Standard wisdom on parlays: they're a tax. The math is brutal — combine multiple bets and the book gets to charge vig on each one, eating any edge before you place the slip. Standard wisdom is right about uncorrelated parlays. It's wrong about correlated ones, when you know what to look for.
Why most parlays lose money
When you parlay two independent bets, the payout is the product of their individual odds minus the compounded vig. Two -110 bets parlayed = +260, but the fair-juice price is +300. You're handing the book a 10% overround on top of the vig you already paid on each leg.
For genuinely independent outcomes, parlays are -EV almost by definition. The book wins the vig stacking.
What correlation actually means
Two outcomes are correlated when one happening makes the other more (or less) likely than chance.
A few examples:
- Positive correlation: a team's quarterback going over 300 passing yards correlates with the team's total going over. The conditions that produce 300 passing yards also tend to push the game total up.
- Positive correlation: a star receiver going over receiving yards correlates with his QB going over passing yards. The throws to the receiver are the same throws contributing to the QB's yardage.
- Negative correlation: a heavy-favorite team covering a big spread correlates with the underdog quarterback going under passing yards.
How books price correlation
Sportsbooks try to model correlation in their same-game parlay (SGP) pricing. The book's engine adjusts for correlation in real time.
The book is doing this badly enough that there's edge in the corners. Specifically:
- Books over-adjust on obvious correlations the public hammers (QB-WR stacks in marquee games) — the SGP gets too expensive, killing the edge
- Books under-adjust on less-obvious correlations (a team's defensive line getting sacks correlating with the opposing QB going under passing yards) — the SGP is too cheap
Your job: find the under-adjusted ones.
Worked example
A Sunday NFL game: Ravens favored by 7 against the Bengals.
You consider an SGP:
- Lamar Jackson over 220 passing yards (modest line, his average)
- Ravens to win (favorite, moneyline -300)
In an independent world, the parlay price would be roughly the product of the two prices. But in this game, the conditions that produce Lamar over 220 yards (sustained drives, the Ravens controlling the game) also tend to mean the Ravens win. Positive correlation.
A sportsbook's SGP engine will recognize this and price the SGP slightly worse than independent — maybe at +180 instead of the +220 that independence would imply. They've taken some correlation out.
But: is +180 enough adjustment, or is the true correlation stronger than the book modeled? If your read is that Lamar over 220 + Ravens win is more correlated than the book priced, then +180 might still be +EV.
The skill is identifying when book correlation modeling is conservative — when the conditional probability of leg B given leg A is higher than what the book's SGP price implies.
Where to look for under-priced correlation
Markets where books tend to under-model correlation:
- Game-script SGPs — a side spread + the same team's totals (rushing for a run-heavy team that's favored)
- Defensive correlations — a team to win + the opposing QB under passing yards in a defense-heavy matchup
- Pace-driven correlations — a game total over + multiple players over their counting stats (rebounds in NBA, shots in NHL)
- Weather-driven correlations — outdoor MLB / NFL games with extreme weather
Where books over-model correlation
Avoid these because the book has already priced the obvious read:
- QB + their own WR over in marquee games — overadjusted
- Team to win + their top scorer to score first — overadjusted
- NBA team to win + their star over points — usually overadjusted
If everyone on Sunday is talking about an SGP, the book knows. The price reflects it.
Pitfalls
- Payout caps — books cap SGP payouts ($25K, $50K, $100K depending on book). A monster +EV SGP on a small stake is fine; a large stake gets capped.
- SGP restrictions — some leg combinations are banned by the book. Build the SGP in the slip before doing the math.
- Variance is huge — correlated SGPs hit less often than their effective probability suggests, because correlation works both ways: when the game scripts wrong, you lose multiple legs together. Stay small on stakes.
Sharp's stake rule for SGPs
Halve your normal stake size on correlated SGPs. Variance is much higher than single bets at the same EV. You'll have multi-week stretches where every SGP misses. The Kelly math accounts for it if you size conservatively.