built a graph of wallet interactions, connecting wallets that traded with each other, and used network clustering to identify groups of wallets that exhibited suspicious coordination patterns. The signature they were looking for was wallets cycling trades between each other at high frequency with no directional exposure change, the onchain fingerprint of an account trying to generate volume without taking market risk. 14% of Polymarket’s wallet population, about 176,000 accounts, was flagged as exhibiting activity consistent with wash trading. One cluster alone, more than 43,000 wallets, generated close to $1 million in combined volume, mostly at sub-cent prices, almost all of it flagged.
The distortion was not static:
The category breakdown matters because it maps almost perfectly onto which market types were most likely to contribute to an eventual token airdrop allocation.
The researchers stopped short of accusing Polymarket of orchestrating the activity. They pointed instead to three structural conditions that made it viable: no identity verification requirements, no trading fees for most of the platform’s history, and speculation around a future POLY token as the plausible driver. That third condition is the one that explains the timing pattern. Peaks in wash trading correlated more closely with rumors about token issuance than with news about the events being traded.
In October 2025, Polymarket chief marketing officer Matthew Modabber publicly confirmed that a POLY token and an airdrop were coming, without disclosing a snapshot date, an allocation formula, or eligibility criteria. That vacuum is what converted volume inflation into a rational strategy.
⚡️ @Polymarket's CMO confirms $POLY token and airdrop
“There will be a token, there will be an airdrop,” @MatthewModabber said during an interview on the Degenz Live podcast. Polymarket is currently focused on its U.S. app launch and eyes between $12B to $15B valuation in new… pic.twitter.com/t23B5UWv26
— CryptoDep (@Crypto_Dep) October 24, 2025
If a future distribution rewards trading activity and no one knows the cutoff or the filter, manufacturing volume costs next to nothing relative to a potential five- or six-figure airdrop payout. The fee structure introduced in 2026 adds a 0.75%-1.80% marginal cost on taker orders. Against a potential airdrop allocation, that cost is a rounding error.
Active sports market-makers who spoke to researchers and analysts noted that the crude wash trading patterns visible in 2024, large block trades moving directly between two connected accounts, have largely disappeared. The activity has not stopped; rather, it has evolved.
Operators now spread coordinated trades across more than 100 wallets to evade wallet-concentration detection, a more sophisticated version of the same strategy that is harder to flag statistically because it produces resting order activity that looks identical to genuine liquidity provision at the individual wallet level. Crucially, this shift began well before Polymarket introduced trading fees in March, suggesting the change was driven by anticipated POLY token criteria rather than the fee structure itself.
Polymarket introduced trading fees in three phases rather than all at once. Crypto markets received taker fees in January 2026, followed by sports on Feb. 18.
With the rollout of fees across most @Polymarket markets on the 30th, current estimates put them flipping @Pumpfun's revenue by summer. https://t.co/aX9br9mTMG pic.twitter.com/Ahktw8WgWy
— Pine Analytics (@PineAnalytics) March 25, 2026
On March 30, the schedule expanded to finance, politics, economics, culture, weather, tech, mentions, and a catch-all “other” category, leaving only geopolitics markets fee-free.
By the time the rollout was complete, however, the most sophisticated wash trading strategies had already evolved, limiting the fees’ effectiveness as a deterrent.
The formula is symmetric and probability-weighted: The fee equals shares traded, multiplied by share price, multiplied by a category-specific rate, multiplied by price times one minus price. That last term is the variance of a Bernoulli distribution, so it peaks at a 50% probability and falls toward zero at the extremes. Peak effective rates run from 0.75% on sports to 1.56% on mentions to 1.80% on crypto.
At high-confidence prices near 90 cents, effective fees compress to a fraction of those headline figures, which is relevant because the sub-cent price activity that Columbia’s cluster of 43,000 wallets generated would face near-zero fees even under the new schedule.
Makers, meaning resting limit orders, pay nothing and collect a rebate funded by taker fees, with the rebate share running 20%-25% by category. Reuters reported in June that analysts projected the schedule would generate $800,000-$1 million in daily revenue based on trailing volume, roughly $300 million annualized.
The Washington Post described the change as increased protection against wash trading rather than a simple fee hike, positioning the redirect toward makers as the mechanism doing the work. That framing is the one that needs examining.
The textbook wash trading technique involves two colluding wallets crossing a trade with each other. Under a maker-taker fee model, if both sides post resting limit orders that happen to match, both sides are makers. Neither pays a fee. The fee only applies when one side crosses the spread and immediately takes liquidity.
A wash trader with patience does not need to take anything. Two of their own limit orders simply need to cross, which costs nothing under the current fee schedule and may generate a rebate through the maker rewards program.
The March fee structure taxes the least sophisticated version of wash trading, rapid market-order round trips between accounts, while leaving the maker-maker version structurally untouched. The maker-maker version is also the harder one to flag statistically because resting order activity looks identical to genuine liquidity provision unless you can trace wallet clustering at the depth the Columbia team applied.
Two additional structural limits compound this. The fee schedule applies only to markets deployed on or after each category’s activation date. Pre-existing markets, including those that carried the heaviest wash-trading load through 2024 and into early 2025, sit entirely outside the new rules. The 43,000-wallet cluster Columbia documented was largely active before any fees existed. The rules that now apply would not have applied to the trades the study analyzed.
No public data set yet breaks out wash-trading share specifically for the post-March-30 period by category. That analysis is the test that would separate the fee’s actual deterrent effect from the airdrop-farming adaptation already underway beforehand. Without it, the claim that the fee structure reduced wash trading is an assertion rather than a measurement.
Polymarket has not gone on record about maker-maker collusion as a detection gap. No outlet covering this story appears to have put that question to the company directly. The platform announced after the Wall Street Journal’s fake-win influencer investigation that it would conduct a comprehensive audit of promotional content. Whether that audit extends to wash-trading detection methodology, and whether it produces public findings, has not been confirmed.
The POLY token eligibility criteria, whenever published, will be the real signal. If wash-flagged wallets are excluded retroactively from any airdrop distribution, that exclusion does more deterrent work than any fee curve Polymarket could design. A 1.80% taker fee raises the cost of farming.
A retroactive exclusion that strips volume-weighted eligibility from flagged accounts eliminates the incentive entirely. Until that exclusion is confirmed and its detection methodology made public, sophisticated wash traders are operating under a reasonable assumption that the airdrop criteria remain achievable at a 0.75%-1.80% marginal cost.
Polymarket generated about $300 million in annualized revenue under the new schedule. It crossed $100 billion in lifetime notional trading in June. It is the largest prediction market in the world by volume and has produced genuinely accurate price signals on major political, sporting, and economic events across its history. The platform’s commercial success and its integrity gap are not competing narratives. They are the same story.
Prediction markets derive their value from the assumption that prices reflect genuine information aggregation. If 25% of historical volume, and potentially a meaningful share of current volume, comes from wallets’ manufacturing activity rather than expressing views, the price signal degrades. A platform where 45% of sports volume was likely wash trading at its peak was not producing a clean sports prediction market. It was producing a market contaminated by noise at roughly the same magnitude as its signal.
The fee structure addressed the revenue gap. It has not yet addressed the detection gap, the pre-existing-market gap, or the airdrop incentive that the Columbia study identified as the primary driver.
Three things would meaningfully change the integrity picture: a public post-March breakdown of wash-trading share by category, a public statement from Polymarket on maker-maker collusion as a detection gap, and confirmed retroactive exclusion of flagged wallets from POLY token eligibility. None of the three exists as of July 14, 2026.
Share
preferred on
