On September 18, 2026, an account displayed as Papeasy led Polymarket’s monthly profit ranking with $6.13 million and $0 of trading volume. The ranked wallet was collecting contracts bought by other wallets, redeeming them and returning nearly all the cash to the suppliers. Its monthly figure closely followed gross payouts, while the contracts’ purchase costs sat elsewhere in the connected network. Papeasy records
This was one of five cases documented between May and September in which public rankings diverged from profit earned by the ranked wallet during the stated period. Other cases placed lifetime earnings into a daily result, brought old accounting amounts into recent rankings or credited gains unsupported by reconstructed trading. A separate group appeared with large, fixed values whose origin remains unresolved. The cases share a measurement problem; they do not establish one common technical cause.
A payout becomes profit only after costs
A winning prediction contract can pay $1, but that dollar is not all profit. In a hypothetical example, a trader pays 90 cents for a contract and receives $1 when it wins. The payout is $1; the gain is 10 cents before fees and other costs.
Wallet boundaries matter too. If wallet A buys the contract and wallet B receives and redeems it, the purchase cost can remain on A’s record while the payout appears on B’s. Looking at B alone can make a payout collector look like a highly profitable trader. Understanding the connected operation requires following both the contracts and their costs.
The same dollar can describe a payout or a much smaller gain
Hypothetical one-contract example · before fees and other costs
| Measure | Amount | Meaning |
|---|---|---|
| Purchase cost | $0.90 | Money spent to acquire the contract. |
| Winning payout | $1.00 | Cash received when the contract wins. |
| Economic gain | $0.10 | Payout less purchase cost. |
This illustration explains the accounting distinction. It is not a measured return for Papeasy or any other account.
The time window creates a second distinction. A day’s profit needs a starting value and an ending value on the same accounting basis. If lifetime profit stays at $4 million throughout the day, the day’s gain is zero. Treating a missing starting value as zero can instead make the entire $4 million look newly earned. An old accounting difference can produce a similar ranking effect when it is inserted into a recent period.
Papeasy’s monthly figure closely followed payouts
The September 18 leaderboard showed $6,127,169.04 in monthly profit for the collecting address. The records examined over an 18-day transfer window showed 194 wallets supplying it with prediction contracts. Direct supplier transfers accounted for 99.87% of incoming contract transfers; the collecting wallet generally pulled the contracts using permissions the suppliers had granted. Contract transfers and ranking captures
The collector received about 7.4 million shares and approximately $3.99 million in payouts during that window. The roughly 54 cents paid per share received is a payout ratio, not a profit margin: the suppliers had bought both winning and losing positions, and their purchase costs belong in the economic calculation.
In a separate 30-day window ending September 18, the collector converted about $6.00 million of payouts into Polymarket’s dollar-denominated trading token and sent $5.93 million onward. Of the outgoing amount, 99.98% went to 145 wallets that had supplied contracts. Each of the 15 largest suppliers had received its first funding, exactly $200, from the collector within a 48-second interval on April 24. The transfers, funding and return payments support a connected operation. They do not identify its operator. Cash distribution and funding records
Purchase costs and payouts appeared on different wallets
The observed roles in the connected network
| Stage | What the records showed |
|---|---|
| Suppliers bought contracts | The supplier wallets carried the acquisition costs. |
| Contracts moved to the collector | The collector received contracts directly rather than buying them through the exchange. |
| The collector redeemed contracts | Payouts appeared on the ranked address. |
| Cash went back to suppliers | 99.98% of outgoing cash in the 30-day window went to supplier wallets. |
The 194 suppliers belong to the 18-day contract-transfer window. The 145 cash recipients and $5.93 million belong to the separate 30-day window.
Transfer and cash records ↗Sampled gains were much smaller than gross receipts
The sampled return estimate combines purchase costs from six supplier-wallet days with a win rate measured from traced contract transfers. On that basis, each $1 spent would return about $1.09 to $1.13: a gain of 9 to 13 cents per dollar of cost, or approximately 8% to 12% of receipts. These are estimated returns, not six directly observed cash returns. If the modeled returns generalized to about $6 million of payouts, the implied gain would be roughly $500,000 to $700,000. That is a conditional illustration, not a reconstruction of the operation’s total profit. Sample calculations and their limits
October 21 update: two differences from September
Papeasy’s collector still receives contracts from supplier wallets, redeems them and returns cash. The October check found two differences from the September observations.
Direct buying appeared. In the September 21–October 21 records examined, the collector’s own purchases first appear on October 10.
The monthly ranking no longer 100% matched the earlier payout fit. The $1.40 million figure captured on October 21 was not reproduced by the simple payout relationship observed in September.
We have not established when or why that ranking relationship changed, or recovered the current calculation. October check
Four other cases exposed different ranking problems
Twelve inactive wallets appeared with unexplained profits
In nine distinct archived server responses between September 15 at 19:29 UTC and September 17 at 10:24 UTC, 12 wallets appeared with fixed positive monthly figures. One led the ranking with $5,869,917.50; the others’ values started at $189,512. Each retained its own figure to the fraction of a cent across the approximately 39-hour episode. Dated episode records
The sources examined showed no recorded Polymarket-contract activity since 2024, no profile page and no all-time leaderboard entry for these wallets. Three other Polymarket sources did not provide comparable profit figures. The group was absent in surrounding captures, but the archived responses cannot establish uninterrupted visibility between observations. Its appearance is established; the origin of the values remains unresolved. These were saved page-data responses, not screenshots proving what every visitor saw.
Old accounting amounts repeatedly entered recent rankings
A separate pattern affected 13 dormant accounts. Fixed historical amounts, ranging from $4,403.78 to $102,266.87, reappeared in daily, weekly, monthly or category results. For one account, the discrepancy traced to different accounting treatment of a winning position in a 2021 Super Bowl market. That amount came from an old difference rather than a new winning bet. Historical accounting and archive records
Archived copies of the website’s monthly leaderboard placed members of this group in the top 20 on nine dates from July through September. On September 1, four were in the top 20 and the highest was No. 6. Appearances clustered near the start of the month, when active traders had accumulated fewer new gains. Some accounts nevertheless appeared, disappeared and returned while the entry threshold was falling. The old values explain their ability to rank highly; the intermittent inclusion remains unresolved.
A reconstructed loss appeared as a $980,000 monthly gain
On May 6, the website ranked a marketing account linked to the Gravia copy-trading service third for the month with $980,352.94 in profit. Its reconstructed trades and payouts instead showed a loss of approximately $5,268. May reconstruction and dated ranking
The account repeatedly created and redeemed matching Yes and No shares. A complete pair backed by $1 can return that same $1; returning the deposit does not itself earn a dollar of profit. The unsupported credits followed the April 29 introduction of new settlement software. Six of 42 accounts examined closely in the following nine-day window had confirmed unsupported credits. That is a selected sample, not a platform-wide rate.
At 18:00 UTC on May 8, the inflated totals were reset; an unrelated account’s $65,855.16 credit disappeared in the same hour. The published numbers changed. The reset does not establish that funds left the wallets, or that an affected wallet intentionally caused the error. The evidence supports a settlement-related accounting problem without recovering every internal calculation.
A missing starting value turned lifetime profit into daily profit
On June 23, the public feed ranked Inaccuratestake first for daily profit with $3,987,825.07, although the account had made no gain that day. Its accumulated lifetime profit appeared as that day’s result. Across 64 examined accounts, 35 with no daily gain were shown with a combined $6,186,805.38 of daily profit. These rankings come from the public feed; the investigation did not recover a usable website archive for the same episode. June records and accounting comparison
A separate, mechanically drawn comparison of 103 accounts tied the error to missing midnight starting records. Six of seven accounts without a baseline showed lifetime profit as daily profit. None of the 96 with a baseline did. This strongly supports the missing-baseline explanation, while the May case concerned unsupported credits added through a different accounting pathway.
The error concentrated in accounts missing the daily baseline
103 mechanically selected accounts · saved June episode
| Starting record | Accounts examined | Lifetime profit shown as daily profit |
|---|---|---|
| Missing | 7 | 6 |
| Present | 96 | 0 |
The comparison supports the mechanism in this episode. It does not estimate the error rate across all Polymarket accounts or dates.
Baseline definitions and comparison ↗Different feeds could send different winners downstream
The website changed its leaderboard source on September 10. The older public feed remained available without a retirement notice. Its monthly results counted from the beginning of the calendar month; the newer feed used roughly the preceding 30 days for ordinary accounts. Both could be labeled “monthly” while answering different period questions. Papeasy’s fitted running total was a separate finding, not the ordinary-account definition. Feed comparison and switch records
On September 18, the two daily sources shared only 28 of their top 50 names. An outside product using the older source could therefore show a materially different set of leading traders from the website.
The dormant-account figures also reached copy-trading tools. At 01:53 UTC on September 17, 12 of the 13 dormant accounts appeared in the older feed’s daily top 50. In a saved Stand capture, six of its top 10 belonged to that group, each beside a “Copy” button. Three open-source bots used the older feed to select accounts to follow. The evidence shows downstream use of faulty ranking data. It does not establish that a user copied an affected wallet or lost money. Product captures and code review
Other public surfaces could preserve different address scopes or histories. Papeasy’s leaderboard link opened the separate trading wallet’s profile, while the ranking described the collecting address. In September, the Gravia-linked account’s headline result showed a $6,859.59 loss, but its chart retained the erroneous $980,352.94 May peak. Neither comparison establishes a new accounting error; each shows why an address, period and source version must accompany the number.
Rank needs an accounting definition
The five cases show that a field labeled profit can carry gross receipts, a lifetime balance, an old accounting difference, unsupported credits or an unexplained value. Those differences can change who appears to be winning, and they matter when rankings guide trader selection, research or automated tools.
The supported finding is a historical market-data problem. Before treating rank as performance, a reader needs to know which wallet is measured, which costs are included, where the period begins and which source supplies the figure. The evidence does not establish a common cause, deliberate manipulation, ranking-linked rewards or losses to users. It also does not establish that any of these problems remains live today.