Combo Contracts: Why Correlated Legs Don't Simply Multiply

Same-game legs move together, so multiplying their probabilities gives the wrong price for the whole combo.

Published 2026-10-27 · Data as of 2026-10-27 · Market & data intelligence · Educational, not advice.

A combo contract bundles several outcomes into one position. A common shortcut is to price one in your head by multiplying: if each of two legs looks like a coin flip, the pair must be a quarter. That arithmetic is right only when the legs have nothing to do with each other. Inside a single game, they almost always do.

What a combo actually is

A combo is a contract between people: one side buys, another sells. According to the Kalshi Help Center, a combo lets a trader hold a custom combination of several outcomes as one position, and each combo is its own market with its own order book.

The settlement rule is strict. Kalshi's help page says a combo settles at the product of its legs' values. It pays the full $1.00 per contract only if every leg settles Yes, and $0 if any leg settles No. One miss zeroes the whole thing.

Pricing works differently from a single-outcome market too. Kalshi does not post a combo price itself. A request for quote goes out and other participants respond with prices, so a combo's price is whatever a counterparty is willing to accept to take the other side. The same page notes that once a combo is placed and filled, the user cannot cancel or reverse it.

Why multiplying fails

Multiplying probabilities assumes independence: knowing one leg came true tells you nothing about the next. Same-game legs break that assumption. If a team scores a lot, its quarterback probably threw for a lot of yards, the game probably went over its total, and the team probably won.

Delta Arc research measured how strong those links are. Using nflverse data for the 2021–26 NFL seasons, it found a player's yards were positively correlated with that player's own team's points: +0.374 for passing yards (n=2,830), +0.245 for rushing (n=3,579) and +0.191 for receiving (n=9,657). None of those numbers is huge on its own, but none is zero either, and a combo stacks them.

When legs rise together, the chance that all of them come true is higher than the simple product. So Delta Arc's pricing engine treats a multi-leg contract as one question, the probability that all of its legs are true at settlement, rather than several questions multiplied together. Because a filled combo can't be reversed, the joint outcome at the end is the only thing that counts.

Correlation can also work the other way. When two legs pull against each other, the joint probability falls below the simple product.

How the correlation gets modeled

You can't capture these links by applying a formula to each leg separately. Per its research, Delta Arc's engine prices every leg on one shared set of 500 real NFL games from 2018 onward, drawn from nflverse and chosen to sit close to the market's spread and total. Each one is a real quarter-by-quarter score path, so the legs carry over whatever relationship they had in real games.

Reversal combos make a useful test. Out of sample on 855 NFL games (2023–25, leave-season-out), Delta Arc's engine predicted 10.8% for the away team winning the first half and the home team winning the game, against 11.1% realized. For the home team winning the half and the away team winning the game, it predicted 9.6% against 10.3% realized.

What the quoted price contains

Because counterparties quote combos, the price includes whatever those participants need to take the other side, plus fees. Kalshi's fee schedule, effective July 7, 2026, lists combos (series KXMVE, excluding uncorrelated NFL combos) with a maker fee multiplier of 2 and a taker multiplier of 1, applied to its standard fee formula.

The building blocks matter as well. A Delta Arc backtest of 33,347 settled Kalshi NFL and college football markets (2026 season through October 2) found single-outcome prices well calibrated at the mid in both leagues. Kalshi's NFL winning-margin ladders, though, carried about 14–15% overround, with asks across the buckets summing to 1.14–1.15. Accurate midpoints and the cost of trading at the ask are two separate things.

Measuring combo prices directly is harder. When Delta Arc tried to compare Kalshi's 2025 same-game NFL combo prices with how they settled, the 90,685 markets in that series had only about 1 to 5 trades each. That was too few to measure pricing at all, so the test was rejected because of thin trading, not because of the prices.

The takeaway

A combo's price says how outcomes move together, not just how likely each one is alone. Multiplying the legs is a reasonable first guess for unrelated events and a misleading one for legs from the same game. Next on The Line: how request-for-quote pricing differs from a standing order book, and what that means for markets that barely trade.

Sources: Kalshi Help Center, Combos; Kalshi Fee Schedule; Delta Arc research.