The Polymarket “KPMG Wallets” Are a Case Study, Not a Copy Signal

Not on the published record alone. The reported results are worth studying, but they do not establish a repeatable strategy you can copy at today’s price. Nor does the label “KPMG wallets” establish that KPMG or its employees controlled the accounts.

For a beginner, the useful question is not “Did these wallets win?” It is “Would this position still make sense at my executable price, under the contract’s rules?”

What the wallet reports actually show

FinanceFeeds reported that Bubblemaps identified 19 connected Polymarket accounts with 41 wins in 42 earnings bets involving KPMG audit clients. The report explicitly says the record does not prove insider trading and the traders’ identities have not been established. FinanceFeeds

A separate reconstruction by Bitquery, corrected on September 13, 2026, scored the original 19 accounts by distinct markets rather than counting each account’s bet separately:

Earnings-market group	Markets won	Wins implied by entry prices
KPMG audit clients	19 of 21	14.3
Other auditors’ clients	11 of 15	11.3

Those figures are not interchangeable with “41 of 42.” Several wallets betting on the same company’s result do not create several independent outcomes. Bitquery’s 19-of-21 tally also counts a market as won only if every bet in it won; one market had positions on opposing sides. The implied wins are a price-based benchmark, not independently verified probabilities. Bitquery reconstruction

The group’s results are not each wallet’s track record: Bitquery attributed more than half the KPMG-client profit to three bets by one account. It found three additional linked accounts but kept its headline performance figures scoped to the original 19. It could not connect the accounts to KPMG, any person, or the separately reported employee inquiry. Bitquery findings and limitations

There is also a timing problem for would-be followers: that reconstruction records the last bet on March 26, the last sale on March 27 and the last withdrawal on March 28, 2026, with no movement afterward in its September account. That is a historical record, not evidence of an active October trading signal. Bitquery trading timeline

Copying the side does not copy the price

Consider a hypothetical earnings contract. Assume its rules require a specified adjusted earnings-per-share figure to exceed a named consensus estimate for a specified quarter. Before evaluating the trade, check the actual metric, benchmark, reporting source, deadline and treatment of equality. “The company had good earnings” is not enough: Polymarket resolves contracts under their predefined rules, with winning shares paying $1 and losing shares becoming worthless. Polymarket resolution guide

Suppose a wallet bought Yes at 80¢. By the time you notice, the executable ask is 90¢ and the best bid is 88¢. Assume you buy 100 shares as a taker, sufficient depth exists at 90¢, you hold to ordinary binary settlement, and there are no rebates or intermediary costs.

Using the finance-category fee rate in Polymarket’s published schedule, accessed October 4, 2026:

• Purchase: 100 × $0.90 = $90.
• Entry fee: 100 × 0.04 × 0.90 × (1 − 0.90) = $0.36.
• Total cost: $90.36.
• If Yes wins: $100 − $90.36 = $9.64 profit.
• If Yes loses: $90.36 loss.

Verify the specific market’s fee parameters before ordering; this is an illustration, not a quote for a live earnings market.

Your break-even probability is $90.36 ÷ $100 = 90.36%. If your own defensible estimate is only 85%, expected profit is:

0.85 × $100 − $90.36 = −$5.36.

The wallet’s earlier entry could have been attractive while yours is not. The 2¢ bid–ask spread matters too: immediately selling all 100 shares at the assumed 88¢ bid would return only $88 before any exit fee.

A public wallet is not permission to use prohibited information

An unexplained winning record is not proof of misconduct. Equally, public visibility does not override trading restrictions. Polymarket’s policy prohibits using confidential information in breach of a duty of trust, using confidential tips when you know or have reason to know the tipper is prohibited from trading on them, and trading where you can influence the outcome—or at the direction of someone who can. It distinguishes those situations from original analysis of public data. Polymarket market-integrity policy, accessed October 4, 2026

Use a wallet alert as a prompt to investigate, not an instruction to buy. Require an independently explainable thesis, the exact resolution wording, a maximum acceptable entry price and a loss you can afford. If the only thesis is “these accounts usually win,” you have not yet evaluated the trade. The pre-trade reading checklist is a better next step than automatic copying.