# Prediction Market Wallet Concentration and the Wisdom of the Crowd > An August 2026 briefing found the top 1% of Polymarket wallets produced 68% of congressional-market volume. That matches platform-wide data and fits marginal-trader theory, so it does not prove prices are wrong. Our view is that it makes each price depend on informed capital trading against the biggest wallet, which most thin midterm markets lacked. Published 2026-11-10 by PM Crypto Notes. Canonical: https://pmcrypto.blog/wallet-concentration/ ## Key points - ACDC found the top 1% of Polymarket Global wallets produced 68% of 2026 congressional-market volume, with data through August 10, 2026. - 80% of Polymarket congressional markets had fewer than 100 wallets, and only 10 of 7,466 markets reached 1,000 participants. - Platform-wide, Allium found 1% of wallets produced 65% of notional volume, so the midterm figure is the norm. - A 2026 working paper classed 3.14% of Polymarket accounts as skilled; skilled traders and market makers took over 30% of gains. - One midterm wallet held 43% of South Carolina Senate dollars and 32% of Kansas dollars on the Republican side; Republicans lost both. Our thesis is that the 1% of wallets producing about two-thirds of prediction-market volume does not by itself make prices wrong, but it means a price is only as good as the informed capital willing to trade against the largest wallet in that market, and in most 2026 congressional markets that capital was missing. ## What did the August 2026 briefing find? The [Anti-Corruption Data Collective](https://acdatacollective.org/?p=1662) (ACDC), a nonprofit research group, published its first "Gambling on Democracy" briefing on August 17, 2026. It examined 7,466 markets on 2026 House and Senate races across Kalshi, Polymarket and Polymarket US, with data through August 10. Its headline number: on Polymarket Global, the top 1% of wallets accounted for 68% of trading volume on congressional markets. Ten wallets alone produced 17% of volume and had traded in 426 of the 470 seats on the ballot. The participation figures are starker. Only 39,820 Polymarket wallets had bet on 2026 congressional races. Eighty percent of Polymarket's congressional markets had fewer than 100 participating wallets, and only 10 had reached 1,000. ACDC classed 87% of markets as either low-volume (under $10,000 wagered) or high-volume but held by very few traders. Michelle Kendler-Kretsch, one of the authors, said: "That combination makes prediction markets hard to trust as a real signal." [CryptoSlate's write-up](https://cryptoslate.com/just-1-of-wallets-control-133m-in-midterm-odds-on-polymarket-creating-a-dangerous-illusion-of-mass-public-consensus/) on August 22 added the necessary caveat: "the 68% number alone can't show that 2026 odds are wrong or manipulated." Wallets are pseudonymous, one wallet may not be one person, and a large wallet may be hedging or providing liquidity with no political view at all. ## Is 68% unusual for prediction markets? It is close to the platform-wide norm. [Allium Research](https://www.allium.so/reports/prediction-markets-a-new-financial-primitive), in a report published March 31, 2026 covering more than $90 billion of notional volume on Polymarket and Kalshi, found that "65% of all notional volume is generated by just 1% of wallets." A Bloomberg analysis of every Polymarket wallet active since the start of 2025, [summarized by PYMNTS](https://pymnts.com/markets/2026/most-polymarket-users-lose-money-as-top-accounts-win-big) on April 28, 2026, found that 5% of wallets produced 75% of volume. The midterm markets looked the same up close. [Bitquery's on-chain study](https://bitquery.io/investigations/polymarket-midterms-odds-2026) of Polymarket's Polygon trades, published November 1, 2026 with data through October 30, counted 437 wallets that had each traded at least $100,000. Together they supplied about 60% of midterm-market money. Wallets that traded under $100 were about 60% of all wallets and under 1% of the money. So the congressional figure is the ordinary shape of these venues, applied to markets that were mostly small. ## Was the crowd ever supposed to set the price? The academic case for prediction markets never rested on the average trader. In their 1992 study of the Iowa Political Stock Market, [Forsythe, Nelson, Neumann and Wright](https://ideas.repec.org/a/aea/aecrev/v82y1992i5p1142-61.html) found that traders showed substantial judgment biases, yet the market beat the polls on the 1988 presidential election. Their explanation was that bias describes average behavior, while prices are set by marginal traders who keep quoting near the market price. The marginal-trader hypothesis predicts concentration. The most detailed recent test supports it. A working paper by Roberto Gómez-Cram, Yunhan Guo, Theis Ingerslev Jensen and Howard Kung (London Business School and Yale), revised April 25, 2026 and [reported by The Block](https://www.theblock.co/amp/post/398902/skilled-polymarket-traders-are-a-3-minority-and-everyone-else-funds-their-gains-study), analyzed every Polymarket trade from 2023 through 2025: 1.72 million accounts and about $13.76 billion of volume. Using a sign-randomization test that reran each trader's history 10,000 times with trade direction flipped at random, it classed only 3.14% of accounts as skilled winners whose order flow predicted both short-term price moves and final outcomes. Skilled traders and market makers together were under 3.5% of accounts and captured more than 30% of gains. The 67% of accounts classed as unlucky or unskilled losers absorbed all aggregate losses. Skill also persisted: 44% of accounts classed as skilled in one sample stayed skilled in a held-out sample. The authors wrote, [as quoted by Cointelegraph](https://cointelegraph.com/news/prediction-market-showcase-widsom-of-an-informed-minority-not-the-crowd-study), that the majority "does not produce accuracy; rather, it funds it." Our read is that concentration is the mechanism working as designed. What matters is who the concentrated wallets are: market makers quoting both sides, skilled takers correcting prices, or a motivated buyer with a side to defend. ## What does a share of volume measure? Volume concentration is a noisy proxy for price-setting power, for three reasons. First, market makers generate large volume without a directional view. The Gómez-Cram paper puts them inside the informed minority, and Cointelegraph reported that market-maker accounts earned about $11,830 each on average. Second, some volume is not trading in any economic sense. A Columbia study led by Allen Sirolly, [covered by Decrypt](https://decrypt.co/347842/columbia-study-25-polymarket-volume-wash-trading) on November 7, 2025, estimated that about 25% of Polymarket volume over three years was likely wash trading, peaking near 60% of weekly volume in December 2024. [CNBC reported](https://www.cnbc.com/2026/09/30/kalshi-polymarket-trading-volume-scrutiny.html) on September 30, 2026 that the same research put wash-like trading at a negligible level by April 2026, so this matters more for older data than for the midterm cycle. Third, volume piles up in odd places. CNBC found about $158 million traded on Egypt winning the World Cup, a contract never priced above 0.5%, against about $152 million on Spain's odds. Polymarket attributed such patterns to "sharps" correcting mispricings. Whatever the cause, dollars traded are a poor measure of dollars at risk on a belief. ## When does one large wallet move a price? The record has both kinds of whale. In Intrade's 2012 presidential market, Rajiv Sethi [wrote](https://rajivsethi.substack.com/p/the-romney-whale-13-09-26) that "a single trader was responsible for about a third of all bets on Romney to win" in the final two weeks, losing close to $4 million, with large orders that "effectively placed a floor on the Romney price and a ceiling on the price for Obama." The [published paper](https://www.ubplj.org/index.php/jpm/article/view/1179) by David Rothschild and Sethi (Journal of Prediction Markets, 2016) presents evidence that this trader may have manipulated the market. In 2024 the large wallet was the informed one. The pseudonymous trader "Théo" told [The Free Press](https://www.thefp.com/p/french-whale-makes-85-million-on-polymarket-trump-win) that he commissioned neighbor-style polling in Pennsylvania, Michigan and Wisconsin and then raised his stake on Trump to $80 million across 11 accounts. The outlet reported he made $85 million. Concentration in that case carried private research into the price. The Gómez-Cram paper describes a third type: 1,950 accounts that opened shortly before a single event and went dormant after it resolved. These insider-like accounts moved prices roughly 7 to 12 times more per dollar than skilled traders, but their activity was too isolated to drive overall accuracy. ## What did the 2026 midterms show? Bitquery's data gives a clean case of one wallet dominating single races. A wallet labeled lomneivan held Republican positions across several Senate contests and accounted for 43% of traded dollars in South Carolina and 32% in Kansas. For October 23 to 29, Polymarket priced a non-Republican win at 24% in South Carolina and 42% in Kansas, each up 10 points from early October, with about $722,000 traded and roughly 1,350 wallets in each race. The two races then split. [CryptoTicker reported](https://cryptoticker.io/en/polymarket-midterms-2026-senate-odds/) that Adam Hamilton, the Democrat in Kansas, was priced at 64.5% an hour before the first polls closed on November 3, and he won. In South Carolina, Democrat Annie Andrews was at 30.5% and won, the only miss among the favorites in nine Senate races called by early November 4. We cannot say the whale caused the South Carolina miss, and a 30% outcome should happen about three times in ten. What the pair shows, in our reading, is the condition that matters. In both races, opposing money moved the price 10 points against a wallet holding a third or more of the dollars, which is the self-correction the theory promises. In Kansas it went all the way. In South Carolina, with the same dollar volume, a single balance sheet held 43% of it, and the price on election night still sat on that wallet's side. In a market shaped like that, the price says a great deal about one trader's position and much less about the electorate. ## What is the strongest case against our view? The strongest counter-argument is that a mispriced market is a standing invitation. If a whale props up a price, anyone with better information profits by selling to it, and the finding that skill persists suggests such traders exist and come back. On this view concentration corrects itself, and the midterm data above partly supports it. We accept this for deep markets. It weakens where the reward is too small to attract the correctors. ACDC found that most congressional markets had fewer than 100 wallets and that 87% were either under $10,000 in volume or dominated by a few traders. On our rough arithmetic, correcting a 10-point error in a market with $10,000 of total volume earns at most around $1,000 before fees, which rarely pays for original research on a House primary. Marginal-trader theory needs the marginal trader to show up, and across 7,466 markets there are not enough of them to go around. ## What would change our mind - A market-level study showing that 2026 markets with fewer than 100 wallets were as well calibrated as deep markets. - Evidence that the top 1% of wallets by volume are mostly two-sided market makers, with directional large wallets holding a small share of open interest. - Calibration data showing that races where one wallet held more than a third of traded dollars resolved in line with their prices at the expected rate. - A replication of the Gómez-Cram skill classification on Kalshi or Polymarket US showing a much larger informed minority on regulated venues. ## FAQ ### What percentage of prediction market volume comes from the top 1% of wallets? The Anti-Corruption Data Collective found the top 1% of Polymarket Global wallets produced 68% of volume on 2026 congressional markets, with data through August 10, 2026. Allium Research found a similar 65% across more than $90 billion of Polymarket and Kalshi volume. ### Does wallet concentration mean prediction market odds are manipulated? Not by itself. Large wallets include market makers and skilled traders who correct prices, and one wallet is not necessarily one person. Concentration raises manipulation risk mainly in thin markets where few traders can trade against a large position. ### What is the marginal trader hypothesis? It is the idea, from a 1992 study of the Iowa Electronic Markets, that prices are set by a minority of unbiased traders quoting near the market price, so the average trader's biases do not decide accuracy. It predicts that a small group of wallets will dominate trading. ### Who makes money on Polymarket? A 2026 working paper by Gómez-Cram, Guo, Jensen and Kung classified 3.14% of Polymarket accounts as skilled. Skilled traders and market makers, under 3.5% of accounts, took more than 30% of gains, while 67% of accounts absorbed all aggregate losses. ### How many wallets bet on the 2026 midterms on Polymarket? ACDC counted 39,820 Polymarket wallets that had bet on 2026 congressional races as of August 10, 2026. Bitquery later counted 437 wallets that each traded at least $100,000 and together supplied about 60% of midterm-market money. ## Sources - [A Handful of Bettors Are Setting the Odds on the 2026 Midterms, New Analysis Finds](https://acdatacollective.org/?p=1662), Anti-Corruption Data Collective, 2026-08-17 - [Just 1% of wallets control $133M in midterm odds on Polymarket, creating a dangerous illusion of mass public consensus](https://cryptoslate.com/just-1-of-wallets-control-133m-in-midterm-odds-on-polymarket-creating-a-dangerous-illusion-of-mass-public-consensus/), CryptoSlate, 2026-08-22 - [Prediction Markets: A New Financial Primitive](https://www.allium.so/reports/prediction-markets-a-new-financial-primitive), Allium Research, 2026-03-31 - [Most Polymarket Users Lose Money as Top Accounts Win Big](https://pymnts.com/markets/2026/most-polymarket-users-lose-money-as-top-accounts-win-big), PYMNTS, 2026-04-28 - [Polymarket Midterms 2026: Can Democrats Flip the Senate?](https://bitquery.io/investigations/polymarket-midterms-odds-2026), Bitquery, 2026-11-01 - [Anatomy of an Experimental Political Stock Market (American Economic Review, 1992)](https://ideas.repec.org/a/aea/aecrev/v82y1992i5p1142-61.html), IDEAS/RePEc - [Skilled Polymarket traders are a 3% minority, and everyone else funds their gains: study](https://www.theblock.co/amp/post/398902/skilled-polymarket-traders-are-a-3-minority-and-everyone-else-funds-their-gains-study), The Block, 2026-04-26 - [Prediction markets reflect 'wisdom of an informed minority,' not crowd: Study](https://cointelegraph.com/news/prediction-market-showcase-widsom-of-an-informed-minority-not-the-crowd-study), Cointelegraph, 2026-04-27 - [A Quarter of Polymarket Volume May Be Wash Trading, Columbia Study Finds](https://decrypt.co/347842/columbia-study-25-polymarket-volume-wash-trading), Decrypt, 2025-11-07 - [Kalshi, Polymarket trading volumes on some products raise questions amid massive growth](https://www.cnbc.com/2026/09/30/kalshi-polymarket-trading-volume-scrutiny.html), CNBC, 2026-09-30 - [The Romney Whale](https://rajivsethi.substack.com/p/the-romney-whale-13-09-26), Rajiv Sethi, 2013-09-26 - [Trading Strategies and Market Microstructure: Evidence from a Prediction Market](https://www.ubplj.org/index.php/jpm/article/view/1179), The Journal of Prediction Markets, 2016-09-20 - [French whale makes $85 million on Polymarket Trump win](https://www.thefp.com/p/french-whale-makes-85-million-on-polymarket-trump-win), The Free Press, 2024-11-26 - [Polymarket Midterms: The Odds Against the Results](https://cryptoticker.io/en/polymarket-midterms-2026-senate-odds/), CryptoTicker, 2026-11-04