The two months Kalshi priced parlays closest to fair were the first two months anyone could buy one. Combos opened to everyone on 15 December. Through December and January the markup over fair value nearly vanishes, and from February it sits near 7% and hasn't left.

Buy a contract on Kalshi at 95 cents and the fee costs you 0.35% of your money. Buy one at 2 cents and it costs 6.9%. The fee is worked out from the contract itself, so the less you paid for it, the bigger a share the fee takes. Parlays are cheap by design. That's what stacking legs does to the price. The average parlay dollar goes in at 31 cents, where the fee takes 4.8%. Across every combo Kalshi has sold that comes to $140m.

The worst time to buy a parlay on Kalshi is the evening. The best is mid-morning. In July it ran 4.9% over fair value at 10am and 19.2% at 10pm. That holds when you compare like with like: inside matched cells of the same series, leg count and price band, 45 of 47 rise from morning to evening. Overnight, when almost nothing is trading, it sits around 7%.

Every leg you add to a parlay makes the price worse, and it compounds. On Kalshi a two-leg parlay costs about what its legs cost. Four legs, 2% over. Ten legs, 14% over. Twenty-five legs, 55% over. We checked 55 million parlays against the price of their own legs at the moment each one traded. Eleven months, all from Kalshi's public data.

👨🏾‍💻 We are cooking and preparing for the LAUNCH We’ll be revealing the key features of our platform FOLLOW US so you don’t miss what’s coming 👉🏼 And join the WAITLIST to get PREMIUM access at launch - waitlister.me/p/prediction-t… Prediction Markets Will Win 🏆

Steam Log In on @Polymarket 🤔 Сapturing the gamer market, nice move btw

-inf club

Be the first to try a best-in-class analytics tool for prediction markets like @Polymarket @Kalshi and others 🔥 Join waitlist today 👇👇👇 waitlister.me/p/prediction-t…

This market is trending all over the feed today and many people clearly misunderstand or don't read the resolution rules. We're experimenting with alternative ways to show and simulate resolution criteria, check this out 👇