Correlated Parlays On Kalshi: The Team-To-Advance Plus Striker-To-Score Stack
The most interesting question I see in prediction market communities every tournament cycle is some version of the same stack: a team to advance paired with its star to score, bought as one combo, and the poster asking whether the payout was luck or an actual edge. Kalshi correlated parlays are the rare case where the honest answer is that it can be a real edge, because correlation between legs is one of the few forces that can push a combo's fair value above the price the market quotes. It can also be a story you told yourself. The difference is one conditional probability, and this article is about computing it instead of feeling it.
In Summary
- Independence Is The Default Math. A combo of a 60% leg and a 45% leg is worth 60% x 45% = 27% if the legs have nothing to do with each other. Fair price: 27 cents.
- Correlation Changes The Second Number. The right formula is P(A) x P(B given A). If the striker scores 55% of the time in the runs where his team advances, the stack is worth 60% x 55% = 33%, six points more than the naive product.
- That Gap Is The Entire Edge. A combo quoted near its naive product while the legs are positively correlated is priced below its true chance. No gap, no edge.
- To-Advance Already Includes Extra Time And Penalties. Both sides are shorter to advance than to win in 90 minutes. It is not a longer-priced version of the moneyline, and stacks built on the wrong read of that market start broken.
- Correlation Is Local. Same team, same run, same game state: real. Different games, different weeks: an illusion that costs you the premium you paid for it.
The Independence Math Every Combo Price Starts From
Kalshi is a CFTC-regulated event-contract exchange, so every market in a stack is a Yes/No contract that settles at $1.00 or $0.00, and its price in cents reads directly as an implied probability. That makes combo pricing unusually transparent. If two legs are independent, the chance both hit is the product of the two probabilities, and the fair price of the combo in cents is that product.
Two independent legs at 60 cents and 45 cents:
0.60 x 0.45 = 0.27, a fair combo price of 27 cents
At 27 cents the combo is roughly a +270 proposition in sportsbook language, against +203 for the 33-cent version we will get to shortly; the Kalshi odds guide covers the cents-to-American conversion if that bridge helps. Independence is the right assumption more often than stackers want to believe: outcomes from different games, different days, and different sports do not move each other, and their combo is worth exactly the multiplication, minus the fees you pay to hold it.
The whole subject of correlation stacking is about the cases where that multiplication is wrong.
Correlation: When One Leg Carries The Other
Two legs are positively correlated when one hitting raises the probability the other hits. The formula stops being P(A) x P(B) and becomes:
P(A and B) = P(A) x P(B given A)
The second term is the conditional probability, and it is where all the money hides. A striker's overall scoring rate blends every version of the tournament: the runs where his team controls games and generates chances, and the runs where it is eliminated early or plays from behind. Condition on the team advancing and you have thrown away most of the bad worlds. The scoring rate inside the surviving worlds is higher than the blended rate, and the stack's fair value rises with it.
One market-reading note before the arithmetic, because it breaks more stacks than bad math does. On Kalshi and everywhere else, the team-to-advance market settles on advancing by any route, extra time and penalty shootouts included. Both sides of that market are therefore shorter to advance than they are to win in 90 minutes, and to-advance is not a longer-priced version of the three-way moneyline. If your mental model of the leg is "win the match," you are conditioning on the wrong event, and the conditional scoring rate you attach to it will be wrong too.
Worked Example: Pricing The Advance-Plus-Score Stack
Illustrative numbers, chosen to keep the arithmetic clean; every row is the formula, not a quote from any live market.
A knockout-round setup: the team is 60% to advance. The striker's unconditional tournament scoring probability for this window is 45%. Film and minutes say that in the worlds where this team advances, it plays more attacking football and he takes the set pieces, so your conditional estimate is 55%.
| Input | Independent Read | Correlated Read |
|---|---|---|
| Team to advance, P(A) | 60% | 60% |
| Striker to score, P(B) or P(B given A) | 45% | 55% |
| Joint probability | 0.60 x 0.45 = 27% | 0.60 x 0.55 = 33% |
| Fair combo price | 27 cents | 33 cents |
If the combo is quoted near 27 cents while your honest conditional math says 33%, the gap is six points of probability you are being paid to accept. On 100 contracts at 27 cents, the cost is $27.00 plus the standard taker fee, round up of 0.07 x 100 x 0.27 x 0.73, which is $1.38, for $28.38 all-in. At a true 33%, the expected settlement value is $33.00, an expected edge of $4.62 on the position. The Kalshi fees breakdown covers the full fee curve; the shorthand here is that the fee is real but small next to a six-point probability gap, and combos, like any Kalshi order, pay it when the order executes.
Reverse the exercise before you trust it. If the quoted combo price is already at or above your correlated number, there is no edge in the stack, no matter how good the story sounds. The market may be conditioning better than you are.
Your conditional probability is only as honest as the inputs underneath it. The live odds screen is where I build mine for the sports side: the tool surfaces the de-vigged, no-vig fair probability on every market across DraftKings, FanDuel, and the rest of the major books, so each leg starts from a market-grade number instead of a fan's number. Correlation adjustments belong on top of honest inputs, never instead of them.
Where Correlation Is Real, And Where It Is Illusory
The strongest correlations are local: same team, same game, same run, same underlying cause. A team advancing and its striker scoring share machinery, because the game states that produce one produce the other. A quarterback's passing production and his top receiver's receiving production share snaps and drives. When one event is partly made of the other, conditioning moves the number a lot.
Stretch the distance between legs and the correlation dies fast. Two teams in different matches do not transmit probability to each other. A Tuesday result does not lean on a Sunday result. Stacks built across unrelated games are independence wearing a narrative, and the naive product is their true worth; pay a premium above it for imagined momentum and you have donated the difference. The candid test I use: can I name the specific shared cause, in one sentence, that makes P(B given A) different from P(B)? If the sentence is about vibes rather than mechanism, the legs are independent and I price them that way.
There is also a quieter failure mode: real correlation, overstated. Conditioning on the team advancing might honestly move a striker from 45% to 51%, not 55%. The formula does not defend you from feeding it a hopeful conditional. When my correlated read and the naive product differ by more than a few points, I make myself argue the case for the smaller gap before sizing anything, and our strategy guide is blunt about how often the market's price is the better estimate.
Fees, Sizing, And The Exit
Three practical rails keep correlation stacking from becoming an expensive hobby.
Count the fee before the edge. As computed above, the taker fee on a mid-priced combo runs near its maximum, because the P x (1-P) curve peaks toward 50 cents. A real edge survives $1.38 per hundred contracts; a two-point imagined edge mostly does not.
Size it like the longshot it is. A 33% stack loses twice for every win. That is the deal you signed. It belongs in the small, flat-capped slice of a bankroll, sized so that the expected losing streaks are boring instead of destabilizing.
Remember the position is live. A combo on an exchange is still a contract you can sell before settlement. If the first leg banks and the price of the stack rerates upward, selling into the order book is a real exit, and the hold-or-sell framework applies to a stack exactly as it applies to a single: your updated probability against the net bid, nothing else.
FAQ
What is a correlated parlay on Kalshi? A combo where one leg hitting raises the probability the other leg hits, such as a team advancing and its main striker scoring in the same run. When legs are correlated, the fair price of the combo is higher than the product of the two individual probabilities, so a combo priced off simple multiplication is priced below its true chance.
How do I price a correlated parlay? Multiply the first leg's probability by the second leg's probability conditional on the first hitting, P(A) x P(B given A), not the two standalone numbers. A 60% leg and a 45% leg multiply to 27% if independent, but if the second runs at 55% whenever the first hits, the honest joint probability is 33%.
Does the Kalshi team-to-advance market include extra time and penalties? Yes. To-advance markets settle on advancing by any route, including extra time and penalty shootouts, so both sides are shorter to advance than they are to win in 90 minutes. To-advance is not a longer-priced version of the three-way moneyline, and treating it that way misprices every stack built on it.
When is parlay correlation an illusion? When the legs live in different games or different weeks. Outcomes from unrelated matches carry no causal link, so their joint probability really is the naive product, and paying a premium for imagined correlation is donating expected value. Real correlation is concentrated in same-team, same-run stacks.
How much should I put on a correlated combo? Longshot rules apply. Even a well-built stack in the 27 to 33 cent range loses two times out of three, so it belongs in the small, flat-capped bucket of a bankroll, sized so a normal run of misses changes nothing about how you trade the next one.
Stack With A Number, Not A Story
Every correlated parlay worth making can be written as one line of arithmetic: P(A) x P(B given A), compared against the quoted price, with the fee counted and the size capped. If you cannot fill in the conditional term with something you would defend out loud, the stack is a story, and the market charges for stories. Kalshi correlated parlays reward the trader who does the boring conditioning work, prices the to-advance leg for what it actually settles on, and walks away the moment the quote already contains the correlation. That trader wins the argument either way: cheap combo, take it; fair combo, keep the powder.
Build the inputs before the stack. The odds comparison turns any price into an implied probability in one step, and the live odds screen shows the no-vig fair number on every sports market, so both legs of your next combo start from prices you can trust.
Event contracts involve risk and are not appropriate for everyone. Any probabilities discussed here are model estimates, not predictions of fact and not financial or trading advice. 18+. Availability varies by state. Trade responsibly.
