CrowdIntelFour wallets with no names, funded 58 seconds apart, are up $618K.
A single on-chain address seeded four accounts inside one minute. None ever chose a username. Together they beat the entry-price odds across thousands of markets.
The funding
On 2025-09-24, a single on-chain address — 0x5487…a014 — created five Polymarket proxy wallets. Four of them were funded inside 58 seconds, between 13:26:53 and 13:27:51 UTC. The funding is on-chain and independently checkable.
0x5487559b → 0x33615155 · $200 USDC · block 76835231 · status 0x1 · verified via public RPC 2026-07-10
The wallets
Polymarket’s interface shows these as four unrelated users. On-chain, they are one wallet’s children, born seconds apart, running the same strategy in the same markets.
| Wallet | Funded (UTC) | Resolved n | Win rate | PM lifetime PnL |
|---|---|---|---|---|
| 0x3361…4363 | 13:26:53 | 2,147 | 62.1% | +$310,346 |
| 0x7846…c08d | 13:27:23 | 732 | 79.4% | +$118,088 |
| 0xa7ec…79fb | 13:27:39 | 431 | 72.6% | +$100,713 |
| 0x212a…83cf | 13:27:51 | 567 | 76.7% | +$88,697 |
None of the four ever chose a username — all display as raw 0x addresses.
The behavior
These are pure buy-side bots. Each account placed between 60,000 and 107,000 buy fills and made ≤6 sells ever — one side per market, held to resolution. Together they purchased $22–43M gross across 10–16K markets each. The concentration is in crypto micro-markets — the 15-minute “Bitcoin/Solana Up or Down” series — plus a long tail of thin markets.
Timeline
- 2025-09-24 · 13:26:53 UTC
The funder seeds the first of four wallets.
- 13:27:51 UTC
The fourth wallet is funded — the whole ring is live in 58 seconds.
- then
First bets placed: one side per market, across thousands of markets.
- 2026-07-10
Snapshot — +$617,844 combined, and still not one username between them.
How we computed this
The measure, the multiple-testing correction, and what we do not claim.
Skill vs entry odds, never raw win rate
For each wallet, over resolved positions, expected wins = Σ(entry price) and variance = Σ p(1−p). The z-score is (actual wins − expected) ÷ √variance. Entry price is the market’s own probability at purchase, so buying 95¢ favorites earns no credit — the null already expects those to win. A raw win rate is meaningless on its own.
Multiple-testing honesty
We scanned 102,263 wallets meeting the filters (≥20 resolved positions, ≥$10K staked, entry price 2–98¢). At z ≥ 3 chance alone would produce ~138 flags; we found 740. The Bonferroni-corrected line for this scan is z ≥ 5 — every headline account here clears it.
The dollars are Polymarket’s own
Every dollar figure comes from Polymarket’s public APIs (leaderboard profit + activity redemptions). Our internal ledger is used only to find wallets and compute the odds-based statistics, never to quote profit. Named per-market claims are re-derived from Polymarket’s activity records — buys, sells, and redemptions — so an account that scalped out early is not credited with a jackpot it never held.
What we do not claim
Addresses are pseudonymous; usernames are quoted as displayed. We assert no identity, no intent, and no wrongdoing — trade data cannot prove those. The words used are statistically extreme, coordinated, and anomalous. A shared funder plus an improbable record is evidence worth examining, not a verdict.