Gambling Insider’s analysis of Kalshi ether perps found that unusually concentrated fixed-size transactions showed a much smaller pricing advantage than other trades, challenging the exchange’s explanation for the activity. An economist who advanced a similar theory says he is now reconsidering it.
A Gambling Insider investigation of 5.35 million trades found that traders responsible for unusually concentrated activity on Kalshi’s ether futures market did not consistently capture the pricing advantage expected under the exchange’s explanation for the transactions.
For two months, trades clustered around six specific dollar amounts accounted for $9.81 billion, or 55.9%, of the notional value traded in Kalshi‘s ether perpetual futures contract. Kalshi says the pattern arose when fast traders repeatedly executed against a market maker’s outdated quotes.
But Gambling Insider’s analysis found that the fixed-size traders beat the spot-implied price of ether on 33% of transactions, against a 35% benchmark for randomly selected buy-or-sell directions at the same moments and spreads. Allowing up to one second for timing differences improved their result to 39%, against a 36% chance benchmark.
Other traders performed substantially better under the same timing allowance, beating the spot-implied price on 54% of trades against a 42% benchmark.
The findings challenge the expectation that the concentrated trading was predominantly driven by participants exploiting stale quotes, although they do not establish what caused the pattern.
Rajiv Sethi, an economics professor at Barnard College who had advanced a similar explanation, told Gambling Insider after reviewing the results:
“I think only the account identifiers could settle this decisively, but I am starting to have doubts about my tolerated sniping explanation.”
The findings do not establish wash trading, in which traders trade with themselves or coordinate transactions to inflate reported volume. None of the three experts consulted by Gambling Insider concluded that wash trading had occurred. Kalshi declined to comment on the new analysis.
Six Trade Sizes Dominated Kalshi’s Ether Trading, Far Exceeding Rival Exchanges
A perpetual future, or perp, is a bet on an asset’s price with no expiry date.
On Sept. 22, The Wall Street Journal reported more than $5 billion in almost one million near-identical trades of about $5,500 in Kalshi’s ether contract. The Journal also reported that the Commodity Futures Trading Commission (CFTC) was examining the trades, citing a person familiar with the matter.
Gambling Insider expanded the analysis to every trade in Kalshi’s public data for the contract between Aug. 1 and Sept. 30. Across 5.35 million trades, transactions within $50 of a dominant trade size accounted for $9.81 billion of the contract’s $17.56 billion in notional value.
The concentration rose to 61% in September. The dominant trade size changed five times, moving through $5,000, $4,000, $4,500, $5,500, $5,425 and $4,025. Four of those changes occurred between 11:07 p.m. and 11:58 p.m. ET.
Chart 1 notes: Gambling Insider analysis of Kalshi public trade data; publication time from WSJ. Source: Gambling Insider
To establish whether this concentration was unusual, Gambling Insider asked crypto market data company Kaiko to apply a comparable test to ether perpetual futures traded on Binance and Hyperliquid.
Kaiko found that the day’s most-traded size, measured by dollar volume, accounted for 1.68% of notional value on Binance’s ETH/USDT contract, 1.61% on Binance’s ETH/USDC contract and 3.79% on Hyperliquid.
The comparable figure for Kalshi was 55.8%. Even Kalshi’s least concentrated day, at 9.3%, exceeded the highest daily concentration recorded on the three comparison contracts, which reached 3.5% on Binance and 7.8% on Hyperliquid.
“We have not seen this degree of concentration at a single trade size on either venue,” said Thomas Probst, a research analyst at Kaiko.
Probst noted that Binance and Hyperliquid are mature markets with many competing market makers quoting at different sizes. In a younger market with fewer participants, he said, “the behavior of a single large participant can show up much more strongly.”
The comparison establishes that Kalshi’s trade-size concentration was exceptional among the contracts examined. It does not establish manipulation, since a market with fewer participants could produce more concentrated trading without coordination.
Chart 2 notes: Binance and Hyperliquid figures from Kaiko tick-level data supplied to Gambling Insider. Source: Kalshi figures from Gambling Insider analysis of Kalshi public trade data (script 20_kaiko_method_concentration.py).
Kalshi Attributes Concentrated Trades to Market-Maker Incentives and Faster Traders
Kalshi attributed the trading pattern to its market-maker incentive program, which paid participants to maintain orders on the exchange’s order book.
Responding to the Journal on Sept. 22, Kalshi said the pattern was “consistent with a single maker putting up resting orders of a fixed size.”
Under this explanation, a market maker repeatedly posts orders at the minimum size required to qualify for incentive payments. When ether prices move on other exchanges, faster traders can execute against those orders before the market maker updates them. The resulting transactions would repeatedly occur at similar dollar amounts.
Kalshi said the counterparties were “hundreds of distinct traders” who were “pretty consistently right,” while the market maker was “pretty consistently wrong.” It said it had seen “no evidence of collusion or wash trades.”
Sethi subsequently developed a similar explanation, which he called “tolerated sniping.” He suggested that market makers might knowingly accept small losses on stale orders because incentive payments more than compensated them.
In emails to Gambling Insider, Sethi explained:
“The most benign explanation I can think of is tolerated sniping. When there are large differences between the perp and spot prices the market makers cancel and replace orders, when these differences are small they allow takers to make modest profits but come out ahead anyway by maintaining their commitments under the reward program.”
Both explanations imply that traders taking the fixed-size orders should benefit from favorable prices often enough to make the strategy worthwhile. Gambling Insider tested whether that advantage appeared in the trading data.
Fixed-Size Traders Captured Less Short-Term Price Movement Than Other Traders
Gambling Insider first examined how Kalshi’s estimated market price moved after each trade, using a method known as markout analysis. Traders exploiting stale quotes would generally be expected to benefit when prices subsequently moved in the direction of their transactions.
Over intervals extending to five minutes, fixed-size trades showed positive markouts, but their gains were approximately half those associated with other trades. After one minute, the fixed-size trades had generated an aggregate positive markout of approximately $106,000 on $9.81 billion in notional value, equivalent to 0.11 basis points, before fees.
Sethi said five minutes was too long: snipers’ gains should show up in seconds.
At his suggestion, Gambling Insider measured the first five seconds after every trade in 100-millisecond steps.
At every interval, fixed-size traders were less likely than other traders to have a positive markout.
A tenth of a second after trading, 17% of fixed-size trades showed a positive markout and 65% a negative one. After one second, 30% were positive and 55% negative. After five seconds, 43% were positive and 49% negative, with the remainder approximately at break-even.
In aggregate, fixed-size trades showed a negative markout of approximately $380,000 after one-tenth of a second, recovering to roughly break-even after five seconds. Other trades performed better throughout the period.
Chart 3 Notes: Both panels measure how far the estimated mid-price moved past the taker’s trade price: per trade on the left, summed in dollars on the right. Before fees, rebates, funding and incentive payments. Mid-price is estimated from the last taker buy and sell prints. Source: Gambling Insider analysis of Kalshi public trade data (script 17_markouts_short_horizon.py).
Sethi called the short-interval results “really interesting and a bit surprising.”
The results do not show that the traders lost money overall. Markouts measure subsequent price movements against execution prices, and these calculations exclude fees, rebates, funding and incentive payments. Kalshi’s public data also lacks an order book, requiring the analysis to estimate the mid-price from trade prints.
Nevertheless, the weaker short-term results prompted a more direct test of whether fixed-size traders were exploiting prices that lagged the wider ether market.
Fixed-Size Traders Showed Little Advantage Over Chance in Spot-Price Tests
Gambling Insider matched Kalshi’s trades against Binance’s public one-second ether prices to assess whether traders were executing at prices that had fallen behind the spot market.
The analysis estimated the spot-implied value of Kalshi’s contract and compared each trade with a chance benchmark. That benchmark measured how often randomly selected buy-or-sell directions would have produced a favorable result at the same moments, preserving the observed spreads and estimated differences between futures and spot prices.
Fixed-size traders beat the spot-implied price on 33% of transactions, compared with 35% under the chance benchmark.
Allowing up to one second for timing differences, a method that favors finding evidence of sniping, increased their result to 39%, against a 36% chance benchmark.
The same allowance produced a much larger advantage for other traders, who beat the spot-implied price on 54% of transactions against a 42% benchmark.
The test does not show that fixed-size takers did worse than chance. It shows that they had little or no edge where a sniper should have one. When spot prices were taken from the second before each trade, neither group stood out: one-second prices are too coarse to catch trades decided in milliseconds.
With the one-second allowance, other takers beat the spot-implied price 12 percentage points more often than chance would predict, against 3 points for fixed-size takers.
On larger pickoffs, worth more than one basis point, fixed-size takers fell below chance under both timings (7% against 10%), while other takers beat chance with the allowance (22% against 18%).
Chart 4 notes: Spot-implied price = Binance one-second close x 0.001, adjusted by the median perp-spot gap over the previous 10 minutes. Look-ahead rows use the close of the second containing the trade, which favors finding sniping. Chance level: the share expected if each taker’s direction were a coin flip, keeping the taker’s actual spread and the perp-spot gap at that moment. Perp price is estimated from the last taker buy and sell prints. Source: Gambling Insider analysis of Kalshi public trade data and Binance public data (script 19_spot_gap.py).
The nine-percentage-point difference between the groups’ advantages over chance challenges Kalshi’s explanation that the concentrated trades were primarily generated by traders exploiting stale quotes.
The analysis relies on estimated prices and one-second spot data, which cannot capture every millisecond-level trading opportunity. However, the difference between fixed-size traders and other traders persisted across 12 variations of the test, including changes to the perp-spot adjustment, estimated mid-price and timing assumptions.
Economist Questions His Own Theory After Reviewing Gambling Insider’s Findings
After reviewing the spot-price results, Sethi told Gambling Insider he was reconsidering his earlier explanation.
His reassessment could be significant because he had independently proposed a legitimate market-making mechanism that could explain the concentrated trading. It does not mean he concluded that the trades were manipulative.
Kalshi’s public data shows trades but not who made them. Sethi said he would want to know exactly how many accounts took the fixed-size trades, and whether market makers also appeared on the taking side.
His own research, he said, found that wash traders take liquidity on almost exactly half their volume, so Kalshi publishing that share for its market makers would also help.
The Journal reported that Kalshi shareholder Jump Trading and crypto trading firm Wintermute were among the parties to the trades, but not that they traded with each other. Jump told the Journal it “does not coordinate its trading with anyone.”
On Oct. 14, Gambling Insider asked Kalshi how many accounts took the fixed-size trades and whether its market makers also took liquidity. Kalshi’s head of communications, Elisabeth Diana, initially said the company would respond the following day but subsequently declined to comment. An earlier request for comment was sent on Oct. 6.
The CFTC did not respond to questions sent on Oct. 14, including whether it had requested account-level data from Kalshi. A CFTC spokeswoman told the Journal the commission “can’t comment on whether an investigation is happening or not.”
Diana told CoinDesk in September that Kalshi had not been contacted by the CFTC.
Arash Aloosh, an assistant professor of finance at DCU Business School who studies trade-size clustering, described the pattern as “largely inconsistent with typical wash-trading patterns.” He nevertheless said the trading “warrants further investigation” by regulators.
On Sept. 28, Kalshi filed to end its Volume Incentive Program, effective no earlier than Oct. 13. Gambling Insider first reported the filing on Oct. 7. The filing does not establish whether the decision was connected to the unusual trades.
The analysis establishes that fixed-size trades accounted for an exceptional share of Kalshi’s ether futures volume and that the traders responsible showed a much smaller pricing advantage than other participants. The findings challenge the exchange’s explanation, but identifying the counterparties and determining whether trading was coordinated would require account-level information that is not publicly available.
How We Did It
Gambling Insider used Kalshi’s public trade data for its KXETHPERP contract, Aug. 1 to Sept. 30, and Binance’s public one-second ETHUSDT prices.
Because Kalshi’s feed has no order book, we estimate the mid-price from trade prints. Profit figures exclude fees, rebates, funding payments and incentives.
To check whether the results depend on those assumptions, Gambling Insider reran the spot comparison 12 ways. The reruns changed the window used to adjust for the usual gap between perp and spot prices (two, 10 and 30 minutes), the mid-price estimate and the timing allowance.
In every version, other takers beat chance by about 12 percentage points whenever the one-second allowance applied. Fixed-size takers beat chance by no more than 3 points, and fell below chance on pickoffs worth more than one basis point.
One-second spot prices are not granular enough to catch sniping at millisecond speed, but that limitation applies equally to both groups, and the same test found a clear edge for other takers.
The analysis measured concentration around dominant trade sizes, short-term markouts and execution prices relative to a spot-implied benchmark. The chance comparison preserved observed trading times and estimated spreads while varying trade direction.
All scripts, results and the full trade file are available to download, with a README explaining the order in which to run the scripts.
