Extremely long parlays, which have become an ever-growing part of Kalshi’s activity, are creating by far the biggest losses for Kalshi takers, with the prices of the longest combinations more or less completely detached from their probability.
That the biggest market maker edge happens on the longest parlays may not be a surprise, but the extent of the difference is stark.
Users taking the yes side on the longest parlays, those priced below one cent, have lost well over half their stakes. Thousands of multi-game Kalshi parlays priced at around 0.1 cents, implying a one-in-a-thousand chance of success, likely had less than a one-in-a-billion chance of hitting, based on the odds of their legs. A parlay of five underdogs to win, each priced at 20% odds, is likely to be placed at almost the same odds as a parlay of 15 or even 50 underdogs.
Kalshi's parlay price floor
Since Sept. 3, Kalshi has allowed a minimum price of 0.01 cents on parlays, the equivalent of +1000000 odds. However, once parlays get below the 0.1 cent mark, there is a marked change in what it takes to move the price further down. In practice, 0.1 cents often remains the floor price for a parlay. Many parlays of 20 or more legs, each with a slim chance of winning, are still priced at roughly 0.1 cents, suggesting that takers of the longest parlays display virtually zero price sensitivity.
On Kalshi’s website and app, users can only take the “yes” side of a parlay and cannot act as the market maker, meaning that parlay takers are overwhelmingly recreational users. Users of the Kalshi API can act as the market maker. Some of these users are smaller-time “garage-band” market makers, while others are large institutions such as sportsbooks or Wall Street funds.
Looking at leg prices
For multi-game parlays, where the odds of independent legs are not likely to be correlated, an expected price can be calculated by multiplying the odds of the parlay’s constituent legs. Doing so reveals that multi-game parlays priced above 6 cents more or less correspond to the prices of all their legs hitting.
However, at longer odds, the parlay price starts to diverge from the price of its legs, and below 0.1 cents this divergence becomes dramatic. Eventually, the real price of parlays hits a wall close to 0.1 cents, while the expected prices continue to get longer.
There likely is some level of correlation between legs in many of the longest combinations. Many of these bets are parlays of college football underdogs, and especially early in the season it is likely that the performances of these underdogs could be correlated.
However, the difference between actual and expected odds goes well beyond what could be explained by this kind of correlation. Even if a correlation between underdog performances is assumed, the longest parlays are still priced at dramatically shorter odds than would be expected from their legs.
At their most extreme, some parlay bettors have taken odds of around 1,000-to-1 on events where multiplying the legs would suggest odds of all legs hitting would be as long as one in 10,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000.
Even if single-sport parlays are excluded, there have been several ultra-long parlays of events across 10 or more different competitions still priced at odds of around 0.1 cents. Being spread across different competitions, correlation in these parlays should be dramatically lower.
No parlay with leg-implied probability longer than one in 28,000 has won. The longest winning parlay by leg-implied probability, priced at 0.08 cents, was made up of 15 legs of soccer and baseball, across nine different competitions.
Long-odds parlays are big share of volume
Because of how prediction markets count volume, including the amount staked by both the taker and the market maker, these very long parlay trades make up a particularly high portion of Kalshi’s daily volume figures.
Looking only at the amount staked by takers on the yes side of parlays — a closer equivalent to sportsbook handle — they represent a much smaller share of Kalshi’s activity. A little over $112.5 million has been staked on parlays at 1 cent or less since June. These trades make up $37.2 billion in volume, or 60% of Kalshi’s volume over that time period.
Due to how Kalshi’s fee structure works — with higher fees for trades close to 50/50 odds and lower fees for big favorites or longshots — these trades have not been a particularly high share of Kalshi’s fee revenue either. The business has made $11.7 million in fees on these trades, or about 1% of its total fee revenue over that time period. Instead, the main financial beneficiaries are market makers.
Analysts find sportsbook combo prices beat Kalshi
In a note to clients sent this Monday, Citizens analysts Jordan Bender and Isabelle Slavin noted that Kalshi's pricing on the combos it tracked, made up of 15 legs and therefore likely to be at long odds, was worse than that of sportsbooks. In contrast, Bender and Slavin have found that Kalshi's prices beat sportsbooks for non-combo bets.
"We tracked combos on Kalshi by taking the average odds and implied vig for 1) the over, and 2) the favorite across every NFL game, representing 15 data points," the note said. "Kalshi’s implied vig was 26.4%, 11% higher than the 23.9% implied vig at DraftKings and 6% higher than FanDuel, before accounting for transaction fees.
"Kalshi pricing last season was worse on average, and we are watching to see if Kalshi's implementation of a combo market making fee in August will impact spreads (and the vig) for market makers. Kalshi combo volume has seen no impact from a volume perspective to this point, as combos have increased to $26bn in the last 30 days alone, or 53% of its total volume, compared to its 90-day average of 44%."
