On-Chain: 152 Wallets, a 97.2% Win Rate — and One Indictment

A senior Polymarket official told CNN on Thursday that the platform had referred dozens of accounts that showed signs of potential military insider trading to the Justice Department — referrals that had never been disclosed and that, the official said, predated the watchdog analysis published the day before. Grade the exhibit before you grade the story: this is one unnamed company official speaking while the platform is under federal investigation. Asked to confirm or deny, the Justice Department did not respond. "Referred" is a verb with a lighter burden than "indicted," and the distance between those two words is where the real story lives.

That distance has a number attached to it, and the number is not "dozens." It is 152. Polymarket is a prediction market — event contracts where traders buy and sell shares in a yes-or-no outcome — and in research released Thursday the Anti-Corruption Data Collective, a nonprofit group of academics and illicit-finance investigators, identified 556 wallets on Polymarket International that fit what it calls an "Orca" profile: freshly created accounts that snap open, place one well-timed long-shot wager, cash out, and disappear. Of those, 152 bet on military and defense markets, earned about $8 million combined, and carried an average win rate of 97.2%. No legitimate strategy explains a 97 percent hit rate on long shots except information its holder was not supposed to have. That is why this reads as an intelligence problem rather than a betting quirk, and why the joke about prediction markets stops the moment the win rate does.

The report — Classifying Insider Trading Risk, released Thursday — grades itself. ACDC worked from Polymarket's settled markets through May 5 — trades on Polymarket International settle on a blockchain, so every wager, wallet, size, and timestamp is public even though the humans behind them are not — and screened for long-shot behavior defined tightly: wagering $2,500 in an hour or less on an outcome priced at 35 percent or less. Across the 78,496 long-shot wagers it examined, 12,355 transactions involving Orca wallets sat in military markets, visualized in the report's own "Bubblemaps" clustering of related wallets. The examples ACDC picks out are specific because they have to be. Before U.S. military action in Iran in June 2025, an Orca bet landed hours before the strikes, and a trading bot then copied the position with $200,000 and a deep-pocketed whale with $100,000. Before the U.S.-Israeli strikes on Tehran in February, Orca wagers were followed by a flurry of first-time long-shot bets from the same species of copycat. Every copy multiplies the market's move, which is to say every copy makes the leak louder.

That amplification is the piece of the story the referrals headline buries. A prediction market is a pricing machine: an insider's bet moves the price, and the price is public before the news is. The same blockchain that lets ACDC and Polymarket spot the Orcas in hindsight hands the tip to everyone in real time — including those copycat bots and whales, who owe no duty to anyone and break no law by front-running a signal they are allowed to read, and including the thought, in ACDC co-founder David Szakonyi's warning, that foreign-intelligence agencies are not monitoring these markets — an assumption he called naive. A classified fact, once it becomes a liquid contract, changes legal identity: before the trade it is a secret protected by law; after settlement it is a market price published to every feed on earth. Insider trading on a state secret is a crime with a dollar figure. Broadcasting the secret is an intelligence loss with no dollar figure at all, and the second of those is the larger damage.

The enforcement that has actually happened gives the other half of the picture, and it is worth keeping as a timeline beside the Orca math.

Date	What happened
Mar 24	CNN reported a trader banked nearly $1 million from dozens of accurate bets on U.S. and Israeli strikes on Iran, some placed hours before the news.
Mar 30	Federal prosecutors in Manhattan were exploring whether prediction-market bets had violated insider-trading and other laws.
Apr 23	SDNY and the CFTC charged Master Sgt. Gannon Ken Van Dyke, who helped plan the January capture of Nicolás Maduro, with trading on classified timing information for more than $400,000.
May 27	DOJ unsealed a complaint against a Google employee, Michael Spagnuolo, with more than a million dollars in profits from confidential internal data.
Jun 26	The CFTC began investigating Polymarket itself, the same week two senators demanded a federal probe tied to reporting about fabricated winning bets.
Aug 20–21	ACDC's report lands; Polymarket's referrals surface.

Now watch what the timeline does not contain. The April 23 case was the first application of the Commodity Exchange Act's insider-trading provisions to a prediction-market trader, in a joint announcement by the U.S. attorney for the Southern District of New York, the acting attorney general, and the FBI director; the CFTC filed a parallel civil action the same day, and Van Dyke now faces five federal charges, including wire fraud, and has pleaded not guilty. The legal theory is real, and so is the enforcement appetite. What the ledger has not produced is a person drawn from any of the 152 flagged wallets. Van Dyke is the one military indictment on the board, and ACDC notes he built his position gradually — which is precisely why he did not trip the long-shot wires that flag Orcas. The statistical net that caught 152 wallets caught zero persons; the case that produced an indictment came through a different door, a security clearance and a planning table, not a pattern. Polymarket's own official concedes the Orca signals are "clues" that do not by themselves prove wrongdoing, and ACDC concedes that pure luck is theoretically possible. Two adversaries describe the same wall: on-chain, you can see everything except who was holding the mouse. A wallet owner who was never asked to verify identity cannot be named, which is the design, not the oversight. A referral is conduct; an indictment is a burden; the medium converts the first into a statistic and the second into a hard problem. ACDC says its research surfaced dozens of wallets that had not previously been reported to authorities, and whether those overlap with Polymarket's "dozens" cannot be checked from outside the room — no public ledger reconciles the two counts.

The innocent reading deserves to be stated plainly, because it is real. Polymarket has spent the year being pounded by scandals and has responded visibly: in March it announced enhanced market-integrity rules banning trades based on insider tips, CEO Shayne Coplan told the CFTC the company hired a former FBI official to build proprietary surveillance, and the company says it referred dozens of trader wallets to the authorities, including in the Maduro case. On that telling, this is a compliance machine doing its job — spotting, flagging, referring, cooperating — and the Orcas are the proof of concept. On the other telling, the platform is a channel for secrets that its own ledger broadcasts, and referrals are the cover story for a business model that cannot stop the leak without ending itself. Both readings can be true at once, which is the uncomfortable part, and it is why the House Oversight Committee is now scrutinizing prediction markets for insider trading while a bipartisan group of 44 state attorneys general pushes for regulation on national-security grounds.

Map the situation onto the only episode ordinary readers half-remember: the trader who read the war dispatch before the crowd, back when the price moved at the speed of a physical wire. In that world the leak ran through a wire that belonged to someone, so finding the wire meant finding the leak. Prediction markets have no wire to cut. The tip is posted to the order book itself — public, machine-readable, and unattributable to a person. That mapping holds until you hunt for the owner of the wire, and then it detonates: nobody owns this channel, so you cannot close the leak without closing the market. It is why ACDC's own recommendations run to market design rather than more referrals — mandatory identity proofing for traders, withholding payouts on suspicious trades, and banning the markets where non-public information is most actionable. Enforcement cleans up after the signal goes out. Only design keeps the signal from being posted in the first place.

Read the market signal, meanwhile, for what it is not. As of Thursday, the broad crypto complex was rallying: the crypto fear/greed index sat near 72, firmly in greed territory; total crypto market cap was up about 5.8 percent in a day to roughly $2.6 trillion; bitcoin traded near $77,600, up about 23.7 percent over the previous 20 trading days though still far below the $125,500 high of the past year, per AInvest market data. Prices are saying this is a venue-level legal story — a USDC-on-Polygon problem for one platform — not a token-complex event. That indifference is itself information. The damage in the Orca story does not register as a dollar of token price, because the price that leaked is denominated in secrets, not in tokens; there is no ticker for it, which is exactly why the market shrugs.

So run the grades, because they decide what this dossier actually establishes. Established: 152 military wallets, roughly $8 million, a 97.2 percent win rate, documented copycats, and a platform that says it referred "dozens" of accounts to the Justice Department. Alleged, in the strict sense: that the Orcas committed insider trading — the only adjudication on the military side remains one soldier who does not appear in the flagged set at all. The finding that would change the read overnight is an indictment naming a wallet from the 152, or a referral ledger released that matches ACDC's set exactly. Either one converts "statistically overwhelming" into "legally established," and answers the only question that matters: whether the ledger that shows everyone everything will ever show us who.