Over the past 48 hours, discussion density around prediction market Polymarket has clearly increased, but the attention isn’t driven by it launching any new market. An on-chain settlement statistic, a primary-market financing round, and two sets of regulatory rules have brought the old question of “whether prediction markets are indeed a good business” back into the spotlight.
1. $338.9 million: turning “retail users lose money” into verifiable records
According to public discussion, a statistics report on international platforms’ complete on-chain settlement records shows that among approximately 2.9 million retail accounts, 69.2% failed to return above the break-even line, with total losses of $338.9 million. This statistic covers only international platforms and applies filters to accounts under the criterion of having completed “at least a certain number of trades,” excluding addresses used just to try things out. The data was compiled by an oracle network.
What’s worth reading more than “70% of people lose” is the distribution structure. The median loss among remaining accounts is about $3—meaning most people lose small amounts, in fees, and in emotions; while the bottom 1% of accounts lose at least $4,804. The loss curve isn’t evenly spread out—it’s more like a long tail: most people lose limited amounts, and a tiny minority bears the vast majority of the absolute value.
The second behavioral detail: among accounts that experienced losses within 30 days, 15.2% didn’t trade again afterward; but among profitable accounts, the figure is only 6.1%. In other words, winners exit faster, while losers are more inclined to “try again.” This isn’t unique to prediction markets, but for the first time, the full settlement dataset has quantified it at this level of granularity.
Second, money comes in from the other side: $300 million vs. an estimated valuation of about $21 billion.
Appearing almost at the same time as the retail-loss narrative is another line from the funding side. According to public discussion, the number of publicly discussed crypto primary-market financing projects in September was broadly stable month over month, but the financing amount rose more than 71% month over month. Among them, Polymarket received a $300 million investment, implying a valuation of about $21 billion. Related claims still need cross-verification from more independent sources.
If this deal is ultimately confirmed, the signal it sends is clear: what institutions are pricing is not the “average profit of retail traders,” but rather nominal trading volume, settlement infrastructure, and the ability to expand into new categories. The business model of prediction markets charges for flow, bid-ask spread, and market-making revenue; it is not responsible for making users money—something that is often confused in market sentiment.
The two lines aren’t contradictory: users are losing money, while the platform and market makers are making it. The real question is how long this “pump-priming” structure can hold when regulatory pressure and peer competition both rise, and whether user retention will become a problem before revenue does.
Third, two sets of rules put “whether states can regulate” front and center.
There has also been real progress on the regulatory front. According to public discussion, in late September the U.S. Commodity Futures Trading Commission submitted two sets of rules to the White House’s Office of Information and Regulatory Affairs for review. The first is a proposed rule that would clearly include event contracts in the definition of “swaps,” after which it would enter a public notice-and-comment process. The second is an interim final rule that excludes casino-type entertainment products from “swaps,” and can take effect once the review is completed. Both sets of rules are categorized as “non-material economic impact,” with relatively low review thresholds.
Once event contracts are considered “swaps,” they fall under federal jurisdiction, making it hard for state regulators to reach in. Multiple states are currently suing prediction market platforms. New York has filed a lawsuit against Polymarket, while federal agencies are suing state governments in the opposite direction. Court rulings vary across jurisdictions, and the disagreement has been pushed to higher levels.
The key here isn’t the fine details of the rules, but the timing gap: the interim final rule can almost take effect immediately, while the proposed rule still has to go through the full notice-and-comment process. That means that, in the short term, a “federal priority, casino-type excluded” framework may appear first, and then the full definition of event contracts will be filled in gradually. For the platform, the compliance path will become clear first, and the business boundaries will follow.
Fourth, what the market is talking about: a 21% swing—and a checklist of pitfalls for copy-trading.
Beyond macro narratives, there are two more ground-level signals.
First: in a sub-market about the number of tweets, the win rate of an option dropped from 61.5% to 40.5% within one hour—a swing of about 21%. This kind of drastic volatility shows that even when the event itself has nothing to do with asset prices, if the order book depth is insufficient, any new piece of information can trigger repricing. It isn’t evidence of manipulation, but it is evidence of liquidity.
Second: someone compiled a checklist of pitfalls for copy-trading on Polymarket, including: the official API latency is uncertain—sometimes you can check a trade within one or two seconds, other times you may need to wait 10 to 30 seconds; on-chain raw logs are faster, but you have to decode them yourself; the platform has more than one trading contract producing trades, and if you only scan one contract you’ll miss over 20% of orders; the same match can have multiple similarly named markets—if key fields are missing, you may trade the wrong market; a timeout error when placing an order doesn’t mean no trade happened, and blindly retrying may buy twice; reconciliation must be done retrospectively from the chain—after a restart, using old balances can overwrite the true positions.
These details don’t amount to accusations against the platform itself, but they do indicate one thing: the user-side tooling stack in this market is still in its early stage. Information asymmetry isn’t just in the events themselves; it also exists in interfaces, contracts, and reconciliation methods. For a market that sells the idea of “rational pricing,” this is friction cost that needs to be taken seriously.
Fifth, where the disagreement lies—and under what conditions the narrative could be rewritten.
The bullish camp believes: endorsement from the primary market, ongoing category expansion (elections, weather, sports, and personality statements), and an increasingly clear regulatory framework—when these three reinforce each other, they will push prediction markets toward mainstream financial market infrastructure. The bearish camp believes: long-term negative returns for retail traders will erode retention. If a 15.2% loss-related churn rate scales upward as the market grows, customer acquisition costs will keep rising. At the same time, the tug-of-war between state-level lawsuits and federal rules still leaves room for reversals, and legal certainty is far from already in hand.
There are four points of falsification that need to be watched. First, whether the $300 million investment and the implied valuation of about $21 billion have obtained confirmation from more independent sources. Second, the final text of the two sets of rules—especially whether the wording “event contracts equal swaps” has been diluted. Third, the direction of rulings in New York and other states’ lawsuits—whether effective decisions emerge that directly conflict with the federal position. Fourth, whether on-chain active accounts and the retention structure improve—especially whether the proportion of users who stop trading 30 days after incurring losses continues to rise.
Before that, this market will tell two stories at the same time that don’t really conflict: users lose money, while the platform gets more expensive. Which story ultimately defines Polymarket depends on how regulation is written—and whether the product can turn “most users with negative returns” into “most users who are willing to keep participating.”