The Lab · Reading the market · updated 2026-10-22
Same-game parlays — the correlation is real, and the book charges for it
Bets from one game win and lose together, so multiplying their odds gets the price wrong. We measured what actually moves together — and why a same-game parlay is only worth it at its fair price, which books rarely pay.
A parlay of bets from different games pays their odds multiplied together. Bets from the same game tend to win and lose together, so the book prices them its own way and keeps the difference. Some combinations really do ride together; the book knows which ones, and charges for them.
A parlay is the most popular bet there is: two or more bets, one ticket, and a payout that grows with every leg. A same-game parlay puts all the legs in one game — the Chiefs to win, the over, and Kelce over 5.5 catches. They feel natural, because those things seem to go together.
Sometimes they do. That's the whole story of how they're priced, and who keeps the difference.
A normal parlay: multiply
When the legs come from different games, they have nothing to do with each other, and the math is simple: multiply the decimal odds.
Two bets at −110 (decimal 1.909) pay 1.909 × 1.909 = 3.64, or +264. If each is truly a coin flip, the chance both hit is 50% × 50% = 25%, which would be fair at +300. The gap between +264 and +300 is the book's fee, charged on each leg and multiplied together. That's why parlays cost more than single bets: the fee compounds.
Why same-game legs break the multiplication
Take a quarterback's passing yards over and his top receiver's yards over. If the quarterback has a big day, the receiver probably did too. The chance of both is higher than the two chances multiplied — the legs are correlated.
If a book priced that by multiplying, it would pay far too much for a combination that comes in more often than the multiplication assumes. So no book does. Same-game parlays get their own price from the book's own model of what moves together, plus its margin.
What actually moves together — measured
We built a model to answer exactly this. It replays every past game of a sport — 11,700 games across the NBA, NHL and NFL, from seasons before this one — re-centred on tonight's lines, and measures which bets land together inside each way a game can go. Three findings matter for anyone building a same-game parlay.
1. Who wins and the total barely move together. Once you know the lines, the winner and the total are almost independent. The correlation between the surprise in the margin and the surprise in the total:
| Sport | Correlation |
|---|---|
| NBA | −0.005 |
| NHL | +0.023 |
| NFL | +0.015 |
Zero, to two decimal places. A parlay of "this team to win + the over" has no correlation in it to buy. If a book trims that price for correlation, it's charging you for something that isn't there.
2. A player's numbers ride with his team's score — stars much more than role players. How tightly a player's stat moves with his team scoring more than expected (a regression weight; 0 means unrelated):
| Stars | Starters | Role players | |
|---|---|---|---|
| NBA points (by minutes: 32+, 26–32, 20–26) | 0.287 | 0.198 | 0.150 |
| NHL points (by ice time: 19+, 16–19, 12–16 min) | 0.415 | 0.386 | 0.279 |
That's what makes "his team goes over and he goes over" a real combination. Inside the way a game goes where his team scores a lot, a prop's chance rises by about 1.17× for an NBA scorer's points (a coin-flip prop becomes roughly a 58% shot), 1.30× for a hockey point, and nearly 1.5× for an NFL quarterback's touchdown passes when his team wins a high-scoring game. Game script shows up too: a running back's yards rise when his team wins, even in a low-scoring game.
3. The same player's props are nearly the same bet. One player's points and his points-rebounds-assists move together at about 0.91. Stacking both buys almost nothing the first leg didn't already buy, and books price it that way.
What the correlation is worth — and who keeps it
We ran the model on 60 made-up games per sport, with lines and props shaped like real ones, and read each game all four ways. A game outcome (a winner and a total side) plus the three props that ride with it landed together 1.49× as often in the NBA as they would if the props had nothing to do with the game, and 2.22× as often in the NHL. Paid as a straight five-leg parlay — multiply the legs — that would be worth roughly +20% and +78% of every stake.
Which is exactly why you can't get straight-parlay odds on them. Same-game pricing exists to take that value back, and then add the book's margin. The correlation is real; by default, it belongs to the book.
The rule
- Work out the fair price. Estimate the chance of all the legs together — not each one multiplied — and convert it to the price that would break even. That's the most a same-game parlay can be worth.
- Bet it only at that price or better. If the book's price is shorter than fair, the correlation went to the book.
- Use fewer legs. Every leg adds the book's margin. The best combination is the one that carries the most correlation in the fewest legs — a star's over beside his team's over, say — not five legs for a bigger number.
For the numbers people: how the replay works
Each past game contributes its margin surprise and total surprise — what happened minus what its closing market expected. Re-centring those on tonight's moneyline and total cuts every replayed game into four outcomes (each team winning, over or under) so that the chance of each matches tonight's market exactly.
Each prop's chance is the market's. What the model adds is the link: a player's stat surprise, in normal scores, regressed on his team's points surprise and the margin surprise, fitted by usage tier on seasons before 2025-26 (NBA from 2024-25 box scores; NHL from 2021–25 skater logs; NFL from 2017–22). Leftover links between teammates enter through a Gaussian copula, and the chance of several legs together is summed game by game with the tetrachoric series.
The replay has no random draws, so the same lines always give the same answer. The 1.49× and 2.22× figures average 240 plans per sport (60 made-up games, each read four ways), with props and prices spread across realistic ranges and a book's margin on every leg — a description of the method, not a promise about any particular night.
What this doesn't tell you
We can't see same-game prices in our odds feed. We know what the legs together are worth; we don't yet know how often a book's same-game price beats that. We're logging it by hand, price by price, and will publish what we find — whichever way it goes.
The weights come from earlier seasons. Roles change during a season, a trade can move a player from star to role overnight, and one game is not an average.
Correlation is not a forecast. A combination that rides together still loses most of the time; a five-leg same-game parlay that lands one time in twenty is still a long shot, priced fairly or not.
And the bigger picture from everything else in this Lab: the price is usually right, and the fee is charged on every leg. A same-game parlay is the one bet where knowing what moves together can matter — but only if someone pays you for it.
Model output is informational and entertainment content, not betting or financial advice. If you bet, bet what you can afford to lose.
Log what you actually bet — units and dollars — and see it graded beside a model that publishes every call it makes, wins and losses alike. Free, no card.
Start a free tracker