The Public Interest Prong: Prediction Markets Will Be Constrained by Gaming Law, Not Securities Law

The wire moved at 04:12 Chicago time. One sentence: most Americans oppose betting on elections.

No sample size. No field dates. No margin of error. No number attached to "most." No respondent breakdown by party, age, or trading experience. No platform named, no protocol, no token. The word "crypto" appears nowhere in it. The industry it describes is compressed into a subordinate clause — platforms that let people bet on elections and sports.

I read it four times looking for a denominator.

There isn't one. What exists is a proposition, an attribution to Reuters, and a functional description of a category. That is the entire payload.

My default posture toward artifacts like this is to file them as null. Nine years of reading filings instead of press releases does that to you. A poll item with no quantitative disclosure is not information. It is a signal with the data stripped out. I trust the null set, not the influencer.

Then I did what I actually do, which is not read the sentence as journalism but map it onto the statute it will eventually be litigated under.

That is where it stopped being noise.

"Majority opposes betting on elections" is not primarily a data point about public opinion. It is the precise evidentiary form that one specific statutory prong demands. The Commodity Exchange Act does not ask whether an event contract is well-engineered, liquid, or epistemically useful. It asks whether listing the contract would be contrary to the public interest. A poll is an industrial process for manufacturing exactly that finding.

So the collision is this: a legislative standard drafted for agricultural futures in the 1930s, asking a question that a modern polling firm answers on a five-point scale, routed — in the on-chain case — through a resolution layer arbitrated by token holders whose neutrality is an assumption, not a proof.

Nothing in that chain is a code problem. The code is the least interesting part of the system.

Silence in the code speaks louder than hype.

Context: Three Layers, One Statute

A prediction market is not a single piece of software. It is three distinct machines bolted together, each with a different failure surface and a different regulator.

Layer one — collateral and settlement. Where the money sits. Polymarket settles in USDC on Polygon. Kalshi settles in dollars held at a banking partner and clears through a CFTC-registered designated contract market. The difference is not cosmetic. One is a bearer asset with no counterparty; the other is a custodial claim inside a supervised entity. Different failure modes, different insolvency regimes.

Layer two — matching. How orders find each other. Polymarket runs a hybrid: off-chain central limit order book, on-chain atomic settlement. Kalshi runs a conventional centralized matching engine. Throughput is irrelevant here. Event contracts do not stress a sequencer. A market that trades a few million dollars a day across a few hundred binary instruments is a rounding error against any L2's gas profile. Anyone who tells you prediction markets are a scaling problem has not looked at the load.

Layer three — resolution. Who decides what happened.

This is the whole game. Layers one and two are commodities. Layer three is where the money, the liability, and the politics live.

On-chain, the dominant pattern is the optimistic oracle. A proposer asserts an outcome and posts a bond. If nobody disputes within a liveness window — typically two hours for low-stakes, longer for high-stakes — the assertion finalizes. If someone disputes, the question escalates to a token-holder vote. UMA's Data Verification Mechanism is the canonical implementation: a dispute goes to a 48-to-96-hour vote among UMA holders, weighted by stake, to determine the truth of a factual claim about the world.

Read that sentence again and substitute "election outcome" for "factual claim about the world."

Off-chain, the regulated model resolves differently. Kalshi's contracts are self-certified with the CFTC under Rule 40.11, and resolution follows a rulesbook that specifies the designated source, the determination timing, and an internal escalation path. It is slower, more bureaucratic, and — critically — it produces a named legal entity that can be sued.

Both models answer the same question: who has the authority to declare what happened. One answers with a governance vote. One answers with a compliance department. Neither answers with cryptography, and that asymmetry is the thing the market keeps mispricing.

Now the statute.

The Commodity Exchange Act, Section 5c(c)(5)(C), carves out event contracts from the general listing regime. The Commission may decline to permit a contract that involves activity unlawful under federal or state law, involves gaming, or is otherwise contrary to the public interest. Rule 40.11 implemented that prohibition. In 2013, the D.C. Circuit vacated the Commission's interpretation of the "public interest" prong in ISDA v. CFTC, holding that the agency had read the phrase too loosely — effectively substituting its own policy preference for the statutory text. The rule survived in part; the interpretive gloss did not.

What that means operationally: the public interest prong is not a settled standard. It is an open question that gets re-litigated every time a new contract category is listed. Which means the evidence that bears on it — including polling — has legal weight without having legal finality.

In 2024, Kalshi sued the CFTC after the agency disapproved its congressional-control contracts. The district court held the Commission had exceeded its authority. The contracts went live, the D.C. Circuit affirmed in 2025, and the Commission subsequently withdrew its appeal. In parallel, state gaming regulators in New Jersey, Nevada, Maryland, and Ohio moved against event-contract platforms, arguing that state gaming law governs regardless of federal designation. Kalshi obtained injunctions in several of those actions. A preemption fight is now running on two tracks simultaneously.

That is the terrain. A one-sentence poll landed on it.

Core: The Information Audit

Before any of this becomes an argument, the source has to be audited. This is not pedantry. In a domain where a five-point swing in a poll number can move a rulemaking docket, the difference between 51% and 85% is the difference between a talking point and a statute.

| Field | Status in source | Downstream consequence of the gap | |---|---|---| | Sample size | Missing | Cannot assess statistical significance or subgroup stability | | Field dates | Missing | Cannot rule out recycled data from a prior election cycle | | Margin of error | Missing | Cannot distinguish signal from sampling noise | | Magnitude of "majority" | Missing | 51% and 85% imply materially different legislative urgency | | Questionnaire wording | Missing | Wording alone can invert the result | | Polling firm (executor) | Missing (only "Reuters" attributed) | Cannot evaluate methodology or house effects | | Party / age / trading-experience crosstabs | Missing | Cannot identify whether opposition is broad or a coalition artifact | | Industry response | Missing | Cannot determine whether counter-framing is already organized | | Named platforms or projects | Absent | No tradable instrument can be identified | | Occurrence of "crypto" or "blockchain" | Absent | Domain classification is inference, not fact from source |

The last row matters more than it looks. The source does not contain the word crypto. The classification of this story as a Web3 story is my extrapolation, and it carries medium confidence, not certainty. Anyone building a thesis on the assumption that the article was written about crypto should note that the article was not written about crypto.

The wording row matters most. Consider two instruments that measure the same underlying construct:

Q1: "Do you support allowing people to bet on election outcomes?" Q2: "Do you support allowing people to express forecasts about election outcomes with real money, in order to produce more accurate probabilities than polling?"

Q1 and Q2 will not produce the same number. The first embeds a casino frame. The second embeds an information-market frame. A well-constructed instrument would ask both and report the delta, because the delta is the entire policy question. The source reports one sentence and no wording, which means the reader cannot tell whether the reported result is a measurement of public sentiment or a measurement of the question's grammar.

That is not a minor methodological complaint. It is the difference between a regulatory input and a regulatory artifact.

My working assumption — flagged as inference, medium confidence — is that the poll was a fast-turnaround omnibus item, the kind fielded alongside horserace numbers, where a single question is slotted in and the crosstabs never leave the topline deck. The absence of a magnitude is consistent with editorial compression in a wire format. But it is equally consistent with the question never having been a priority in the instrument. Those two explanations point in opposite directions for policy probability.

Core: The Word Is the Instrument

The source uses "betting." Not trading. Not forecasting. Not price discovery. Betting.

Lexical choice is not decoration in a regulated domain. It is the load-bearing element of the frame. Under "betting," the activity maps onto state gaming statutes, which are older, more numerous, and enforced by regulators with no obligation to defer to a federal agency. Under "trading," the same activity maps onto derivatives law, where the framework is federal, preemptive in tendency, and staffed by people who understand basis risk.

The identical binary contract, described with two different nouns, lands in two different legal universes.

This is why the absence of the word crypto in the source is not an oversight. It is a tell. Prediction markets are being described in mainstream coverage as gaming products, not as cryptographic applications. That description does two things at once, and they cut in opposite directions:

It reduces regulatory hostility, because a betting product is not a securities offering and does not trigger token-distribution analysis.

It destroys the category's only non-gaming justification, because the argument that event contracts constitute information infrastructure depends on the word "information" surviving in the description. Once the frame is "betting," the epistemic value proposition has no home in the sentence.

I have watched this pattern before, from the other side. In 2021 I spent months auditing ERC-721 metadata schemas and published a gas analysis showing that a majority of top collections were overpaying for storage because of naive string handling and redundant on-chain writes — a Merkle-based schema would have cut the cost materially. The technical community read it. The market did not, because the market was not evaluating schemas; it was evaluating narratives, and my document contained no narrative. Metadata is just data waiting to be verified. The frame determines who reads the verification.

Prediction markets are in the same position now. The verification exists. The frame does not.

Core: Why Securities Law Is the Wrong Lens

The reflexive analytical move on any crypto-adjacent regulatory story is to run a Howey test. It is the wrong instrument here, and running it produces a false sense of resolution.

| Howey element | Assessment for event contracts | Residual risk | |---|---|---| | Investment of money | Present — purchase of event shares | Neutral | | Common enterprise | Weak to absent — counterparties are directly opposed, not pooled | Low | | Expectation of profit | Share price traverses 0 to 1; appreciable | Moderate | | From efforts of others | Weak — outcome is exogenous to the issuer | Low | | Composite | Low-to-moderate securities exposure | Not the binding constraint |

The binding constraint is elsewhere. Event contracts are far more likely to be characterized as swaps or commodity derivatives under CFTC jurisdiction, or as wagers under state gaming law, than as securities under SEC jurisdiction. Analysts who model this as a token-distribution problem are optimizing the wrong variable.

The real risk sits in the public interest prong and in state gaming enforcement. And here the poll becomes structurally relevant, because the public interest prong is the one place in the entire framework where public opinion is an admissible input rather than background noise.

A regulator weighing whether a contract category is contrary to the public interest has three kinds of evidence available: internal staff analysis, industry comment letters, and public sentiment data. The first two are symmetric and expensive to produce. The third is cheap, external, and quotable. A wire-service poll with a clean topline and an attribution to a major agency is the highest-leverage form that third category takes.

Which means the operative question is not "is the poll accurate." It is "is the poll citable." And a one-sentence item with no methodology disclosure is extremely citable precisely because it has no methodology disclosure — there is nothing in it to argue with.

That is a failure mode, and it is not a technical one.

Core: The Resolution Layer Is the Actual Attack Surface

Here is where I stop caring about the poll and start caring about the architecture.

Every prediction market has a resolution function R: (market, source, time) → outcome. The security property that matters is not that R is correct. It is that R is expensive to corrupt relative to the value at stake.

For most markets, that inequality holds comfortably. The value at stake in a market on a quarterly CPI print is small, the factual question is well-defined, and the cost of mounting a dispute exceeds the gain from winning it. Optimistic oracles were designed for this regime and they work well in it.

Election markets violate every assumption in that sentence.

| Parameter | Typical market | Election market | Consequence | |---|---|---|---| | Value at stake | Low to moderate | High, concentrated in a short window | Dispute economics invert | | Factual determinacy | High (a number is published) | Moderate (certification, recounts, contested slates) | Ambiguity becomes the attack surface | | Resolution latency tolerance | Hours | Days to weeks | Liveness windows become strategic | | Number of interested parties | Few | Millions, including non-traders | Governance capture incentive | | Political salience | Low | Maximum | Every dispute becomes a public event |

The last row is the one nobody models. In a normal optimistic oracle dispute, escalation to a token-holder vote is an obscure governance event that a few dozen delegates participate in. In an election market dispute, escalation is a news story. The set of actors with an incentive to influence the vote expands from "holders of the governance token" to "anyone with a political preference and a willingness to acquire the governance token."

That is not a hypothetical vulnerability class. It is a straightforward cost calculation. If a governance token has a liquid market and a dispute resolves on stake-weighted voting, then the cost of influencing a resolution is approximately the market depth required to acquire a decisive stake, minus the value recovered by selling it back afterward. For small-cap governance tokens, that number can be startlingly small.

I have seen this movie. In 2020, during the DeFi composability experiments, I built a local testnet harness to stress liquidation cascades under volatility and found a manipulation vector in an early aggregator integration: the price feed was sourced from a pool whose depth was thin enough that a borrower could move it with a fraction of the value of the position it secured. The feature was not yet public. I documented it in a forty-page writeup anyway, because the structural lesson generalized even though the specific exploit did not. The lesson was that security assumptions are denominated in the attacker's budget, not in the designer's intent.

In 2022, during the ZK winter, I spent eight months on Groth16 and built a circuit in Circom, and I found something with the same shape: a privacy pool whose entropy source was weaker than its proof system. The cryptography was immaculate. The randomness was not. A perfect proof over a predictable witness is not a proof of anything.

Resolution is the entropy source of a prediction market. It does not matter how elegant the collateral layer is, how fast the matcher is, or how tight the spread is. If the resolution function has a predictable or purchasable input, the entire market inherits that weakness, and the weakness is invisible in every metric that gets reported.

Verification is the only trustless truth. And in election markets, verification is the part that gets outsourced to a vote.

Failure Modes

I put a failure-mode table in every analysis. Here is this one.

| Failure mode | Mechanism | Observable precursor | Blast radius | |---|---|---|---| | Governance capture of dispute resolution | Acquire decisive stake in the oracle token during escalation | Abnormal token accumulation ahead of a scheduled resolution | Single market, catastrophic if unaddressed | | Liveness arbitrage | Proposer asserts a favorable outcome in the final minutes of the window | Assertions clustering near window close | Systematic, low visibility | | Ambiguity exploitation | Market rules under-specify the resolution source for contested outcomes | Rule text with undefined terms | Category-wide reputational damage | | Geo-blocking bypass | IP-based restriction defeated by a VPN | No measurable effect on reported volume from restricted regions | Compliance liability, not user harm | | Frame collapse | Election vertical removed, category re-described as gaming | Coverage vocabulary shift | Structural re-rating of the whole sector | | Poll-driven rulemaking | A single topline number enters a rulemaking record unchallenged | Absence of methodology disclosure | Federal, slow-moving, hard to reverse |

Row four deserves a sentence. Geo-blocking is the compliance strategy of choice for permissionless platforms, and it is a security theater implementation. It restricts by IP address, which is a routing artifact, not an identity. Any restriction that can be defeated by a consumer VPN is not a control; it is a declaration of intent. Regulators know this. The reason it survives is that it provides plausible deniability in the interval between the conduct and the enforcement action, and that interval is where the platform's revenue lives.

Row five is the one I would hedge a position on.

Core: Architecture Will Bend Toward Compliance

The poll, if it reflects real pressure rather than questionnaire grammar, has a predictable architectural consequence. It does not kill prediction markets. It re-shapes them.

The direction of that reshaping is already legible in the 2025-2026 buildout. Three patterns:

KYC gating at the front end. Identity verification before order entry. This is a hard architectural commitment because it converts the platform from a protocol into an intermediary. Once you know who your users are, you can be compelled to act on that knowledge, which means you inherit the full obligations of a financial institution rather than the thinner obligations of a software publisher.

Jurisdictional product slicing. The same contract, listed or not listed depending on the user's state or country. This is the direct consequence of the federal-state preemption fight: until the preemption question is resolved, a platform that wants to operate in the United States must maintain a per-jurisdiction contract registry. That is an operational cost that scales with regulatory fragmentation, and fragmentation is not shrinking.

Hybrid settlement. On-chain settlement for auditability, off-chain compliance controls for legal defensibility. The chain provides the receipt; the compliance layer provides the permission. This is not a compromise of the technology — it is a recognition that the technology was never the constraint.

| Architecture | Identity model | Settlement | Resolution | Dominant risk | |---|---|---|---|---| | Permissionless on-chain | Pseudonymous, IP-restricted | On-chain atomic | Optimistic oracle with governance escalation | Oracle capture, regulatory classification | | Regulated DCM | Full KYC | Custodial clearing | Rulesbook with named determination source | Listing approval, state preemption | | Hybrid | KYC at entry, on-chain after | On-chain atomic | Optimistic oracle, compliance override | Split-brain liability between layers |

The hybrid row is where the industry is converging, and it carries a failure mode that neither pure model has: when the compliance layer overrides the settlement layer, which one is authoritative? If an oracle finalizes an outcome that the compliance layer determines is impermissible, the contract has two truths. There is no clean resolution to that, because the whole point of on-chain settlement is that it is not overridable.

I benchmarked something adjacent in early 2026 — a hybrid rollup model where the proof verification was fast but the execution layer introduced a twelve-second finality delay under load. The interesting finding was not the delay. It was that the delay was invisible in the aggregate metrics and only appeared when you instrumented the transition function directly. Hybrid systems hide their seams. Prediction markets with a compliance override will hide the seam between legal truth and settled truth until the first contested election, at which point the seam becomes the story.

Core: Pulse Revenue and the Token That Isn't There

The source contains no token, no supply schedule, no unlock calendar, no treasury. Zero. That absence is itself a structural fact worth analyzing.

Event-contract revenue is event-driven. It concentrates in windows — a presidential election, a major sporting event, a macro shock — and decays sharply afterward. This is not a flaw in execution; it is the shape of the demand curve. People do not want to forecast the price of eggs in March with the same intensity that they want to forecast who wins in November.

The consequence for valuation is direct. A revenue stream that is a series of pulses cannot be discounted as if it were an annuity. Applying a terminal growth rate to a spike is the single most common modeling error in this vertical, and it is why peak-event volume gets cited as a valuation anchor when it should be treated as a ceiling that is touched for six weeks every four years.

| Revenue characteristic | Structural reading | Implication | |---|---|---| | Event-concentrated | Pulse, not annuity | Terminal value assumptions must be suppressed | | Sports-adjacent | Year-round but competitive | Lower margin, weaker moat | | Election-adjacent | Concentrated, high margin, high regulatory beta | Discount rate must embed policy risk | | No protocol token at the compliant leader | Deliberate regulatory strategy | Asymmetric risk transferred to tokenized competitors |

The last row is the underappreciated one. The most compliance-forward platforms in this category have conspicuously not issued tokens. That is not an oversight or a maturity signal. It is a legal position: a token adds securities-analysis exposure and distribution compliance surface to a business that is already fighting a gaming-law case. Choosing not to issue is choosing not to open a second front.

But the risk does not disappear. It redistributes. If the compliant leaders stay tokenless while smaller platforms issue governance tokens to fund development, then the entire weight of "tokenized gaming exposure" lands on the smaller platforms. The regulatory surface is not shared evenly; it is allocated by balance sheet. That produces a structural asymmetry in which the well-capitalized players absorb the market and the tokenized players absorb the enforcement.

There is no Ponzi structure here, and I want to be precise about that because the accusation gets thrown loosely. Event-contract revenue comes from trading fees and market-making spread, not from new entrants paying earlier entrants. What does exist is narrative-premium valuation: multiples anchored to peak event volume rather than steady-state revenue. That is not fraud. It is a modeling error with a marketing department.

Contrarian: The Consensus Reading Is Wrong in Three Directions

The standard interpretation of this story writes itself: bearish for prediction markets, election vertical at risk, watch for downside. That interpretation is wrong in three separable ways, and each error is instructive.

First error: treating a sentiment item as a market signal. There is no identifiable tradable instrument in the source. No platform is named. No token is named. Even if one were named, a qualitative poll with no magnitude has low pricing power — my estimate is under 20% incorporation, because markets do not reprice on adjectives without numbers. The correct expression of a view here is not shorting a specific asset. It is a relative posture: overweight compliance-clean verticals such as sports and economic-indicator contracts, underweight election-concentrated exposure. And even that posture is weakly actionable, because the two exposures usually sit inside the same platform.

Second error: running the securities analysis. As the Howey table above shows, securities exposure is low-to-moderate and is not the binding constraint. The binding constraints are the public interest prong and state gaming law. Every analyst who spends the week modeling token distribution is spending the week on the wrong variable. This is a recurring pattern in crypto coverage: the industry's own vocabulary colonizes the analysis, so a gaming-law story gets read as a token story.

Third error, and the one that matters: assuming the damage is to volume. It is not. The damage is to the category's only non-gaming justification.

The entire intellectual case for prediction markets rests on one claim: that event-contract prices aggregate dispersed information better than alternatives, and that this aggregation has public value. Elections are the single strongest demonstration case for that claim. They are high-information, widely followed, frequently polled, and — crucially — the domain where the gap between prediction-market odds and traditional polling has been most visible and most discussed. Media outlets cite prediction-market odds as a data source during election cycles. That citation is the industry's real moat. Not the code. Not the liquidity. The fact that a Reuters or a Bloomberg treats the odds as a legitimate input.

A poll showing majority opposition to election betting attacks that moat directly. It says: the public does not accept your odds as information; it accepts them as gambling. Once that framing settles, the citation stops, and the information-infrastructure story loses its evidence base.

Strip out elections and what remains? Sports. Entertainment. Cultural events. Those are real markets with real volume, and they are also — without exception — domains where licensed gaming operators already have decades of brand equity, state-level relationships, and customer habit. Competing there means competing as a betting product against betting companies, with none of the information-market premium that made the category interesting. It is a quality-for-compliance trade, and the quality being traded away is the entire thesis.

There is a fourth blind spot, less discussed. The governance mechanism that resolves disputes is a plutocracy by construction. That is fine for factual questions with low stakes. For a contested presidential certification, a stake-weighted vote among token holders is a legitimacy vacuum wearing a governance costume. If an election market ever resolves in a way that a large fraction of the public considers wrong, the reputational damage will not be recoverable through a post-mortem. It will be the thing the category is remembered for.

And a fifth: the reverse narrative is unhedged. If the next twelve months bring a favorable preemption ruling, a clear CFTC rulemaking, or a second poll showing majority support when the question is framed as forecasting rather than betting, this story becomes the "bad news exhausted" anchor — the reference point against which a re-rating happens. Nobody is building that branch into their model because it requires holding two contradictory readings of one sentence at once.

Takeaway: What to Instrument

I do not forecast prices. I instrument systems and watch for the transition where a parameter crosses a threshold.

Four things to watch, in order of signal density:

The wording of the next poll. If a follow-up instrument asks the forecasting frame and the betting frame separately, the delta is the most valuable number in the entire debate. A large delta means the opposition is grammatical, not substantive, and the policy risk is far smaller than the headline implies. A small delta means the opposition is real and structural. Everything downstream keys off that single disclosure.

The preemption docket. The federal-state question is the actual determinant of operating architecture. A preemption win collapses the jurisdictional slicing cost to near zero and permits a genuinely national product. A preemption loss locks in fragmentation permanently and makes the compliance layer the dominant cost center in the business.

Resolution governance parameters. Bond sizes, liveness windows, escalation thresholds, and — most importantly — whether the escalation mechanism can be insulated from spot-token acquisition. This is a pure engineering problem with a clean answer, and the fact that it is not being solved tells you where the industry's attention actually is. Proofs don't lobby, but they also don't vote. The gap between those two facts is the systemic risk.

The vocabulary. Watch whether coverage continues to say "betting" or begins to say "event contracts." The word is the regulatory instrument. It will move before the rules do, and it will move before the price does.

The sentence that crossed the wire had no denominator. That is the whole story. A legislative standard written to protect grain merchants from corners is now being adjudicated by a topline number with no sample size, in a market whose resolution layer can be purchased, in a category whose only defensible claim is that its prices are worth citing.

Which part of that chain, exactly, does anyone believe is secured by the code?