Score: A real market with a real record — wearing three numbers that aren't. Polymarket called the 2024 election better than the polls, and that part checks out. But "Polymarket" is now two different products with two different rulebooks, the "free trading" era ended in 2026, the famous "$2 billion from ICE" was $1.6 billion actually paid, and the "$21 billion valuation" is a round that, last anyone confirmed, hadn't closed — the last confirmed close is $15 billion. The platform is legitimately big. You just have to read it in the right units.
What a prediction market actually is
Strip the branding off and Polymarket is a betting exchange for yes/no questions. Each market asks one question about a future event — who wins an election, whether a rate cut happens, who lifts a trophy. You buy "Yes" or "No" shares on an order book, the same way you'd trade a stock. Shares trade between $0.00 and $1.00. If you're right, each share pays out exactly $1.00. If you're wrong, it pays $0. A price of $0.62 is the market saying, roughly, "62% chance." You don't have to hold to the end — positions can be sold any time before the market resolves.
Two things follow from that design, and they're the whole game. First, Polymarket never takes the other side of your trade. It's peer-to-peer: your winnings are another trader's losses, minus fees. Second, somebody has to officially declare what happened — and on Polymarket, who declares it turns out to matter as much as the event itself. More on that below, because it's the soft spot in the whole product.
One more mechanic worth knowing: markets are curated, not user-created. Polymarket's internal Markets Team writes the questions, names the resolution source, and publishes the rules for each market before trading starts. You can suggest a market; you can't list one.
Two products, one brand — don't mix them up
This is the distinction almost every headline gets wrong. Since December 2025 there are two Polymarkets:
- Polymarket (global) — the original, at polymarket.com. It runs on the Polygon blockchain, settles in crypto (USDC, increasingly labeled "pUSD" in the platform's own docs), uses pseudonymous wallet accounts, and is not CFTC-regulated. It has blocked US users since 2022, and Reuters reported in October 2026 that the international platform is run through an entity registered in Panama.
- Polymarket US — the regulated twin, at polymarket.us, run by QCX LLC (the exchange) and QC Clearing LLC (the clearinghouse). It's a CFTC-designated contract market (DCM) and derivatives clearing organization (DCO), trades in plain US dollars with full identity verification, and launched December 3, 2025.
Different fees, different accounts, different market catalogs — and, critically, different ways of deciding who won. Any sentence that says "Polymarket does X" without saying which Polymarket is probably wrong about one of them.
Who decides who won? On the global platform: a token vote
On Polymarket US, settlement is boring in the good way: official results, exchange rulebook, winners get $1.00, losers get $0, done.
The global platform resolves through UMA's Optimistic Oracle, and it's worth walking through because your money rides on it:
- After the event, anyone can propose the outcome — but they have to post a bond, typically ~$750.
- A 2-hour challenge window opens. If nobody disputes the proposal, it stands and the market resolves — about two hours after the proposal.
- If someone disputes it (posting a counter-bond), there's a second proposal round. Dispute that too, and the question escalates to UMA's Data Verification Mechanism — a vote of UMA token holders. Debate runs 24–48 hours, voting about 48 hours, and a fully disputed market takes 4–6 days to settle.
- The bond pays the winner: a successful proposer recovers their bond plus half the disputer's, and vice versa. An "Unknown/50-50" ruling settles every share at $0.50.
Most of the time this is a non-event. UMA says roughly 99% of assertions have gone undisputed since 2021, and an independent on-chain forensic measured a dispute rate of just 0.157% across ~205,000 proposals in May 2026 — about 1 in 640. Both numbers can be true at once.
The problem is what happens in the tail. Polymarket logged more than 1,150 disputed markets in the first five months of 2026 — already past its full-year 2025 total (Galaxy Research, June 2026). And a May 2026 Wall Street Journal investigation found that in most disputed markets, more than half of the UMA votes came from the 10 largest wallets, at least 60% of active voters could be linked to live Polymarket accounts, and roughly 1 in 5 disputes had a voter holding a position in the very market being judged. An independent on-chain rebuild put that conflicted-voter figure at 18.65% — and called it a floor.
That's not hypothetical. In the 2025 market on whether Ukraine's President Zelenskyy would be photographed wearing a suit (~$237M in volume), an initial "Yes" was disputed and UMA finalized "No" — after numerous outlets had described the outfit as a suit. Critics alleged large UMA holders were positioned on "No"; that's an allegation, not a proven finding, but Yes holders lost either way. In June 2026, a market on whether Strategy (formerly MicroStrategy) had sold bitcoin resolved "No" even though the company's own SEC filing disclosed a sale of 32 BTC inside the market window — the ruling turned on when the sale was confirmed, and Yes holders sued in New York court in July. You can call the event right and still get paid wrong. That's resolution risk, and it's priced into nothing.
Who's behind it
The founder. Shayne Coplan, born in 1998, a New York City native who enrolled in computer science at NYU and left after about one semester. He'd been in crypto since buying Ethereum in 2014 at around $12. He founded Polymarket in 2020 and launched it in June 2020, building the early version essentially alone out of a makeshift office in his apartment bathroom during lockdown. The original operating company is Blockratize Inc., a Delaware corporation. After the ICE deal, Bloomberg's Billionaires Index estimated Coplan's net worth at roughly $1 billion, with a stake of about 11%.
The backers. The cap table is a tour of crypto and tech money: a $4M seed led by Polychain Capital (2020); $25M Series A (General Catalyst) and $45M Series B (Founders Fund) announced together in May 2024 — Ethereum co-founder Vitalik Buterin participated in the funding, and Peter Thiel invested personally; a ~$200M Founders Fund-led round at a ~$1B valuation in June 2025; and 1789 Capital in August 2025, the deal that also brought Donald Trump Jr. in as a strategic advisor — a political connection worth naming out loud on a platform whose biggest markets are elections. Pollster Nate Silver, founder of FiveThirtyEight, has been an advisor since July 2024.
Control. On the global platform, the two levers that matter are market curation (Polymarket's own team decides what's listed and how it's worded) and resolution (the UMA vote above, plus Polymarket's ability to publish "additional context" clarifications after trading has begun — the mechanism critics point to when close calls break one way). On the US side, control is the conventional kind: a CFTC-regulated exchange and clearinghouse under a rulebook.
Three things people still get wrong
This is where the record needs correcting, because all three errors flatter the platform.
1. "Polymarket is free to trade." Not since 2026. For most of its life the global platform charged zero trading fees. That era ended in stages: taker fees started in January 2026 on 15-minute crypto markets, spread to some sports in February, and went category-wide with "Fee Structure V2" on March 30, 2026. Makers still pay nothing. Takers pay a fee that peaks when a share costs $0.50 — exactly where uncertainty, and volume, is highest:
| Category (global platform) | Taker fee rate | Fee on 100 shares at $0.50 |
|---|---|---|
| Crypto | 0.07 | $1.75 |
| Sports / Economics / Culture / Weather | 0.05 | $1.25 |
| Finance / Politics / Tech | 0.04 | $1.00 |
| Geopolitics / world events | 0 | $0.00 |
The one honest exception: geopolitics and world-events markets are still explicitly fee-free — Polymarket's own docs say it does not charge fees or profit from trading on them. The category that made the platform famous is the one category that still makes it nothing per trade. Polymarket US runs a similar curve (a taker coefficient of 0.0695, maxing at about $1.74 per 100 contracts at $0.50).
2. "ICE invested $2 billion." ICE has paid $1.6 billion. On October 7, 2025, Intercontinental Exchange — the company that owns the New York Stock Exchange — agreed to invest up to $2 billion at roughly an $8 billion pre-money valuation, and became the global distributor of Polymarket's event data to institutional investors. "Up to" was doing heavy lifting. ICE's own March 27, 2026 update broke it down: $1 billion invested in October 2025, a second $600 million closed March 27, 2026, plus authority to buy up to $40M more from existing holders. That's $1.6B actually in, against a $2B ceiling. Still an enormous vote of confidence from the NYSE's owner — it just isn't the number in the headlines.
3. "Polymarket is worth $21 billion." Not in any closed round. The last valuation Polymarket has confirmed as closed is $15 billion (April 2026). The $21B figure comes from an August 31, 2026 WSJ/Bloomberg report of a ~$1B raise led by 1789 Capital — and Bloomberg reported on October 20, 2026 that the round was still being negotiated and had not been announced by Polymarket. Some tracker sites list it as closed in September anyway. Until the company confirms it, $15B is the number; $21B is a report about a negotiation. (The title of this post uses $15B for exactly that reason.)
The numbers — and why you can't compare them
Prediction-market "volume" is the most abused stat in the industry, because three different yardsticks circulate under the same word: on-chain volume (trades counted on the Polygon order book, the Dune/The Block dashboard method), notional volume (face value of contracts, often counting both sides — runs several times larger), and company-reported figures (not independently audited). We label each number below with its yardstick. Anyone who doesn't is selling something.
- 2024 presidential election winner market: ~$3.7 billion — the breakout that made the brand (Polymarket's event tally, widely reported November 2024; figures from $3.6B to $3.7B circulate depending on what's counted).
- 2026 World Cup winner market: $4.23 billion — per Dune's post-tournament analysis (July 20, 2026), making it the largest single market in Polymarket's history, passing the 2024 election winner market ($3.69B in Dune's accounting). Polymarket's total World Cup volume in that same analysis was $5.4B. A separate sector report using a different methodology printed $12.56B for Polymarket's World Cup — the two cannot be reconciled, and we go with Dune.
So the honest summary: the biggest single market Polymarket has ever run is a soccer market, it did about four and a quarter billion dollars on the most conservative count, and the election that made it famous is now #2.
Is it actually accurate?
The claim built the brand, so we checked it against the measurements instead of the marketing page. (Polymarket's own "Accuracy" page says its odds are right over 90% of the time a month out and over 96% within four hours — self-published, treat as marketing.)
An independent study (PredictDesks, October 2026) went through 358,236 resolved markets across Polymarket and its US rival. The findings, fairly stated:
- A day before resolution, contracts priced at 90¢ or more won 98.0% of the time, and on standard markets the calibration error a day out was small — under a percentage point.
- On long-running markets, the favorite won 87.7% of the time at 30 days out and 92.8% a day out.
- But restrict the sample to genuinely uncertain markets — priced between 10% and 90% a day before resolution — and the favorite won only 65.9% of the time. The headline accuracy is flattered by the fact that two-thirds of sampled Polymarket markets resolved "No": cheap longshots that were never going to happen are easy to be "accurate" about.
The 2024 election is the fairest test, and it's genuinely a win: on election day Polymarket priced Trump at roughly 60–62% while major poll aggregators showed a tie; he won the Electoral College 312–226 and all seven swing states, and Polymarket's price hit ~95% around midnight, well before the AP call. A Vanderbilt University study found its prices more accurate than polling averages in all seven swing states.
Inside that famous win sits the famous miss: the popular-vote market favored Harris late (sources put it anywhere from ~55% to ~73% in the final stretch — print the range, the miss is the fact), and Trump won the popular vote. Post-cycle reviewers put that market's error at roughly 15–20 points.
And then there's Theo — the French trader ("Fredi9999" and about ten linked accounts, per Chainalysis patterns) who made a reported ~$85 million betting on Trump in 2024. His edge wasn't a secret model. He spent under $100K commissioning YouGov "neighbor effect" polls — asking people who their neighbors would vote for, to get around shy-voter bias — and believed them over the published polls. One trader with better data took $85M out of the biggest market in the world. That's the market working as advertised, and it's also the clearest picture of who you're trading against.
The skeletons, at true weight
The 2022 CFTC order — the founding fine. On January 3, 2022, the CFTC ruled that Polymarket (Blockratize Inc.) had been offering swaps/binary options off a registered exchange — 900+ markets since mid-2020. Terms: a $1.4 million penalty (reduced for cooperation), a cease-and-desist, and a wind-down of the non-compliant markets. Polymarket blocked US users and operated offshore for the next three years. That's the documented record, and everything after is the comeback from it.
The road back ran through an acquisition. The CFTC designated QCX LLC as a contract market on July 9, 2025; on July 21, 2025, Polymarket announced it was buying QCEX — QCX plus its clearinghouse — for $112 million. CFTC no-action relief followed in September, was amended in December, and Polymarket US launched December 3, 2025, sports-first and invite-only, opening to everyone in May 2026. The platform that got fined for running an unregistered exchange bought a registered one. Clean, legal, and not cheap.
The FBI raid — real, and closed. On November 13, 2024, the FBI searched Coplan's Manhattan home and seized his phone and electronics, in an investigation into US-user access and settlement compliance. No arrest, no charges. In July 2025, the DOJ and CFTC closed their investigations without charges. Print it at that weight: a raid happened, the files were closed, that's the whole sentence.
The wash-trading study — an estimate, unanswered. Researchers from Columbia Business School and Barnard (Kanoria et al., SSRN working paper, November 2025 — a preprint, not peer-reviewed) used graph analysis of on-chain trades to flag coordinated wallet clusters trading with themselves. Their headline: ~25% of Polymarket's volume over three years flagged as artificial, peaking near 60% in a December 2024 week; by category, sports ran ~45% flagged, elections ~17%, crypto only ~3%. Their suspected motive was users farming a rumored future token airdrop — not the platform itself. One detection method, estimates rather than proof. And note the other half of the record: at publication, Polymarket was reported to be reviewing the study, and no substantive public rebuttal has been located since. A quarter of the volume you're quoted may never have been real two-sided interest, and the company's answer so far is silence.
The renewed CFTC probe — live as of last week. Reuters reported on October 27, 2026 that the CFTC opened a new investigation earlier in 2026 into whether Polymarket complied with the 2022 settlement — geolocation controls, VPN circumvention, US-directed advertising, and onboarding. No charges have been announced; Polymarket declined to comment. The offshore era's US-user leakage is still a live federal question even while the regulated US product operates.
The insider-trading prosecution — alleged, and going to trial. Per the DOJ indictment: US Army Master Sergeant Gannon Ken Van Dyke allegedly accessed the offshore platform through a VPN and placed $33,934 across 13 trades (Dec 27, 2025 – Jan 2, 2026) on markets about whether the US would capture Venezuela's Nicolás Maduro, using nonpublic information from his military duties. Maduro was captured January 3, 2026; Van Dyke allegedly realized ~$409,881. He was indicted in April 2026 (five counts, including wire fraud and commodities fraud), has pleaded not guilty, lost a motion to dismiss on October 26, 2026, and trial is tentatively set for December 7, 2026. Those are allegations in an indictment, not a conviction — but note where the alleged fact pattern starts: with a VPN into the platform Americans aren't supposed to be on.
The founder's spending — reported, attributed. Bloomberg reported (October 20, 2026), from internal records and interviews, that Coplan used company money before Polymarket had meaningful revenue for private-jet flights with friends, New York apartment rent, and other personal expenses; that Founders Fund found it during 2025 due diligence, required repayment of certain expenses and tighter controls, and that the company hired its first CFO in September 2026. That's investor-discipline reporting — not a regulatory or criminal finding — and we print it as exactly that.
The honest math: zero-sum, minus a slice
Here is the part of prediction markets nobody puts on the homepage. Polymarket does not pay you. Other users pay you — every dollar a winner takes home is a dollar a loser put in — and since 2026 the platform takes a slice on the way through.
Worked example, using the new fee schedule. You buy 100 "Yes" shares on a politics market at $0.50 on the global platform:
- Cost of the shares: 100 × $0.50 = $50.00
- Taker fee: 100 × 0.04 (politics rate) × 0.50 × 0.50 = $1.00
- All-in: $51.00
- If you're right: the shares redeem for $100.00 → profit $49.00, not $50.00.
- If you're wrong: the shares are worth $0 → you lose the whole $51.00.
Same trade on a crypto market costs $1.75 in fees (rate 0.07); on Polymarket US, about $1.74. The fee formula peaks at $0.50 on purpose — a coin-flip market is where the trading is, so that's where the tax bites hardest. And because the market is zero-sum before fees, the average trader now loses the fee by construction. To profit long-run you don't just have to be right more than the crowd — you have to be right by more than the slice. (Geopolitics markets remain the exception: no fee, so those stay pure zero-sum.)
As for where the platform's money comes from, now that we've established it isn't paying winners: taker fees (of which 15–25% is recycled back to market makers as rebates, depending on category), the institutional data business — ICE's deal made it the global distributor of Polymarket's probabilities to Wall Street — and, for most of its history, simply the valuation: the equity went from ~$1B to $15B during years when trading revenue was zero.
The risks, plainly:
- Contracts go to $0. A 90¢ favorite that loses loses everything. There is no partial credit for almost right.
- Resolution risk is separate from being right. The suit market and the Strategy bitcoin market both paid the side the plain-English reading argued against, because a token vote was applying fine print.
- Thin markets move on one wallet. Theo moved the biggest market in the world with better polling and ~$60–80M. Smaller markets take far less — and if the wash-trading researchers are even roughly right, some of the "interest" you see isn't real.
- Geo-blocks can strand you. In close-only countries (the US is one, for the global platform) you can exit positions but not trade. The block list changes.
- Circumvention is the charge, not a loophole. The live insider-trading case starts with VPN access to the offshore platform. "Just use a VPN" is how the indictment's fact pattern begins.
- Custody and contract risk (global product). Your collateral sits in Polygon smart contracts as USDC/pUSD. Understand what you're holding before you deposit.
- US tax treatment is unsettled. We found no specific, verified authority on how event-contract winnings are taxed. That's uncertainty, stated as uncertainty — ask a tax professional, not a blog.
Score
Legit platform, bogus headline numbers. Polymarket the market is the real thing: peer-to-peer, transparent order book, a 2024 election call that embarrassed the polling industry, and a regulated US exchange that cost $112M of license-by-acquisition to build. Polymarket the story needs three asterisks: it has charged fees since early 2026 on everything except geopolitics, its marquee investor has put in $1.6B rather than the $2B in the press release headline, and its confirmed valuation is $15B while the $21B round is still, by the last check, a negotiation. Add a resolution system where token holders can vote on markets some of them are betting on, a federal probe into the old settlement reopened four days ago, and an insider-trading trial on the December calendar — and you have the accurate picture: the biggest prediction market in the world, genuinely useful as a signal, and absolutely not the clean, free, $21B machine the headlines describe.
Get posts like this by email: Subscribe to BogusAPY.
Follow BogusAPY on X: @BogusAPY.
Check it yourself: links and sources
- Polymarket (global platform): polymarket.com
- Polymarket US (regulated exchange): polymarket.us
- Polymarket developer docs — fees and resolution pages: docs.polymarket.com
- Polymarket US docs — what it is, fee schedule, market settlement: docs.polymarket.us
- UMA (the oracle behind global-platform resolution): uma.xyz
- CFTC — the 2022 Polymarket order (Docket 22-09) and QCX's designation filings: cftc.gov
- Polymarket on X: @Polymarket
- Dune Analytics — the World Cup 2026 analysis behind the $4.23B single-market figure: dune.com
Not financial advice.
Comments
Post a Comment