Parlay Anatomy
How Kalshi parlays are built and priced: the house edge by parlay length, the rise of multi-leg betting over time, which real games draw the most parlay money, and whether the implied odds match what actually happens — split by same-game (correlated) vs multi-game (independent) tickets.
House edge climbs with parlay length
Cost to the bettor per $100 staked (taker P&L, before fees — real cost is worse). Each added leg compounds the margin on same-game tickets; the multi-game series is noisier and dips at 7 legs before climbing again at 10+.
Win rate collapses with length
The rise of multi-leg betting
From a standing start in late 2025 to billions of contracts a month. Stakes is the money bettors actually paid, and a long-shot ticket is a lot of contracts and very little of it
Mean legs per parlay crept up past 6 and is still rising —
Composition on two different bases: share of volume in 4+-leg parlays, and share of tickets that are same-game (correlated). The same-game ticket share dipped sharply in Feb–Mar 2026.
Parlay volume: correlated vs non-correlated over time
Monthly parlay volume (contracts), split by leg-level correlation — same-game (correlated) tickets versus multi-game (independent) ones. Classified from the actual legs, not the ticker name; brand-new tickers still awaiting leg-mapping sit in a small "unclassified" band.
Cumulative bettor P&L: correlated vs non-correlated
How much money parlay bettors have transferred to Kalshi (and market-makers) over the period, net of fees. Classified using the same audited correlated/non-correlated split as the charts above.
How much sports dominates parlays
Almost every parlay contract is a pure-sports parlay. Each month's bar is split by mix — the green is all-sports, and the thin slivers on top are cross-category "mixed" and all-non-sports parlays.
All-sports parlays are
Mixed and non-sports parlays over time
The slice that isn't pure sports: parlays mixing sports with non-sports legs, and all-non-sports parlays. These only really appeared once Kalshi enabled cross-category combos.
The games that drive parlay money
Top 20 underlying games by parlay volume touching them. A parlay's volume is counted for every distinct game it includes, so totals are non-exclusive (a measure of how "parlayed" each game is). Raw game keys shown for non-dated/futures markets.
The most popular parlays
The 30 most-traded parlay tickets in the window you pick below, ranked by number of trades. The colored chip is the audited leg-level correlation (same classifier as the charts above) — not Kalshi's product family. Taker stakes is the money yes-takers actually put in (taker-yes dollars). Avg price is the stake-weighted price bettors paid to get in — parlays are longshots, so most sit at a few cents or less (a 1¢ ticket ≈ a 1% implied chance). Result is the settled outcome. Covers parlays with ≥100 lifetime trades; recent tickets may still be pending.
The "lottery ticket" parlays
Trading on parlays with 8 or more legs, priced under 2¢ to win — true longshots, not just "unlikely." A ticket only counts on the days it actually traded at that price, so the same parlay can appear here on one day and not another as its odds move; this is a snapshot of longshot trading activity, not a fixed list of tickets. Taker stakes (the money yes-takers put in) is shown starting June 7, 2026. Volume isn't affected and covers the full history back to Jan 2026.
Are parlays priced fairly?
Calibration: implied probability (entry price) vs the win rate that actually occurred. Points below the dashed line = bettors overpaid (actual worse than implied). Bubble size = parlay count.
Calibration gap by price band (actual − implied). Negative bars = embedded margin the bettor pays. Same-game tickets carry a wider gap in every band, and the spread widens toward the favourites.
What a parlay costs vs. multiplying its legs
A non-correlated parlay is worth exactly the product of its legs: three independent legs at 77¢, 66¢ and 74¢ are worth 37.6¢ together. These charts price every leg at the instant its parlay traded — the leg market's last print at or before that moment — and compare that product with what the parlay actually cost. The gap is what a bettor pays over assembling the same bet out of its parts, before fees. Same-game tickets are shown for contrast only: their gap holds real correlation as well as markup, which is exactly why multiplying legs is the wrong way to price them.
Each dot is one price band. Both axes are logarithmic, and both cover the whole window — the date selector above does not apply to this chart or the next. A dot on the dashed line was priced exactly at its legs' product; above it, the bettor paid more. Dot size = number of trades.
Markup by number of legs, whole window. A two-leg ticket is priced almost exactly at its parts; the gap widens with every leg added, and by the double-digit tickets the parlay costs a multiple of what its legs imply.
Daily markup, money-weighted across every priced trade that day — and the one chart in this section that follows the date selector above. On 20 August 2026 Kalshi started charging a maker fee on combos, at double its standard maker rate, and exempted combos built entirely of independent NFL legs, which trade under their own series. Switching between those two below is a sanity check rather than a controlled experiment: the exempt bucket carries well under 1% of the money, so its daily figure is volatile. What does hold up is that the step survives holding the ticket mix fixed — most of it is a repricing of 2-to-6-leg tickets, not a shift in what people were buying.
How this is measured, and what it misses
Leg prices are taken at the instant the parlay traded — the last print in that leg's own market at or before the parlay's timestamp. Both trade in the same tape, so the two are directly comparable. This is not a detail: pricing the same legs at their daily average instead makes parlays look cheaper than their own legs, because a leg's daily average absorbs moves that happened after the parlay printed, and the error compounds with every leg.
A trade counts only if every one of its legs had traded that day
before it —
Markup is money-weighted: total paid ÷ total independence-implied value − 1, over the trades in the band. It is not an average of per-trade ratios — a fifteen-leg ticket's legs multiply out to a number so small that individual ratios run into the millions and any average of them is meaningless.
Below about 0.1¢ the price stops being a probability. The traded price flattens out down there while the product of the legs keeps falling, which is what drives the enormous markups at the cheap end of the first chart. Read that band as a minimum-price effect, not as a judgement about those parlays' odds.
Window. From June 25, 2026 — the first day our leg snapshot covers the tickets that traded, and safely after the June 7 switch to sub-cent price collection, before which a cheap parlay's price was rounded to whole cents and this comparison would be noise. Prices are yes-side taker trades only, which for parlays is nearly all of the flow: they are quoted on request, so the customer is the yes buyer.
About this page & method
Source. The parlay market collector — every Kalshi parlay
market (KXMVE* series) that has traded, from launch (Sep 2025)
through the latest collected day. ~
Same-game vs multi-game. Each leg's underlying game is
derived from its event_ticker. Legs of one game listed under
different market types (moneyline / total / spread) collapse to one game, so
a ticket is same-game (correlated) if any two legs share a game —
meaning it cannot be priced by simply multiplying leg odds. Per-asset crypto
windows and season-long futures are kept distinct (not false-merged).
House edge = −(taker P&L ÷ stake), before Kalshi fees; the real cost to the bettor is somewhat larger. Calibration gap is actual minus implied win rate; the negative gap is the all-in cost the bettor pays (margin/vig), not pure forecast error.
"Lottery ticket" parlays. Qualifies per ticket per day (8+ legs and that day's own volume-weighted yes price under 2¢) — the same ticket can qualify on one day and not another as its price moves, so this is longshot trading activity over time, not a fixed watchlist. Taker stakes starts June 7, 2026: before that date our own price data was rounded to whole cents at collection, which would badly understate stakes for exactly this cheapest slice (a true 0.3¢ leg would show as an exact $0 stake) — Kalshi's own market has always supported sub-cent pricing, this is our data catching up, not a market change. Volume isn't price-derived so it's shown for the full history, and includes contracts on both sides of the market (yes and no) plus a small share (~9%) with no attributed taker side, matching how "Volume" is defined elsewhere on this page. Because the qualifying-day price check itself uses whatever precision was available that day, a small number of tickets right at the 2¢ edge in the pre-June-7 volume figures may be mis-bucketed by rounding — a minor effect on the edge cases, not the same understatement that affects stakes broadly.