Kalshi's Ethereum perpetual futures contract has a favorite trade size, and it's $5,499.
A CoinDesk analysis published Monday pulled 3,450 trades in the contract, a futures product with no expiry that tracks the spot price of ether, from 23 one-hour windows between September 17 and 20. Of those, 1,406 landed within $2 of $5,499. That's $7.7 million of the $13.5 million sampled, or 57% of the dollar volume. Bitcoin perpetuals showed the same thing with different numbers: trades near $2,500 and $5,000 made up 54% of the $8.5 million sampled. Going back to June 19, CoinDesk found a recurring size on 43 of 46 sampled days, with the magic number migrating from $4,999 to $9,999 to $3,999, $4,499 and $5,499.
The fight started when Beni, a pseudonymous quant and co-founder of Stealth Neolab, posted that Kalshi's ether perp was printing $539 million of 24-hour volume against $3.1 million of open interest, a ratio of 174 to one, and called the repeating $5,500 prints "undeniable proof" of wash trading. He also waved around an Artemis chart of prediction-market share, which is a different product from the perps, a mistake Kalshi's crypto chief was happy to point out.
Kalshi's rebuttal the next day was interesting. The exchange says the fixed-size prints come from one market maker resting orders of a set size and getting picked off by "many takers." Its own numbers show the takers winning; in one hypothetical it walks through, the aggressors clear about $98,000. Self-trades are blocked at the matching engine, it runs surveillance for pre-arranged trades, and it has "seen no evidence of collusion or wash trades."
Take all of that at face value and Kalshi has still described a market maker that loses money on every fill and keeps quoting anyway. The rebuttal explains why. Its example liquidity deal pays a firm $100,000 a month to keep bids and offers of at least $5,000 resting within 0.1% of each other, 95% of every hour. The money is for being on the book. A quote of exactly $5,000, or $5,499, is what a contract written that way produces.
The Subsidy Stack
Which raises the question Kalshi's rebuttal doesn't ask of itself: why does the market maker keep doing this? The answer is that Kalshi pays it to, through three separate programs.
The resting-liquidity stipend is the first. The second is a fee rebate. A June 24 filing with the CFTC set up a temporary program that hands self-clearing members back all of their net maker and taker fees on perps each month, with two guardrails: no double-paying incentives on the same volume, and no trade may end up net-negative in fees once the maker and taker sides are combined. Free trading for the biggest accounts, but not paid trading.
The third would go further. A September 2 filing cuts the taker fee on crypto perps to 0.3 basis points, or 0.003%, and pays the maker a net rebate of the same 0.3 basis points. The filing makes that live "upon Exchange notice, but not earlier than 5:00 PM ET on September 16, 2026." Kalshi says the notice hasn't gone out and the program isn't running, so it can't be what generated the September 17-20 tape. Nobody outside Kalshi can currently check that.
Kalshi also runs a retail-facing Volume Incentive Program that splits a cashback pool by each trader's share of volume during reward periods. Market makers with existing agreements are excluded. Whether it touched the disputed markets isn't disclosed.
Put together, the stipend pays one side to post the $5,499 quote and the fee rebate makes it free for the other side to hit it. The repeating number is the incentive structure working as written.
The CFTC Wrote The Memo Five Weeks Early
On August 12 the CFTC's Division of Market Oversight put out a staff advisory, Letter 26-23, whose stated focus is "incentive programs established in connection with prediction markets." It warns that "volume-based rewards with steep tiers or threshold bonuses can encourage participants to trade solely to reach volume targets, heightening risks of wash-trading, pre-arranged trading," and, separately, that "market-maker programs that guarantee net profits or cover participant losses through stipends and rebates may incentivize artificial strategies." Exchanges are told to build surveillance around the specific behavior each program invites.
The advisory names no exchange and finds nothing against anyone. Kalshi's response, that CME, Cboe and NYSE all pay for liquidity too, is true and beside the point.
Real Trades, Fake Signal
Rajiv Sethi, an economist at Barnard College, published the cleanest account of the mechanism on Wednesday - and it doesn't necessarily point to cheating. In his reading, "the volume rewards paid by Kalshi to the market maker are flowing in part to aggressive traders who are able to move before stale orders can be cancelled." The fee refund makes it worse, because refunding the taker's fee "aggravates the adverse selection problem faced by liquidity providers." His verdict: "By channeling funds to aggressive low latency traders through the market makers, the exchange is undoing the liquidity provision that the rewards were meant to boost."