At 16:58 UTC on 16 September 2026 I pulled Polymarket's public Gamma API and counted what was actually on the board. Not what the founders say the product is, not what the attorneys general say it is. Just the open markets, the money moving through them in the last 24 hours, and the tags the platform puts on its own listings.
The answer is that both sides of the argument are right, and that is exactly why this fight has taken two years and still is not over.
What the board looked like
The ranked endpoint stops at 2,100 results, so what I have is the top of the list by 24-hour volume rather than every market in existence. Inside that slice: 2,089 open markets spread across 2,068 events, $59.6 million of volume in a day, $2.35 billion of lifetime volume, and $210.1 million of standing liquidity.
Now the part that matters. One event, "Fed Decision in September?", made up of five markets on what the Federal Open Market Committee would do, took $26.8 million. That is 42.6% of everything traded across every event in the day. The top ten markets took 54.9%, the top fifty took 68.3%, and the median market in the top two thousand did $4,035. On a Fed day, Polymarket is a rates desk with a hobby.
Then look at what is sitting on the shelves rather than what is selling. The Sports tag is on 775 events, 37.5% of everything open. Soccer alone is on 422. There are 204 weather events, of which 177 are markets on a single day's high temperature in a single city. Politics carries 504 events and Economy only 85, but those 85 pulled 45.7% of the day's money, because tags overlap and every Fed market is tagged both ways.
So by count this is a sportsbook with a politics section. By revenue on this particular Wednesday it is a macro venue. Ask which one it "really" is and you have walked into the same trap the courts are stuck in.
For colour: Barcelona's match that day traded $646,800 and Atletico Madrid's $492,270, while a market on whether Elon Musk wins the 2028 Republican nomination did $947,805 and one on United Russia's seat count did $791,710. A prop on next week's temperature in a US city sits on the same screen, priced the same way, settled the same way.
Six states, one federal shield
I stopped trusting the round numbers going around ("twelve states are suing") and went to look for filings I could name. Here is what is verifiable:
- Arizona, 17 March 2026. Attorney General Kris Mayes filed 20 criminal counts against Kalshi, the first criminal charges any state brought. Sixteen counts on sports, four on elections, including contracts on the 2026 Arizona governor's race and the Secretary of State race. A federal judge later blocked Mayes from prosecuting, holding that federal law likely preempts the state.
- Rhode Island, 22 May 2026. Attorney General Peter Neronha sued both Kalshi and Polymarket in state Superior Court. His line is the cleanest summary of the state case anyone has produced: "There is no substantive difference between sports betting and 'events contracts' in this context; Kalshi and Polymarket know that, and we know that."
- New York, 31 July 2026. Governor Hochul and Attorney General James sued Kalshi over an unlicensed gambling operation, unpaid gaming taxes, and letting 18 to 20 year olds trade where mobile sports betting is 21 and over. James: "No matter what they call themselves, prediction markets like Kalshi are gambling platforms, plain and simple."
- Connecticut, 27 August 2026, suing over unlicensed sports betting, after a cease and desist in December 2025.
- Nevada and Tennessee issued cease and desist letters, and in Tennessee a federal court granted a restraining order in January 2026 blocking the state from enforcing its gaming laws.
Notice the pattern. States keep winning the argument on the merits of what the product is, and keep losing on who gets to decide. The platforms are registered with the Commodity Futures Trading Commission, and the CFTC has gone to court against states to keep them off federally registered venues. That is the whole trick: not a claim that this is not gambling, but a claim that a different regulator owns it.
The protections do not come along for the ride
This is where the framing stops being a lawyers' argument and starts costing real people money.
A local reporting project in Pennsylvania worked through what a user actually loses on 15 July 2026. Prediction market exchanges are not licensed or supervised by the Pennsylvania Gaming Control Board. State platforms are 21 and over, prediction markets take 18 year olds. State platforms are bound into the state self-exclusion programme, prediction markets are not. And state regulated gaming paid Pennsylvania about $2.98 billion in taxes and fees in 2025, a number the exchanges contribute nothing to.
Three quotes from that piece do more work than any position paper. Josh Ercole of the Council on Compulsive Gambling: "They don't know that it's a prediction market versus a sportsbook app." Doug Harbach of the Gaming Control Board: "If you are going to make gambling much easier to access, you should expect more people developing problems." And Jody Bechtold of the Better Institute, which is the sentence I would put on the cover of the whole debate: "We're seeing people that have self-excluded with the state, then got into prediction markets, and the cycle started all over again."
A self-exclusion list that a person can step around by opening a different app is not a self-exclusion list. Whatever the correct regulatory theory is, that is a hole in a consumer protection system that took twenty years to build.
When the market knows before the world does
The sports and weather markets are the boring part. The structurally novel risk sits in markets on decisions made by small groups of people behind a closed door.
On 10 October 2025 the Norwegian Nobel Committee announced Maria Corina Machado as the Peace Prize laureate at 11:00 in Oslo. Starting around 00:45 Norwegian time, roughly ten hours earlier, her price on Polymarket went from 1 to 2% to over 43% within about fifteen minutes, and to 73% by 02:00. Decrypt reported that day that the account driving it had been created within the previous ten days and cleared about $80,000; the market had traded over $21.4 million since opening in July. Reporting out of Norway, first by Aftenposten and Finansavisen, put the move at under 4% to about 69% in the final hour with more than $1.8 million flowing in, across three accounts with no trading history that between them made around $90,000. Nobel Institute spokesperson Erik Aasheim: "We're looking into it." By February 2026 the Institute's internal review concluded a cyberattack was the most likely source of the leak, without identifying who was responsible.
Read that as a market design problem rather than a scandal. A contract on a committee's secret decision pays out to whoever learns the decision first. It creates a cash bounty on a leak, denominated in dollars, payable to an anonymous wallet, on a venue that lists such markets by the hundred. A bookmaker taking bets on the same question has exactly the same exposure, which is why most licensed books either refuse those markets or cap them at pocket money. The exchange model does not have that instinct, because in a financial market trading on better information is usually the point.
If you work in this market
Three practical things, since most people reading this sell marketing rather than contracts.
The traffic is not where the headlines are. On the board I pulled, tens of thousands of dollars of daily volume sit on hundreds of long tail sports and weather markets while the money concentrates in a handful of macro events. An audience built on "bet on politics" content is being pointed at the thinnest, most legally exposed part of the product.
The legal ground moves monthly, and it moves per state. Anything you publish with a state list in it is wrong within a quarter. Date your claims and link the filing, or do not make them.
And the category label you use has consequences beyond tone. Ad networks, app stores and payment processors are deciding right now whether "event contracts" inherit gambling restrictions, and the state attorneys general are handing them the argument that it does. On AntiMedia we do not link to operators in pieces like this, prediction markets included, and after a day inside these filings I am more comfortable with that rule than when I started.
The honest summary is not that prediction markets are a scam. It is that they took the most profitable half of the bookmaker's business, kept the parts a financial exchange is good at, and left behind the age gate, the self-exclusion list and the tax bill. Someone eventually has to decide whether that trade is allowed. Until then, the 18 year old who cannot open a sportsbook can trade the same game five minutes later, and nobody in the chain is required to notice.