On 15 December 2025 Kalshi co-founder and chief executive Tarek Mansour announced that combined contracts were live for every user and had traded more than 100 million dollars in their first week. Eilers & Krejcik Gaming reported in its Front Running newsletter on 2 September 2026 that combinations made up 50 percent of tracked United States sports contract volume in August 2026, against 14 percent in February. In eight months a feature became the product.
Over the same year the Commodity Futures Trading Commission published four rulemaking documents on this category. Their raw text runs to roughly 195,800 words. The word parlay appears in none of them. Neither does combinatorial, multi-leg or accumulator. We are not against the product. We think a structure that half the money already uses should be described somewhere other than a help centre.
A combined contract is one instrument, not a basket of them
The clearest public description of how one is built comes from the exchange filing. On 20 May 2026 QCX LLC, trading as Polymarket US, self-certified a Combinatoric Athletic Outcome Contract with the Commission under 17 CFR 40.2(a). The terms are short and worth reading in the original. Each contract has a notional size of one dollar. Margin is 100 percent of the at-risk amount, so nothing here is leveraged. No position limits apply, and the position accountability level sits at 25,000 dollars of notional exposure.
The payout condition in that filing is strictly binary. The contract resolves to one dollar if and only if every leg is satisfied, and if any single leg fails it resolves to zero regardless of the outcomes of any remaining unsettled legs. The filing also provides for accelerated settlement: once one leg has gone the wrong way, the whole contract can be settled before the other events have even been played.
Kalshi describes a different rule for a product with the same name. Its combos help page, dated 20 June 2026, says a combo pays the product of the value of each individual position. If a named player does not play, that leg settles under the rules of its own market, which typically means its last traded price rather than zero or one, and the combo payout is recalculated as the product including that adjusted value. The page gives its own worked example: a leg settling at 0.70 dollars alongside two legs at one dollar produces a combo paying 0.70 dollars.
That difference is not cosmetic. At one venue a scratched player wipes the ticket out; at the other it scales the payout down. Both are defensible designs, and the scalar rule is arguably kinder than the void rules a retail user meets in a sportsbook. What is hard to defend is that a user moving between the two has no obvious way to know which rule they are under.
The price of a combination is quoted, not posted
A single yes or no contract on an order book shows its own spread. A combination does not work that way. The Kalshi help page states that combo prices come from a request for quote system: the platform sends a quote request to the market, other participants respond with a price, quoted prices may change between viewing and execution, fills are not guaranteed, and once a combo is filled it cannot be canceled or reversed. The page also notes that longer or more complex combos attract fewer participants willing to quote them.
So the trader sees one number. They do not see how that number was assembled, and the arithmetic is unforgiving in a way that a single contract is not. Four legs priced at 0.70 each are worth 0.2401 by multiplication. Move each leg by a single cent to 0.71 and the product becomes 0.2541. A one cent step per leg has become a 1.4 cent step on a 24 cent contract, roughly six percent of the price, and it is invisible because the legs are never shown separately. That is arithmetic rather than a measurement of anyone's quoting, but it is the reason the disclosure question here is different from the disclosure question on a single market. We have written before about how spread and slippage carry costs users do not see; a bundled quote compounds exactly that problem.
Colorado publishes what the same idea is worth next door
Regulated sportsbooks have been reporting the combined bet for years, and one state publishes it in a form anyone can check. The Colorado Division of Gaming breaks out a line called Parlays/Combinations in its monthly sports betting proceeds reports. Reading the June and July 2026 sheets directly and doing the division gives a consistent picture.
- In June 2026 combinations took 39.1 percent of all wagers and held 11.74 percent, against 4.48 percent on everything else.
- In July 2026 they took 34.2 percent of wagers and held 13.22 percent, against 7.73 percent on everything else.
- Across the two months together, combinations were 36.5 percent of the money wagered in the state and 53.4 percent of the gross gaming revenue kept from it.
About a third of the handle, more than half of the operator's take. Nobody in that industry finds this surprising, and none of it is hidden. It is published every month by a state regulator, by wager type, and has been for years.
On an exchange the margin does not go to the house
The obvious next step would be to carry the sportsbook number across, and it would be wrong. Mansour made the structural point himself in the same announcement: "there is no house: when you submit a Combo, traders compete in the open market to take the other side." That is accurate and it matters. A designated contract market is not the counterparty. There is no built-in hold to disclose because there is no book taking the other side of the customer as a matter of business model.
Which is what makes the measured outcome worth sitting with. Bloomberg reported on 28 July 2026, from its own analysis of publicly accessible trade data, that users of Kalshi's app and website had lost a net 294 million dollars on combos since the start of 2026, before fees. No house edge, and the money still moved in one direction: from the people requesting quotes to the people answering them.
That is an information problem rather than a rake problem, and it is important to say so, because the two have different fixes. A rake problem is solved by charging less. An information problem is solved by telling the person on the slow side of it what the fast side already knows. Lowering a fee would not touch this. Publishing the distribution would.
The regulator's only sighting of the product is a footnote
The four documents are the advance notice of 16 March 2026, the public interest proposal of 12 June 2026, the data reporting proposal of 1 July 2026 and Conflicts and Affiliations, 91 FR 50926, of 6 August 2026. Searching all four for the vocabulary of the combined bet returns nothing.
There is exactly one place where the product surfaces. In the Conflicts and Affiliations proposal the Commission writes that it "is also aware that a DCM may operate a trade matching system other than a central limit order book", gives a request for quote mechanism as the example, and hangs footnote 200 on that sentence. Footnote 200 is a link to Kalshi's combos help page. The Commission then asks for comment on whether its proposed order-priority subordination rule, which requires an exchange to fill unaffiliated members ahead of an affiliated market maker at the same price, works at all when matching happens by quote rather than by book.
It is a fair question and a serious one. It is also the only appearance in the 2026 record of a product that was half the sports volume by the month that proposal was published, and it enters as a conflicts-of-interest edge case rather than as a thing users trade. Comments on that proposal close on 5 October 2026, so the door is open rather than closed. We have argued before that an exchange trading against its own affiliate needs structural answers; a quote-driven venue where the trader cannot see the book makes that question sharper, not softer.
Appendix F reads contracts together but never sells them together
There is a near miss in the June proposal worth noting, because it shows the concept was in the room. Appendix F sets out the public interest factors the Commission applies under 17 CFR 40.11(a)(5), and its information aggregation factor says that event contracts should generally be considered in the aggregate: if a small group of contracts appears to convey no meaningful information, the Commission will still consider whether they convey it "when combined with or compared to other event contracts".
The proposal therefore contemplates combining contracts analytically, to decide whether a market earns its place. It does not contemplate combining them commercially, as a single instrument a person buys. The public interest test is applied leg by leg to markets that are, by the time the test is run, mostly being traded in bundles.
The explanation of why it complies is not in the public file
Back to the Polymarket US filing, because of what is missing from it. A self-certification under 40.2(a) comes in two parts: the contract terms, and a concise explanation and analysis demonstrating that the product complies with the Commodity Exchange Act and the core principles. The terms were filed as Attachment A and are public. The explanation and analysis was filed as Attachment B and marked confidential under 17 CFR 145.9.
Requesting confidential treatment is lawful and routine, and we are not suggesting otherwise. It does mean that for the first certified combined contract at a United States venue, the public record contains what the product does and not why the exchange believes it belongs on a regulated market. For a product this size, that asymmetry is worth naming out loud. Directive 02 asks for conflicts, fees and settlement sources in plain language on the surface where the trade happens, and the reasoning behind a novel structure is part of that surface.
A smaller thing from the same file, offered as evidence of how new this all is rather than as a complaint: the cover letter calls the product a Combinatorial Outcome Contract, Attachment A calls it a Combinatoric Athletic Outcome Contract, and the Commission's own product filing entry calls it Combinatorial Athletic Outcome Contracts. Three names, one filing, one product.
One clause that should not outlive the opening bell
The last paragraph of the Polymarket US terms reserves to the exchange, if any circumstance arises that may materially affect the reliability or transparency of a contract's source agency, the authority to designate a new source agency and underlying and to change any associated payout criterion and contract specifications after the first day of trading.
Directive 01 of our manifest reads: "Resolution rules are published before a market opens and are not revised while money is at risk." A fallback is not the problem. Data sources do disappear, and a contract with no answer for that is worse than one with an answer. The problem is an unbounded fallback that reaches the payout criterion itself. The version of this clause that satisfies Directive 01 says what triggers it, who inside the exchange signs it, where the change is published and how quickly, and draws a line at the payout criterion once trading has begun. On a combined contract the stakes are higher than on a single one, because a change to one leg moves the value of a bundle the holder cannot unwind leg by leg.
Four numbers a venue could publish this month
None of this requires new law, and none of it requires an exchange to stop offering the product. It requires four figures that every venue already holds and nobody outside can fully reconstruct.
- Share of contract volume from combinations, monthly. Third parties are already estimating it. The exchange knows it exactly.
- Distribution of leg counts. A market where most combinations have two legs is a different market from one where most have six.
- Realised settlement by leg count. What fraction of combinations with two, three, four or more legs resolved to a dollar. This is history, not advice, and it is the single number a first-time user would most want before their first bundle.
- Effective cost of the quote. The filled combo price against the product of the leg midpoints at the moment the quote was requested. Only the venue can compute this one, because only the venue sees both sides at the same instant.
The first three are reporting. The fourth is the one that matters, and it is the one that no analyst, journalist or regulator can produce from outside. It is also the direct answer to the disclosure gap that request for quote pricing creates. A venue that published it would have moved the argument about combined contracts from anecdote to evidence in a single release.
It is worth noting how much of this is measurable at all only because these are exchanges. Bloomberg could calculate a net loss figure across an entire product line because the trade tape is public, which is not true of a single sportsbook parlay anywhere. The raw material for the honest version of this conversation already exists, and it exists because of how prediction markets are built. What is missing is the venue turning it into something a user sees before they trade, rather than something a reporter reconstructs afterwards. That is Directive 04, and the argument that the number you lead with says what you think matters applies here with more force than anywhere else on the product.
Half the volume deserves more than a help page
The combined contract is not the villain of this category. It is a real structure with a real audience, one venue's settlement rule handles a scratched player better than the industry it borrowed the idea from, and the peer-to-peer design removes the house edge that makes the sportsbook version of the product what it is. None of that is in question.
What is in question is a product that took half of the sports volume between February and August 2026 while the entire federal rulemaking record for the category, four documents and 195,800 words, described it in one footnote pointing at a help centre. Directive 05 says standards are set against the jurisdiction that regulates next, not only the one that has not yet. Sooner or later something will be written about combined contracts by someone. The four numbers above are what the category would publish if it wanted that text to be shaped by evidence it produced itself.