Kalshi Charged Its Market Makers a Fee. Its Customers Paid It, Four Times Over.

Editor's note: TrueEdge builds odds tools and earns affiliate commissions from licensed sportsbooks. This piece argues a position on prediction-market market structure, and our commercial interest runs toward the sportsbooks. Weigh it accordingly.

On August 20, Kalshi began charging market makers a fee to quote parlays. Four weeks later the exchange had collected $26 million from that fee alone, and the makers paying it had cleared $24 million in profit after paying it.

Both of those numbers come out of the same book. Somebody funded them.

Daniel O'Boyle published the reconstruction at InGame this morning, and the line in it that matters is not either of those totals. It is the markup. On uncorrelated non-NFL parlays — the category the new fee actually touches, because uncorrelated NFL-only parlays are exempt — the average gap between the price Kalshi's makers quote and the price implied by multiplying the individual legs together was 0.64% between August 6 and 19. From August 20 onward it is 2.76%.

That is a controlled experiment the exchange ran on itself. One category got a fee and one did not. The category with the fee got more than four times more expensive for the customer within days.

What an exchange is supposed to be

The entire rhetorical architecture of the prediction-market industry rests on a single structural claim: there is no house. A sportsbook takes the other side of your wager and profits when you lose. An exchange matches you against another trader and takes a cut of the match. Kalshi's federal defence, its marketing and its case to state attorneys general all lean on that distinction. It is why the product is a derivative and not a bet.

The distinction is real. It is also, on parlays, now doing far less work than it did in June.

Here is the shape of the thing as of mid-September. In the seven days to September 15, Kalshi averaged $12.3 million a day in fee revenue, against a peak of $12.9 million a day during the World Cup — a tournament, not a weekly product. Strip out the parlay maker fees and that average falls to just under $11 million. On September 12 alone, maker fees on parlays produced $1.7 million, and total maker fees hit $2.5 million, more than double the previous combined record. Parlays now generate more maker-fee revenue than every other trade type on the exchange put together.

And the volume did not flinch. On September 13, parlay volume hit $1.49 billion, the highest Kalshi has recorded. Takers staked $68.4 million against it, 11% of that day's taker-side volume.

So: the exchange added a cost, the price to the customer quadrupled on the affected category, volume set a record anyway, and the professionals absorbing the new fee finished the month $24 million up. Add Kalshi's $26 million to the makers' $24 million and the intermediation layer extracted roughly $50 million from the parlay book in four weeks.

That is a house. It has two floors.

The correction we owe

On September 16 we published a piece arguing that Kalshi's headline vig advantage over FanDuel and DraftKings evaporates on combos, and we conceded what looked like the strongest counterargument against us: that Kalshi's maker fee is a 0.0175 coefficient or nothing at all, so a patient trader can sit on the bid and pay far less than a sportsbook's hold.

That concession was already stale when we made it. Parlay maker fees had been live for four weeks. For the product that is more than half of Kalshi's volume, "or nothing at all" stopped being true on August 20, and the rate is not the 0.0175 coefficient — it is half the taker rate, which makes it the most expensive maker fee on the exchange.

For the arithmetic: Kalshi's taker fee is 0.07 × C × P × (1−P), rounded up, which tops out at $1.75 per 100 contracts at a 50-cent price. The exchange's ordinary maker fee is a quarter of that formula. Parlay makers pay half of it — twice the standard maker rate, on the one product where the customer is least likely to be comparing prices. That is not an accident of drafting.

We should have caught it. The fix is not to retract the September 16 argument, which the new data strengthens rather than undermines, but to retire the caveat we attached to it.

The counterargument, which is better than it looks

Two versions of the defence are worth stating properly.

The first is that a maker fee is a feature. Venues charge makers to discipline quote-stuffing, to fund the matching engine, and to stop the liquidity layer from being a free option. Nasdaq does not apologise for its fee schedule. If the alternative to a 2.76% markup is a market so thin that a retail parlay cannot get filled at all, the markup is the price of the product existing.

The second is the comparison. A 2.76% markup is a small number next to anything a sportsbook charges for the same shape of wager. Eilers & Krejcik published a margin table on September 17 showing that an 11-leg parlay priced at 2.5% of margin per leg carries a 24.3% expected margin, and at 4% per leg roughly 40%. Against that, 2.76% looks like charity.

The second argument is the one to be careful with, because it compares two different measurements and hopes nobody notices. O'Boyle's markup is the gap between a parlay's quoted price and the product of its legs — it prices the combination, not the legs. EKG's 24.3% is the compounded margin already sitting inside each leg. They stack; they do not substitute. A customer buying an 11-leg combo on Kalshi pays the per-leg margin and the combination markup and, on the taker side, the trading fee. Our own September 16 reconciliation of the Citizens JMP price sample found Kalshi's combos carrying an implied vig of 23.8% against 22.0% at both DraftKings and FanDuel — before fees. The exchange's combo book was already the more expensive one.

The first argument is stronger and we will concede most of it. Charging makers is defensible. What is not defensible is the marketing. An exchange that spent eighteen months telling regulators and reporters that it cannot profit from a customer's loss has now built a product where the customer's cost rose 212 basis points in a week because the venue changed its own fee schedule, and where takers are down $380 million before fees and $620 million after them since parlays launched in September 2025. The $240 million difference between those two figures is the fee load. It is not a rounding error on a peer-to-peer match. It is the business.

What we do not know

We cannot compute the pass-through precisely, and we are not going to pretend otherwise. Kalshi does not publish maker-side notional broken out by parlay category, so there is no clean way to divide the fee paid by the makers into the portion they ate and the portion they repriced. The 2.12-point markup widening and the fee are denominated differently — one against the combination's fair value, the other against contracts at a price. What the public record supports is directional and strong: the fee arrived, the price moved against the customer, and the makers' profits did not fall.

We also do not know how much of the markup shift is the fee and how much is ordinary September. Football arrives, the retail mix gets worse, and makers widen into worse flow regardless. The reason to weight the fee heavily is the exemption — uncorrelated NFL parlays, which carry no maker fee, are the category where the football-season explanation should bite hardest, and the blow-out is in the non-NFL book instead.

The product is drifting and the dates prove it

None of this happened in one move. Kalshi launched parlays in September 2025. By this April the exchange was featuring a 30-leg combo at roughly +25000 in its Popular Combos rail — Dustin Gouker bought one on April 1, watched twelve legs miss, and noted that half his 40-cent stake went straight to fees. Event Horizon's count of NFL Week 1 has 47% of parlay volume and 26% of parlay combinations coming from tickets with eleven legs or more. On September 13, two block trades put $3.7 million through a three-leg Cowboys-Raiders-Eagles combo with exactly $1 million on the yes side — the largest parlay stake in the exchange's history, almost certainly a sportsbook laying off risk, and it lost when the Giants beat Dallas.

An exchange for hedging weather and elections does not arrive at a 30-leg centipede rail by accident. It arrives there because that is where the margin is, and the margin is where it is because the customer buying an eleven-leg ticket is not price-sensitive.

The fair thing to say for Kalshi is the thing the books cannot match: you can sell your position before the game ends. That is a genuine product advantage, it has nothing to do with vig, and no sportsbook parlay offers it. It is also not what this quarter's revenue growth is built on.

Polymarket went live with parlays to all US users on the same day Kalshi started charging for them — August 20, after a beta that opened August 5, capped at ten legs, quoted through a request-for-quote system. There is now a second venue running the same playbook. Watch whether its markup follows Kalshi's, because if it does, the explanation is not one exchange's fee schedule. It is what the product is for.