Often, we overlook the most important part of an accusation: who made it. Over the past several days, a claim has circulated that Kalshi — a CFTC-licensed designated contract market that has moved into crypto derivatives — inflated the reported volume of its ETH perpetual futures through wash trading. The claim arrived unattributed. No regulator has confirmed an inquiry. No order book has been published. No named employee, competing market maker, or analytics firm has stepped forward to own the evidence. Kalshi has not been reported to have offered a substantive public rebuttal either.
That missing attribution is not a footnote. It is the entire analytical frame. In the years I have spent auditing contracts that hold other people's money, the disclosures that frightened me most were never the ones accompanied by proofs — they were the ones that arrived with none, because an untestable claim can only be believed or disbelieved, never verified. A technical institution should be equally uncomfortable with an unverifiable accusation and an unverifiable denial. That is the condition Kalshi now occupies, and it is a genuinely difficult one, because it forecloses the one response that would end the matter: evidence.
To understand why, we have to start with what Kalshi actually is, because it is not a DeFi protocol and never claimed to be. Kalshi operates as a designated contract market under the U.S. Commodity Exchange Act, authorized by the CFTC to list standardized derivative contracts. Its architecture is a centralized matching engine and clearinghouse — the same basic infrastructure as CME, not the same as dYdX. Users do not custody keys; they hold balances with a regulated intermediary, and the exchange maintains the authoritative record of who traded what, when, and at what price.
ETH perpetual futures are the most heavily traded crypto derivative category in the world — a contract with no expiry that tracks spot price through a periodically settled funding rate. Because there is no expiry, the venue's reported volume becomes the primary signal of its health. Volume attracts market makers. Market makers provide depth. Depth produces tighter spreads. Tighter spreads attract volume. The loop is self-reinforcing, and it is precisely this loop that makes volume the most manipulated number in all of finance.
In the United States, that manipulation has a specific legal name. Under Section 4c(a) of the Commodity Exchange Act, wash trading — transactions in which the same beneficial owner takes both sides, or where parties collude to execute a trade with no change in beneficial ownership — is expressly prohibited, and the CFTC holds direct enforcement authority over it. This is not a grey zone. It is a bright-line rule, and one of the oldest in American derivatives law, because the practice predates electronic markets by a century.
So the analytical question is not whether wash trading is wrong. It is whether a licensed venue can be clean in a way that outsiders can independently confirm. And here the answer is uncomfortable for the entire regulated-crypto thesis: a license is a permission, not a proof.
Consider the mechanical difference. On a transparent on-chain venue, every order, cancellation, fill, and liquidation is a public state transition. Anyone can reconstruct the order book, cluster addresses, identify wallets that repeatedly cross with themselves, and measure the fraction of volume that serves no economic purpose beyond printing a tape. The analysis is tedious, but it is not privileged. It requires nothing more than an archive node and patience.
A centralized DCM works the other way. Its matching engine and its ledger are, by design, non-public. The venue maintains the authoritative record; the public receives a summary. That summary is generated by internal systems no external party can audit, and it is validated, if at all, by the venue's own market surveillance function. This is not a criticism of Kalshi specifically — CME and every other regulated venue operate the same way. It is a structural property: in a centralized matching engine, the only people who can verify the tape are the people who produce it.
That asymmetry explains why wash trading allegations against centralized venues have such a distinctive life cycle. They rarely resolve quickly, because the evidence required to resolve them — the full order-level record, the account-to-beneficiary mapping, the timing correlations between offsetting prints — sits inside the institution. External data vendors see only aggregates. Regulators can compel discovery, but investigations are slow, and the venue has every incentive to characterize the matter as a misunderstanding until compelled otherwise.
What should a research analyst look for, then, when a claim like this lands? Not the accusation — the incentives underneath it.
In my first serious audit — six unpaid months spent working through the MakerDAO stablecoin generation logic remotely from Shenzhen in 2018 — I found three race conditions in the liquidation engine that could have drained collateral during volatility. What that experience taught me was not that the code was badly written. It was that the failure lived in the defaults the developers chose, not in their intent. The system behaved exactly as specified. The specification was the problem.
The same logic applies to volume. A trading venue does not need bad actors to produce a distorted tape. It only needs an incentive structure in which volume is rewarded. New venues chasing liquidity commonly design volume-linked maker rebates, tiered fee schedules keyed to notional, points programs with implied future value, and ranking placements sold on reported turnover. Each of those is a normal, defensible business decision. Together, they constitute a standing economic incentive to trade with yourself — because if a rebate exceeds the cost of a round trip, a rational participant will round trip until the incentive is exhausted.
That is the sentence worth reading twice. When a venue pays for volume, it is not creating a moral hazard; it is creating an arbitrage. And arbitrage is executed, reliably, by whoever finds it first, without malice and without a memo.
Self-trading detection, properly implemented, is not exotic. It requires account-level attribution — mapping every account to its beneficial owner — plus timing analysis to catch offsetting orders entered from related entities within milliseconds, plus a review of prints that show no meaningful price discovery. Exchanges do this. The question is whether the thresholds are calibrated to catch deliberate volume manufacture, or calibrated only to catch the crude cases that would embarrass the venue during an examination. A surveillance system that flags obvious self-matches but ignores offsetting flow between two accounts under common control is technically present and practically absent. Compliance that exists on paper and not in calibration is a category of risk regulators have historically punished precisely because it stays invisible until it fails.
If I were handed this assignment as an analyst rather than a commentator, I would start with a methodology rather than a verdict. I would reconstruct the ratio of cancellations to fills at account level, because manufactured volume leaves a distinctive footprint in how orders die. I would look for intraday clustering around round notional figures, which is where automated incentive farming tends to land. And I would test whether volume spikes correlate with the reset schedule of any rebate or points program, because a genuine liquidity response and an engineered one look identical in aggregate and completely different at the calendar boundary. None of that requires privileged access. It requires knowing what to count.
From an audit perspective, there is one design question I would ask of any venue whose volume is under question: what exactly is the incentive schedule for market makers, and does any component of it reference notional volume rather than resting liquidity or effective spread? Resting liquidity and time-weighted spread are hard to fake — to earn them, you have to actually post quotes and hold them. Notional volume is trivial to fake. A venue that rewards the first and ignores the second has a structural defense it can point to. A venue that rewards the second has a structural vulnerability it cannot argue away, only remediate.
There is a second-order problem that receives far less attention than the accusation itself, and it concerns me more as an analyst: data contamination is an externality, and the venue is not the only party that pays for it. Reported exchange volume does not stay inside the exchange. It flows into aggregators, into index construction, into derivative pricing references, into the liquidity and notional statistics that researchers, ETF sponsors, and institutional allocators use to size a market. If a meaningful slice of a venue's ETH perpetual volume is not economically real, that distortion propagates into every metric built on top of it.
I have watched a version of this before. In 2022, while the market convulsed through the collapse of an algorithmic stablecoin, I spent weeks tracing the oracle feedback loops that turned a depeg into a death spiral. The lesson from that forensics exercise was not that the mechanism was unpredictable — the failure mode was legible to anyone who read the redemption logic carefully. The lesson was that the damage was never confined to the protocol's own balance sheet; it propagated through every venue, lending market, and index that had accepted the protocol's prices as an input.
The parallel here is not severity. A questionable volume print is not a systemic solvency event. The parallel is the propagation path. A contaminated volume figure is an input, and inputs are trusted. Once a number enters a widely used dataset, it acquires the authority of consensus, and correcting it later is slow, contested work. Analysts who build on that number inherit an error they cannot see.
For an ordinary trader, the cost of this uncertainty is not abstract. It shows up in three places. In execution, a trader routing a large order into what appears to be deep liquidity, based on reported volume, may discover that the depth was illusory at the moment of execution and pay slippage the tape never predicted. In funding, perpetual rates are set by the balance of long and short open interest, and if a portion of that open interest is artificial, the rate becomes a distorted signal that every real position pays for. In benchmark selection, an institution choosing where to hedge is choosing where its execution risk lives — and it is choosing on statistics it cannot independently audit. None of this requires the underlying allegation to be true in order to be a genuine problem. It is the cost of the uncertainty itself.
Here is where I depart from most of the coverage. The instinctive reading of a story like this is that it tests one company's honesty. I think the more consequential reading is that it tests a narrative — specifically, the assumption that regulatory permission functions as a proxy for data verifiability.
Those are two different properties, and the industry has been sloppy about conflating them. A DCM license means the CFTC has reviewed a venue's rulebook, capital structure, and surveillance obligations. It does not mean the venue's reported numbers have been independently reconstructed by a third party. Regulated venues are accountable to a supervisor; they are not, in any ordinary sense, verifiable by the public. Tracing the hidden vulnerabilities in the code that constitutes market surveillance means being precise about which of those two guarantees you actually hold.
The converse blind spot is equally real and less popular to state. An unattributed allegation is not evidence. In a market where an unverified claim can move a competitor's valuation, the absence of a named source and a described evidentiary method is itself a signal worth weighing. Both hypotheses — a genuine surveillance failure, and a weaponized narrative — remain live, and the disciplined position is to hold both open rather than selecting the flattering one. Building trust through rigorous, unseen diligence means resisting that selection.
There is one more layer worth naming. The regulated-crypto trade rests on an exchange: accept stricter supervision, disclosure, and capital rules, and receive institutional credibility that offshore venues cannot match. That trade only holds if the supervision produces a measurably different outcome. If a licensed venue's volume is later shown to be no more reliable than an offshore venue's, the premium evaporates — and the capital does not simply stay put. It reallocates toward venues where the tape can be independently reconstructed, which today means on-chain perpetuals with public order flow. That migration is already underway for other reasons. A credibility event at a licensed venue would accelerate it.
The resolution here will not come from commentary. It will come from three specific developments: whether Kalshi publishes an order-level rebuttal or a description of its market surveillance methodology; whether the CFTC opens a formal inquiry; and whether the identity of the accuser is ever established. Quietly securing the layers beneath the hype sometimes means being willing to do nothing until those signals arrive. But if a venue can be licensed and still unverifiable, we should ask what the certification was ever worth.