How 58% of Polymarket Users Lose Money (And How DuelDuck's Model Is Different)

What Percentage of Polymarket Users Actually Make Money?

A Bloomberg News analysis of every Polymarket wallet active since January 2025 found that more than 100,000 accounts lost at least $1,000 - nearly double the number that made the same amount. A separate study by Andrey Sergeenkov found that 84.1% of Polymarket's 2.5 million traders are not profitable. A London Business School and Yale working paper found that only 3% of traders drive price discovery on Polymarket, while the rest lose money around them. The top 5% of accounts - largely automated bots - account for 75% of all trading volume. DuelDuck's creator model is structurally different: creators earn up to 5% net creator fee on every pool regardless of whether they win or lose the underlying prediction.

The Data: Four Studies, One Conclusion

Four independent analyses were published or circulated in April 2026, each using different methodologies and data sources, all arriving at the same conclusion: retail participants on Polymarket are structurally disadvantaged against a small group of informed, often automated, high-volume traders.

Why Retail Traders Lose: The Three Structural Disadvantages

Disadvantage 1: Bots Enter First at Better Prices

Bloomberg's analysis found the typical bot averaged 89 trades on each active trading day versus 2.2 for non-bots. Bots operate at millisecond speed, executing the moment new information (news releases, data publications, social media signals) enters public channels. By the time a human trader reads an article, processes the implications, and enters a position, bots have already moved the market price.

A retail trader who correctly identifies that WTI oil will hit $110 before month-end may be right - but if the bot bought the contract at $0.30 (30% probability) and the retail trader enters at $0.55 (after bots have moved the price), the retail trader's correct prediction produces a much smaller return than the bot's identical conviction. Being right is not enough. Timing and entry price determine profitability.

Disadvantage 2: Retail Traders Trade at Extreme Odds

The University of Toronto/HEC Montreal/ESSEC study found the bottom 95% of earners made 56% of their trades at prices below 10 cents or above 90 cents - the extreme ends of the probability spectrum. The top 0.1% of earners made only 28% of their trades at these extremes.

Trading at extreme odds creates a specific profitability trap: a 5-cent contract (5% probability) must be bought 19 times and win once just to break even. Retail traders who buy long-shot contracts consistently overpay for lottery-ticket-style positions. Professional traders and bots concentrate in the 20-80 cent range where market pricing is most efficient and edge can be extracted without requiring extreme outcomes.

Disadvantage 3: Information Asymmetry

The London Business School/Yale paper found only 3% of traders drive price discovery on Polymarket. These are participants with genuine information advantages: professional analysts who model events better than consensus, energy traders who understand Hormuz shipping data, music superfans who track chart algorithms, and - in some documented cases - participants with non-public information. These 3% move prices toward the correct outcome. Everyone else trades at prices already set by better-informed actors.

The Maduro insider trading case illustrated this in extreme form: three newly created Polymarket accounts bought contracts at 10% probability hours before Operation Absolute Resolve was announced. The accounts made over $630,000 collectively. The retail participants who sold them those contracts at 10% were pricing the event fairly based on public information - and lost to participants with classified knowledge. This is the information asymmetry problem in its most acute form.

The Wisdom of Crowds Myth: What the Data Actually Shows

Prediction markets are justified to regulators, media, and the public on the basis of the 'wisdom of crowds': the idea that collectively aggregating many participants' estimates produces more accurate forecasts than any individual expert. The academic literature from the Iowa Electronic Market supports this claim for well-structured markets. Polymarket's 2024 US election prediction (favoring Trump when polls showed a coin flip) is cited as the flagship example.

The London Business School/Yale paper challenges this narrative directly: prediction market accuracy on Polymarket comes from 3% of traders, not from the crowd. The '97%' are not contributing to price discovery - they are providing liquidity that the informed 3% trade against. The market produces accurate prices not because of crowd wisdom but because a small number of skilled (and sometimes insider-informed) participants move prices toward correct values.

Pablo Martineau, co-author of a related paper from University of Toronto, HEC Montreal, and ESSEC: 'You'd be surprised how many of them did not anticipate that you would see such concentrated gains — that so few make money.' His students who had been enthusiastically trading on prediction markets were 'taken aback' by the findings.

The Social Media vs. Reality Gap

Prediction markets are marketed on social media as a lucrative side hustle. A Polymarket referral program offers 30% of trading fees generated by referred participants - creating a wave of influencer-driven content showcasing large wins. The gap between social media presentation and data reality is significant:

Kalshi's user base is 24% under age 25, compared to just 7% for traditional sportsbooks like DraftKings. Young, first-time traders are the most represented demographic on a platform where 84% lose money. Research shows men aged 30 and under who wager on sports have a 10% rate of gambling problems. The prediction market demographic overlap with this risk group is significant.

The Three Paths to Profitable Prediction Market Participation

The data does not mean all prediction market participation is hopeless. It means the standard retail trading model - buy contracts, hope to win, repeat - is structurally disadvantaged. Three alternative models produce better outcomes:

Path 1: Develop Genuine Domain Expertise

Brandon Fean (music teacher) made $100,000 trading Ariana Grande charts. Caleb Davies made $389,000 on culture markets. Both had domain knowledge that general market participants lacked. Domain expertise converts to information edge that allows entry before the general market reprices. The London Business School paper confirms: skilled participants - those who beat the coin-flip benchmark - make money consistently. The problem is that most participants think they have domain expertise when they do not.

Path 2: Trade in Thin Markets Before Bots Dominate

Bot dominance is concentrated in high-volume markets (major election contracts, NBA finals, WTI oil). Niche markets - a specific Grammy category, a reality TV weekly elimination, a small-cap company KPI contract - have less bot activity because the volume does not justify algorithmic infrastructure. Participants with domain knowledge in niche categories can find edge that evaporates in high-volume markets.

Path 3: Create Duels Instead of Trading Them

DuelDuck's creator model offers a fundamentally different economic relationship with prediction markets: creators earn income regardless of whether their prediction is correct. A DuelDuck creator who designs a Grammy winner duel, distributes it to a music fan community, and earns a 5% net creator fee on a $1,000 pool has made $50 without taking any directional risk on the outcome. The creator income is not a trading profit - it is a platform fee for market design and distribution work.

DuelDuck's Creator Model: Income Without Directional Risk

The structural problem for retail Polymarket participants is that they must predict correctly and enter at good prices to make money. DuelDuck's creator model decouples income from prediction accuracy. Compare the two models:

The Creator Fee Math

A DuelDuck creator who builds a consistent weekly publishing schedule earns predictable income without directional prediction risk:

• 5 weekly duels at $500 average pool: $2,500 pool volume per week, $125 net creator fee per week, $6,500 per year from weekly music/sports/culture duels alone

• One major event duel (Grammy season, $2,000 pool): $100 net creator fee from a single event

• Annual prediction market calendar: Grammys, Oscars, NBA Finals, World Cup, PGA majors, US elections, major economic releases - 52+ weeks of continuous creator income opportunities

The comparison to Polymarket retail trading: a Polymarket participant who deposits $2,500 and trades weekly has an 84% probability of being unprofitable. A DuelDuck creator who publishes 5 duels per week at $500 average pool earns $125/week with zero directional exposure to prediction outcomes.

What to Do If You Are Already Losing on Polymarket

If the data describes your experience on Polymarket or Kalshi, four concrete adjustments change the math:

• Audit your trade timing, not your predictions. Track whether you enter positions before or after significant price moves. If you consistently enter after the market has already moved, you are experiencing the bot disadvantage. Shift toward niche markets where bot infrastructure is less competitive.

• Stop trading long-shot contracts. The bottom 95%'s tendency to trade below 10 cents and above 90 cents is the single most damaging behavior pattern identified in the data. Contracts below 10 cents require 9 correct wins to offset 1 loss. Focus on markets priced 20-80 cents where the probability distribution is most accurate.

• Build a creator income stream instead of or alongside trading. DuelDuck's creator model generates fee income regardless of prediction accuracy. For participants who enjoy the prediction market ecosystem but are not in the profitable 16%, creator income provides a sustainable alternative income stream that does not depend on beating bots at their own game.

• Identify your actual domain expertise. The profitable 3% of traders have genuine informational advantages. Brandon Fean had Ariana Grande HTML code. Caleb Davies had streaming momentum models. Energy traders have EIA data workflows. Before trading, identify what specifically you know that the general Polymarket market does not price correctly. If the answer is 'I follow the news' or 'I have good intuition,' the data suggests caution.

Conclusion: The Industry's Uncomfortable Data

Bloomberg published the wallet analysis on April 29, 2026 - the same week Polymarket was reportedly raising $400M at a $15B valuation. The juxtaposition is striking: the platform is worth $15 billion partly because 84% of its participants consistently lose money to the 16% who consistently win. This is not a bug - it is how financial markets are structured. The question is whether prediction market participants understand which group they are in.

The data from Bloomberg, Sergeenkov, London Business School, and the University of Toronto consortium is not evidence that prediction markets do not work. They work extremely well for price discovery - because the 3% of skilled, often automated, participants move prices toward correct values. The crowd provides the capital. The informed minority provides the accuracy.

For participants who want to engage with prediction markets sustainably, three options exist: develop genuine domain expertise that puts you in the profitable 3%; trade in niche markets where bots have less infrastructure; or build creator income on DuelDuck that decouples your earnings from prediction accuracy entirely.

The question is not whether prediction markets are profitable. The question is which side of the information divide you are on - and whether your model depends on being right or on being useful.

Start Predicting. Start Earning

DuelDuck - P2P prediction market on Solana. Create community duels, earn up to 10% creator fee on every pool, and build income that does not depend on beating bots at their own game.

Start earning as a creator today

Disclaimer: This article is for informational purposes only. Past performance of prediction market participants is not indicative of future results. Prediction market trading involves risk of total loss. Data cited is from independent research; individual results will vary.