Kalshi's $3bn Saturday: Long-Shot Parlays Were 38% of Volume on $4.1m of Stakes

A new daily record of $3.01 billion, another $2.82 billion on Sunday, record fees of $18.2 million and $17.6 million, and the worst two days in Kalshi's history for parlay takers, who lost $46.1 million. Strip out the parlay maker side and neither day was a record for taker volume, InGame's analysis shows.

• Kalshi set a daily volume record of $3.01 billion on Saturday 19 September, beating the $2.46 billion of 13 September, and traded $2.82 billion on Sunday, according to InGame's analysis of the exchange's trade data
• Parlay volume was $1.92 billion on Saturday and $1.76 billion on Sunday; parlays of eight or more legs priced at two cents or less were 38% of Saturday's volume, more than $1 billion, on which yes-takers staked just $4.1 million
• One ten-leg parlay at average odds of 0.3% was traded more than 40,000 times, with takers putting up $219,000 against $71.5 million of recorded volume; it did not hit
• Fees hit records of $18.2 million and $17.6 million, above July's $16.8 million, driven by maker fees introduced on most parlays last month; without them neither day would have been a record, and fees now run near 0.5% of volume against about 1% historically
• Parlay takers lost $23.2 million on Saturday and $22.9 million on Sunday, $46.1 million in total, the worst two-day period on record, and takers on other contracts lost a further $68.8 million; total volume across CFTC-regulated exchanges was almost $8 billion

The Record Is Real and the Number Behind It Is Mostly Market-Maker Liquidity

Kalshi broke its own volume record on both days of the second NFL weekend, and the composition of the record says as much about how the exchange now works as the size does. Volume reached $3.01 billion on Saturday, comfortably past the $2.46 billion of the previous Saturday, and $2.82 billion on Sunday, according to InGame, which analyses Kalshi's public trade data. Polymarket US, DKeX, Crypto.com, Novig and ProphetX all set records on Saturday and beat them on Sunday; Polymarket US exceeded half a billion dollars on both days, and total volume across CFTC-regulated exchanges for the weekend was almost $8 billion.

Parlays produced most of it. Parlay volume was $1.92 billion on Saturday and $1.76 billion on Sunday, and the two days together exceeded Kalshi's volume across every contract type in the first seven months of 2025. The mechanism is the way volume is counted: both sides of a trade are recorded, and on a long-odds parlay the maker side is professional or institutional liquidity. A $1 taker position at 0.1% odds records $1,000 of volume. "Lottery ticket" parlays of eight or more legs priced at two cents or below were 38% of Saturday's volume, more than a billion dollars, on which yes-takers staked $4.1 million. One ten-leg parlay was traded more than 40,000 times at average odds of 0.3%, with $219,000 of taker money against $71.5 million of volume. It lost.

Taker-side volume was still a weekend record, InGame found, but with a caveat of its own: cash-outs on long-odds parlays inflate taker volume too, since a $1 position cashed out at double the odds records $999 of it. Excluding taker-no parlay trades, neither day was a record for total taker volume.

Fees were a record without qualification. Kalshi collected $18.2 million on Saturday and $17.6 million on Sunday, above the previous high of $16.8 million in July, and averaged $12.9 million a day over the week, which annualises to about $1.2 billion a quarter. The records are partly the result of maker fees introduced on most parlays last month; without them neither day would have set one. Because Kalshi charges lower fees as a share of volume at very long or very short odds, fees have fallen from around 1% of volume to closer to 0.5%, and have not risen as fast as the headline number.

The takers paid for the weekend. Yes-side parlay takers lost $23.2 million on Saturday, among the worst days on record, and $22.9 million on Sunday, when an early slate that went badly for parlay bettors was only partly recovered by later games. The $46.1 million two-day loss is the worst in the exchange's history. Takers on non-parlay contracts lost $36.3 million and $32.5 million, closer to average. Rothera did not set a record, Robinhood having directed most of its volume to Kalshi and Crypto.com; CME and ForecastEx have quit sports contracts.

Volume Is Now a Marketing Number, and the Exchanges Know It

A metric on which a $1 stake can register $1,000, and on which the professional side of the book counts equally with the customer's, has stopped measuring what the word implies. Kalshi's records are true and are being reported as if they were handle, which they are not, and the forecasters trying to compare the two channels have already run into the problem. InGame's taker-only and parlay-excluded figures are the closest thing to a comparable number, and on those the weekend was busy rather than historic. The fee line is the honest measure of the business, and there the record stands, at a lower take rate on a much larger notional.

$4.1 Million Staked on Contracts Worth a Billion in Volume Is the Regulatory Exhibit

Thirty-eight per cent of a record day was eight-leg-plus parlays at two cents or less, the product every state gaming regulator would recognise as a lottery ticket and sportsbooks sell at margins they rarely disclose. The exchanges' argument that they are financial markets rests on price discovery and hedging, and a ten-leg parlay traded 40,000 times at 0.3% is neither. It is a retail product, sold to takers who lost $46 million in two days, with the exchange earning maker fees on the professional side that prices it. That is the description Missouri, Connecticut and the tribes will put in front of a court, and the trade data supplies it.

The Take Rate Is Falling Because the Product Mix Is Changing

Fees at 0.5% of volume against about 1% for most of the exchange's history is not a discount, it is a mix shift: long-odds parlays carry lower fees per dollar of notional and much higher notional per dollar of stake. The exchange makes more money in absolute terms, on a customer base that is losing more in absolute terms, while the ratio that analysts have used to value it halves. A $1.2 billion quarterly fee run-rate is the figure that matters for the $21 billion valuations being put on the sector; the volume it is drawn from is the figure that matters for the litigation.

Kalshi had its biggest weekend and its customers had their worst. Both facts come from the same trade file, and which one the reader sees depends on which column they are shown.