Every argument for prediction markets starts in the same place. A contract paying one dollar if an event happens and nothing if it does not turns a price into a readable estimate of likelihood, and a trader who misreads it loses what the contract cost. That bounded downside is why the category can be defended as information infrastructure rather than a faster way to lose money.
Perpetual contracts do not work that way, and since 3 September 2026 the two products have shared a brand. Jose Antonio Lanz of Decrypt reported that day that Polymarket had opened perpetual futures with leverage up to twenty times and no expiry date, routing United States traders to its separate regulated venue instead. We covered that split in The Same Name Covers Two Different Rulebooks. What follows is about the product on the other side of it, read on 22 November 2026 out of its own public instrument feed and documentation.
Ninety Contracts Now Sit Behind the Same Brand
The public instruments endpoint at api.perpetuals.polymarket.com answers without authentication, so the catalogue can be counted rather than estimated. Our own count on 22 November 2026 returned 90 listed perpetual instruments: 41 on crypto assets, 39 on single equities, six on indices and four on commodities. Every one carries a minimum order notional of ten units of quote collateral and settles funding once an hour.
Leverage is not uniform across them. Four markets allow up to fifty times and they are the largest ones, being bitcoin, ether, the S&P 500 and the Nasdaq 100. Eight more allow twenty, covering gold, silver, the two oil benchmarks, three crypto assets and one equity index. Sixty-four sit at ten, where the whole equity list lives, nine allow five, and five allow three, that last group being small crypto assets whose entire per-market notional ceiling is fifty thousand. Polymarket's listed markets page adds pre-IPO perpetuals tracking a private valuation before listing.
A Trader Who Picks Nothing Is Handed the Ceiling
The finding that matters most is not the size of the catalogue but what happens to a trader who never touches the leverage control at all.
Polymarket's margin page sets out a three-step resolution order for leverage and margin mode. An explicit per-instrument setting wins. Failing that, an exchange-managed account group default applies, capped at the instrument maximum. Failing both, the instrument's own maximum sits at the bottom of the ladder. The trading page says it in one sentence: "Without either, leverage defaults to the instrument's maximum and margin mode to isolated."
Read that against the catalogue above. On bitcoin, ether, the S&P 500 and the Nasdaq 100 the instrument maximum on 22 November 2026 is fifty. The documented fallback is therefore the most aggressive setting rather than the gentlest, and it lands on the trader who has formed no view about leverage, because one who had would have set it.
That ordering is a design decision, not a technical necessity, since nothing in a matching engine requires an absent choice to resolve upward. Directive 03 of our manifest says deposit limits, cool-off periods and self-exclusion are product features built into the account rather than concessions granted on request, and the same logic reaches the leverage control. A default is the setting that protects the person who did not read the page.
The Liquidation Buffer Comes From the Market, Not From You
The second layer makes that default heavier than it looks, because maintenance margin is not derived from the leverage a trader chose. Both the margin page and the product FAQ give the rate as 0.5 divided by the market's maximum leverage, flat for every position there regardless of size or of the trader's setting.
On a market whose ceiling is fifty that rate is one percent of notional, against an initial margin at fifty times of two percent. The distance between opening a position and having it closed is therefore half the margin posted, and the documentation says as much: a position opened at maximum leverage is liquidated after losing roughly half of it.
There is a fair reading in the other direction. Because maintenance margin is flat while initial margin rises as leverage falls, a cautious trader on the same market gets a wider buffer. The asymmetry is disclosed and it favours restraint. But the floor under everyone's maintenance rate is still set by the highest number the venue chose to offer there, and moving a ceiling from twenty to fifty cuts that rate from 2.5 percent to one percent for every participant in the market, including the ones trading at two times.
A related clause decides whether money can come back out. The same page requires a withdrawal to leave the greater of existing margin commitments and ten percent of open position notional, and its worked example is blunt: a one hundred position at twenty times needs five of initial margin but ten of withdrawal margin, so with five of collateral the position opens and nothing is withdrawable. That floor bites above ten times, exactly where a trader who chose nothing has been placed.
The Help Page Still Answers Twenty
Both of those numbers are first-party, and on 22 November 2026 they disagree with each other.
Asked "what is the maximum leverage", the Polymarket Perps FAQ answers that crypto, the S&P 500, oil, gold and silver are offered up to twenty times, and real-world assets up to ten. The live instruments endpoint we counted the same day reports fifty on bitcoin, ether, the S&P 500 and the Nasdaq 100. That FAQ then works its maintenance margin example at "2.5% on a 20x market", which is the right arithmetic for a ceiling of twenty. Where the ceiling is fifty the rate is one percent, so the adverse price move that closes a maximum-leverage position is two and a half times smaller than the example implies.
This is the plainest Directive 02 problem in the product. Our manifest asks that fees, spreads, settlement sources, custody arrangements and conflicts of interest be stated in plain language on the surface where the trade happens. A trader who reads the explanatory page rather than the machine feed, and computes the distance to liquidation from it, gets an answer wrong in the direction of safety. The developer pages do tell integrators to read live values from the endpoint. The reader of the plain English page gets no such warning.
Four Percent an Hour Is the Published Limit on Holding Cost
Perpetual contracts have no expiry, so convergence is bought with a recurring payment. Polymarket's funding page documents a one-hour charge window, a premium index sampled every five seconds, and a hard cap on the hourly rate of four percent in either direction, there to prevent extreme funding during sustained dislocation. That cap is a ceiling rather than a forecast, but at fifty times leverage, where initial margin is two percent of notional, one hour at the documented maximum costs twice the margin posted at the default setting.
Two things on that page deserve crediting. Funding is a transfer between longs and shorts with no cut taken by the venue, stated outright, and the premium index methodology is disclosed down to the one thousand of quote notional walked on each side of the book.
A Liquidation Still Adds to a Leaderboard Score
Alongside the contracts runs a competition programme, and its published rules are where the product's incentives are easiest to read. Polymarket's trading competitions page scores entrants on profit and loss multiplied by the square root of boosted volume, defines volume as "the notional size of every fill on qualifying positions, including maker and taker fills, closes, and liquidations", and notes that boosted markets multiply that volume for scoring, giving three times as an example.
Being liquidated therefore contributes to a leaderboard ranking. The profit and loss term is negative and the page is explicit that negative scores still rank, so this is not a claim that losing wins prizes. The point is structural: the volume half of the score counts fills generated by forced closure, and volume is the half that can be multiplied. Directive 06 asks that revenue come from many informed participants over years rather than a few ruined ones over months, and a scoring rule built on volume pulls the other way. That is not a point we are inventing. The Commodity Futures Trading Commission said much the same about incentive programmes in August 2026, which we covered in Bought Volume Is Not Demand.
Changing the Price of a Liquidation While Positions Are Open
On 25 November 2026, three days after this was written, several of the numbers above change. Polymarket published the schedule in advance on a dedicated fee and liquidity reward changes page, and the headline item is that the liquidation fee rises from 0.50 percent to 1.00 percent of liquidated notional, all of it routed to the perpetuals insurance fund. The trading fee table moves the same way at the retail end and the other way at the top: accounts below one million of trailing thirty-day volume see taker fees go from 0.0400 to 0.0500 percent and maker fees from 0.0125 to 0.0200 percent, while the maker rebate at the one billion threshold doubles to a full basis point.
Here the perpetual structure collides with our first directive, which asks that resolution rules be published before a market opens and not revised while money is at risk. An event contract can satisfy that literally, because it has an opening and a close. A contract with no expiry has no moment at which its rules are fixed and none at which they stop applying to money already committed, so every rule change on a perpetual is made while positions are open. That is a property of the instrument, not a failure of one venue.
Which makes advance notice the only substitute available, and it is right to say Polymarket gave it. A dated page five days ahead, a dated changelog whose newest entry on 22 November 2026 was stamped 20 November, and a specific list of what changes, is more than the category average. The floor worth writing down is this. Where an instrument cannot have its rules frozen, the venue publishes the change, dates it, says what it does to an open position, and leaves time to close first.
The Only Reasoned Federal Order Stops at Digital Commodities
None of this argues that perpetual contracts cannot be run properly, and the clearest evidence is that a United States regulator has already approved some. On 29 May 2026 the Commodity Futures Trading Commission issued a nine-page order approving KalshiEX's bitcoin perpetual futures contract, and its reasoning is unusually explicit about what makes the design work.
Expiry is what forces a traditional futures contract to converge, the order explains, and a perpetual needs a funding rate in place of it. The bitcoin spot market supports that substitute for three stated reasons: it trades around the clock across broadly distributed venues, so the reference is continuously observable; it is deep enough that a large position holder cannot cheaply move the reference at the moment a payment is calculated; and continuous trading lets arbitrageurs act on the convergence incentive in real time. The Commission held that listing would not violate the Commodity Exchange Act for Kalshi "and other DCMs", confined the analysis to perpetuals on bitcoin or other digital commodities with deep, active and continuous spot trading, and stated that it "does not extend to underlying asset classes other than digital commodities."
That last clause is the sentence to hold on to. Offshore, the same funding mechanism runs on thirty-nine single equities, four commodities, six indices and a handful of small crypto assets. Directive 05 asks that standards be set against the jurisdiction that regulates next rather than the one that has not yet, and the Commission has already written down which part of this catalogue its reasoning does not reach.
Twenty-Nine Pages, and Not One of the Three Tools
We counted one more thing, because absences are measurable too. The Polymarket Perps documentation index lists twenty-nine pages, roughly 74,000 words downloaded in full on 22 November 2026. In that corpus leverage appears 226 times, liquidation 173 and funding 170.
The terms self-exclusion, cool-off, deposit limit, loss limit, risk warning and suitability appear zero times each. The one occurrence of responsible across all twenty-nine pages is the sentence in the FAQ telling builders they are responsible for verifying a user's location, in the answer confirming that order placement is barred from the United States, Canada, Cuba, Iran, North Korea, Syria, Crimea, Donetsk and Luhansk.
Developer documentation is not where a consumer-facing deposit limit would normally be advertised, and this is no proof that no such control exists anywhere in the interface. What it establishes is that the corpus explaining margin, liquidation and funding in formulas does not explain one of the three tools Directive 03 names. The mechanics of losing money are documented to the decimal. The mechanics of stopping are not documented at all.
What a Floor for Leverage Would Actually Say
We argued in Margin and Leverage that margin and leverage would be the category's next fault line, back when prediction markets were still generally fully collateralised. It has happened, and the useful response is a written standard rather than an objection.
Five clauses would cover most of what this product shows. First, an absent choice resolves down, to the lowest tier the market offers, never to the ceiling. Second, the first leveraged order on an account requires an explicit, recorded choice, not a slider that already has a number in it. Third, the plain-language page and the machine-readable feed state the same maximum, and a divergence is a disclosure defect, not a documentation backlog. Fourth, where an instrument has no expiry, a fee or margin change is published with a date, a stated effect on open positions, and a window to close first. Fifth, a leaderboard does not count volume generated by forced liquidation.
None of that stops anyone offering fifty times leverage to someone who asks for it. It stops fifty times from being what happens to someone who asked for nothing. The market that survives regulation, as our manifest puts it, is the one already behaving as if regulation had arrived, and here the regulator has already said where its own reasoning stops.
Where These Documents Are
The instrument counts, leverage tiers, order minimums and notional ceilings here are our own count of the public endpoint api.perpetuals.polymarket.com/v1/info/instruments on 22 November 2026, and that catalogue can change without notice. The margin, liquidation, funding, fee, competition and FAQ statements come from the Polymarket Perps documentation at docs.polymarket.com, read the same day and linked at each claim above. The KalshiEX approval is the Commission's order of 29 May 2026 on cftc.gov, and the launch date is as reported by Decrypt.