The same-game parlay is the most popular product sportsbooks have launched in the past decade, and it is also the highest-margin one. A typical SGP carries 15 to 25 percent hold, sometimes higher. That is roughly five times the tax of a regular spread bet. And yet bettors hammer them, because they look fun, they pay big, and the screen is colorful. This article unpacks how SGPs are actually priced, why correlation matters more than the legs themselves, and the narrow windows where SGPs are not just a fun stake but actually a +EV one.
What an SGP actually is
A regular parlay multiplies independent legs. If you parlay two -110 bets, the math is simple — multiply the decimal odds (1.91 × 1.91 ≈ 3.65 to 1) and that is your payout.
A same-game parlay does not multiply independent legs, because the legs aren't independent. Mahomes throwing for 350 yards is correlated with the Chiefs scoring 30+. So the book uses a custom model to price the joint probability — accounting for the correlation — and then adds the parlay vig on top. That custom model is opaque, often conservative, and the markup is hefty.
Correlation: the only thing that matters
Two legs in an SGP can have:
- Positive correlation: the legs tend to win together. Example: "QB over 280 passing yards" + "team total over 27.5". When the QB lights it up, the team usually scores.
- Negative correlation: winning one makes the other less likely. Example: "RB over 95 rush yards" + "QB over 280 pass yards" — game scripts where one happens often suppress the other.
- Independent: the legs are basically uncorrelated. Two unrelated player props from different positions.
The book's pricing engine knows this. It pads positively correlated legs more aggressively (because it would lose more often than a naive multiplier would suggest), and negatively correlated legs get less of a discount than the math says they should. The bettor's task is finding correlations the book is under-pricing.
A worked SGP example
Take three legs from a single Chiefs/Bills game:
- Chiefs ML: -150 → ~60.0% implied (devigged ~58%)
- Mahomes over 275.5 pass yards: -110 → ~52.4% implied (devigged ~50%)
- Travis Kelce over 75.5 rec yards: -115 → ~53.5% implied (devigged ~51%)
If the legs were independent, the parlay's true probability would be 0.58 × 0.50 × 0.51 = 14.8 percent, or fair odds of about +578.
But these legs are positively correlated. When Mahomes throws for 275+, Kansas City usually wins, and Kelce usually catches a lot. A reasonable correlation-adjusted true probability might be closer to 22 percent, or fair odds of about +355.
The book's SGP price comes back at +450. That looks better than the "independent" calc would suggest, so it feels like a deal. But the actual fair price is +355. The book is taking about 17 percent on this ticket, even after acknowledging the correlation. That is the SGP tax.
Where SGPs can actually have edge
Real SGP edge requires the bettor to have a better correlation read than the book. A few well-known structures:
1. Underdog moneyline + dog-favorable game script
Example: take the home dog at +160, plus the under, plus opposing QB under passing yards. If the dog wins, the game was likely low-scoring and grindy, so the under and the opposing QB under both have higher win probability than the standalone numbers suggest. The book often under-correlates these legs.
2. QB + WR1 stacks for shootouts
If your edge is on the over, parlaying the QB over passing yards with his WR1 over receiving yards gives you the shootout exposure efficiently. Books usually under-correlate WR1 with QB compared to QB-with-overall-team-score.
3. RB + team total in run-heavy game scripts
Lead-grinding teams stack RB volume and team scoring. The correlation is high but books often price these conservatively.
Where SGPs are traps
- Mass-marketed "Boost of the Day" SGPs. Designed to look good, priced for hold.
- 5+ leg SGPs. Hold compounds. A 5-leg SGP can carry 25 to 35 percent hold. You're not losing the legs, you're losing the rake.
- Negative-correlation legs. Stacking RB and QB rushing yards in the same game often drops your true win rate below the priced odds.
- "Pick Em" leg menus. You combine 6 legs out of a list — these are pure entertainment products with massive hold.
SGP versus straight props — the comparison
Suppose you like the Chiefs, Mahomes over passing yards, and Kelce over receiving yards. Two ways to bet:
- Three straight bets at 1 unit each: total stake 3 units. Win rate per leg ~52%, three independent edges.
- One SGP at 1 unit: total stake 1 unit. Higher payout if everything hits, but 17% hold means you need to be right far more often than the math suggests.
For most bettors, three straights are dramatically better long-term EV. The SGP wins when you have a genuine read on correlation that the book has under-priced. That is rare and specific. We unpack book hold across markets in our vig and hold guide.
How to evaluate an SGP before clicking
- Devig each leg individually to get a fair single-leg probability.
- Multiply for an independence baseline.
- Adjust the joint probability for correlation: positive correlation raises it, negative correlation lowers it.
- Compare to the book's offered odds.
- If the book's offered probability is more than 5 percent above your correlation-adjusted estimate, pass.
Models on Tinker output devigged single-leg probabilities, which makes step 1 straightforward. Step 3 — the correlation adjustment — is where the actual edge comes from, and it requires either a custom correlation model or experience watching how legs co-occur.
SGPs as entertainment vs as edge
It is fine to bet SGPs for fun. They are designed to be fun. Just size them like entertainment, not edge:
- Cap stake at half a unit.
- Cap legs at 3. 4+ leg SGPs are bonfires.
- Avoid boosted SGPs the book promotes.
- Treat any SGP win as luck, any loss as expected.
Bottom line
Same-game parlays are mostly a tax product. The math is genuinely complicated because legs are correlated, but the book has a better correlation model than you do unless you've built one. Stick to 2- or 3-leg SGPs with a clear correlation thesis, devig each leg, compare to a correlation-adjusted true price, and walk if the book's hold is over about 8 to 10 percent. For everything else, prefer straight bets — they win the long game.
If you want the cleanest version of correlated player exposure, our DFS lineups and props page let you stack players with explicit correlation modeling rather than a hidden book multiplier.
Bet responsibly — set limits, never chase losses.