The McDonald's bet: a staged “$100,000 win” checked against the real market

In my first note on the New York City Council's prediction-market investigation, I said I would separate the claims into pieces that public trading data can actually test. Here is one, worked all the way through.

On June 20, 2026, the Wall Street Journal reported that a creator, George Makihara, posted a January video showing him winning $100,000 on a bet that President Trump would say “McDonald's” that month — filmed on a look-alike site, not Polymarket's real one. The Journal also reported something independently checkable: “more than 50 accounts made the McDonald's bet in January, public data shows. All of them lost.” That video is one of the examples behind the Council's Aug 12, 2026 letters to four platforms.

Two things, and they are separable. First, the video's own claim: a real $1,000 stake at roughly 1% odds turning into $100,000. Second, the Journal's claim about the real market: that dozens of real accounts took the same bet and every one of them lost. The first is what the Council is investigating. The second is what the public record can test directly — and testing it also tells you whether the first could ever have happened.

I found the real market: “Will Trump say ‘McDonald's’ in January?”, part of Polymarket's recurring monthly “What will Trump say” event, open Jan 2–31, 2026. (A same-wording market from January 2025, about the inauguration speech, is a different market and was ruled out.) Then I pulled every fill anyone made in it — 85 in total, a single page, well under any pagination cap — and separately pulled the current list of wallets still holding Yes tokens, because a position created directly on-chain never produces a trade row and would otherwise be invisible.

1. Did the market exist and resolve as reported? Yes. It resolved No on Feb 1, 2026. Trump did not say the word that month.
2. How many real wallets took the Yes side? 34 wallets show a Yes-side trade; 26 currently hold Yes tokens; 14 of those 26 never appear in the trade data at all. The union is 48 wallets. The Journal reported “more than 50.” Close, not identical, and I would rather say so than round up to match the headline.
3. Did anyone profit at all — even before resolution? The Journal's “all of them lost” is a resolution-level statement. I went further and rebuilt each wallet's cost basis from every buy and sell during the month. Two wallets out of 34 show a nominal round-trip profit before close — both for amounts too small to be worth writing down. Nobody extracted anything resembling real money.
4. Could the trade in the video have happened? No. The largest single real Yes fill in the market's entire history was under $9. Total Yes-side volume for the whole month was under $90. A $1,000 stake at the ~1% price shown in the video would be roughly 11 times the market's entire lifetime Yes-side volume — not larger than any one trade, larger than all of them combined.
5. Was the ~1% price real? It was — but only in the last two days before close, after the price had fallen steadily from a Jan 21 high near 85% and the No outcome was already close to certain. There was no point earlier in the month when a “shrewd long-shot call” at that price was available.

The real market backs up the Journal's account-level finding in substance, and goes further: nobody profited from this market's Yes side even under a more generous test than “who won at the end.” The trade the video depicts is not merely unverified; at that size it could not have occurred in this market at any point in its life. And the one detail the video got right — the ~1% price — existed only at a moment inconsistent with the story the video tells.

An exact match to the Journal's count. I get to 48; the Journal reported more than 50. The likeliest explanation is that a few more wallets created a Yes position on-chain and exited it before the snapshot, leaving no trade row I could attribute to the Yes side and no current holding — but I have not confirmed that, so I am noting the gap, not the reason for it. Trade-derived notional also runs about 18% below the market's reported aggregate volume, consistent with some activity beyond what these two public endpoints capture.

Anything about what Makihara knew, was told, or was paid. Trading records do not answer that, and this note does not try to.

If you are a reporter working the New York investigation, or checking any other specific prediction-market claim, send it my way — For journalists. No charge.

• Katherine Long, Caitlin Ostroff, Neil Mehta and Brenna T. Smith, “They Looked Like They Were Getting Rich on Polymarket—but None of It Was Real,” The Wall Street Journal, June 20, 2026.
• New York City Council, announcement of the prediction-market marketing-practices investigation, Aug 12, 2026.
• Polymarket public data API — market will-trump-say-mcdonalds-in-january: full trade history and current Yes-token holders, retrieved Aug 16, 2026.