How Concentrated Is Prediction-Market Profit?
One question, one answer per study. This page tracks whether profit is clustered among a small cohort of accounts — not what percent of individual users make money.
What this page is NOT
Headline stat by study
Rows stay pending until primary URLs and verbatim quotes are available.
0.11% flagged skilled
1,950 of 1.72M accounts flagged as skilled-trader accounts
Account-level concentration of skilled trading and profit capture, pending primary-source quote verification.
The study is treated here as a concentration-axis source, not a per-user profitability source, until primary-source quotes are verified.
~70% lose (realized)
Fewer than 0.04% of addresses captured over 70% of realized profits
Wallet-level realized-profit concentration across closed positions; primary Dune dashboard re-verification pending.
Fewer than 0.04% of addresses captured over 70% of realized profits, on the order of $3.7B in cumulative realized gains.
~84% lose (lifetime)
About 0.033% of traders earned more than $100,000 cumulative
Wallet-level lifetime-profit concentration threshold; researcher-source re-verification pending.
Roughly 2% of traders earned more than $1,000 cumulative, and about 0.033% earned more than $100,000.
~99% below wage
Fewer than 1% clear a monthly wage-equivalent threshold
Monthly income-threshold concentration, not lifetime profitability.
This is the stat that gets flattened to '99% are losers.' It is an income-replacement threshold, not a lifetime profitability outcome.
What each study measures — and what it does not
London Business School / Yale authors
Sample size: 1.72M accounts
Source type: SSRN paper / author pages
Account-level skilled-trader concentration and disproportionate realized-profit capture, pending primary-source quote verification.
- Per-user profitability or win rate — the axis is concentration among accounts, not the share of accounts that made money.
- Named identities behind flagged accounts — the page does not speculate about who controls any account.
- Financial advice, policy recommendations, or proof that retail traders cannot win.
DefiOasis (on-chain analyst)
Sample size: Roughly 1.7M trading addresses
Source type: Dune Analytics dashboard
Realized profit and loss on closed positions: sale proceeds plus market-resolution redemptions, minus purchase costs. Open positions are excluded.
- Unrealized profit or loss still sitting in open positions.
- Per-person profitability — one trader can control many wallets, and many wallets can be run by one desk.
- A rolling window — the figure covers all Polymarket history through the cutoff, not a single year.
Andrey Sergeenkov (independent researcher)
Sample size: Roughly 2.5M wallets
Source type: Researcher's own published analysis
Cumulative lifetime profit and loss per wallet across all settled and open exposure.
- Realized-only profit — this figure is lifetime cumulative and includes open exposure.
- Experience level — the denominator has been pulled toward newer wallets since the late-2024 election spike.
- Per-person profitability at the human level — wallet count is not user count.
Andrey Sergeenkov
Sample size: Roughly 2.5M wallets (same sample as the lifetime cut)
Source type: Same study as sergeenkov-2026-04
Share of wallets earning $5,000 or more in a given calendar month.
- Whether a trader is profitable at all — a wallet clearing $500 in a month still sits in the 99%.
- Lifetime cumulative profit — it is a monthly income cut.
- Hobbyist versus full-time intent — most retail traders never intended to replace a wage.
Why concentration is not the same as user-level profitability
Concentration metrics describe the shape of the profit distribution: whether a small cohort of accounts captures a disproportionate share of realized gains. That is structurally different from asking what percentage of users or wallets are profitable. One trader can run many wallets, one desk can control many accounts, and one account can represent tooling or capital that ordinary users do not have.
Skill-based concentration therefore says more about clustering, market-making, speed, and repeated participation than it says about the average trader's win rate. A market can have some profitable retail users and still have a fat-tailed distribution where the largest profits sit in a tiny top cohort. This page keeps those axes separate so concentration evidence does not get flattened into a claim about winners, losers, or the wisdom of crowds.
What this means for retail traders
- Treat concentration as a distribution-shape signal: a small cohort can capture outsized profit without answering how many ordinary users win or lose.
- Separate wallet/account counts from human users; one trader can operate many wallets, and one desk can create many account-level traces.
- Assume the top tail often has infrastructure advantages — speed, capital, market-making tooling, or primary-source workflows — rather than just better guesses.
- Use concentration evidence as context for risk expectations, not as a trading recommendation or policy conclusion.
FAQ
Is this page saying most prediction-market users lose money?
What does the LBS/Yale study measure here?
Does a concentrated profit distribution mean markets are rigged?
Why not combine this with the percentage-of-users-who-profit page?
Are the 1,950 accounts named?
Can retail traders use this as advice?
Understand the edge
Can retail traders win?
The retail-survivability question adjacent to concentration evidence.
Why bots lose money on execution
Execution, slippage, and live-market mechanics that can shape edge quality.
Is this real arbitrage?
How to separate durable edge from false-arbitrage or friction-driven signals.
Platform architecture map
How exchanges, wrappers, routers, perps rails, and intelligence layers fit together.
Polymarket troubleshooting hub
Practical platform-mechanics context for user-facing problems.