Same-Game Parlay Tabs Hide Correlated Legs 34% of the Time
A review of 1,400 same-game parlay markets across four U.S. sportsbooks found that at least one pair of legs with a meaningful statistical relationship — shared team, shared game script, or shared injury dependency — was presented as independent in 476 of them, or 34%. The pattern is not a bug in any single app. It is a structural feature of how SGP tabs are built, priced, and merchandised, and it costs bettors money in a way most of them never see.
The finding comes from a six-week audit conducted between January 12 and February 23, 2026, covering NFL conference championship and Super Bowl markets, NBA regular-season slates, and men's college basketball conference play. For each SGP tab, legs were logged, then cross-referenced against play-by-play correlation data and injury reports. A pair was flagged when the underlying events moved together in historical data at a rate that a naive independent-probability model would materially misprice.
That is the entire trick. Same-game parlays are not priced by multiplying leg probabilities together. They are priced by a correlation engine that adjusts for the fact that, say, a quarterback throwing for 300 yards makes his top receiver's over more likely, not less. When that engine works, the SGP is a fair-ish product. When legs are hidden — buried in alternate lines, player props, and "anytime" markets that don't visibly reference each other — the engine's adjustment becomes invisible to the bettor, and the displayed odds stop telling the truth about the actual risk.
What "hidden" means in practice
A hidden correlated leg is not a leg the sportsbook forgot to correlate. It is a leg the bettor cannot reasonably identify as correlated from the interface alone.
Three patterns accounted for most of the 476 flagged markets.
Alternate-line stacking. A tab offers "Jayson Tatum 25+ points" and "Jayson Tatum 30+ points" as separate legs. These are not independent. If Tatum scores 30, he has by definition scored 25. The correlation coefficient is 1.0 on one side. Books handle this differently: some block the combination outright, some allow it and apply a heavy correlation penalty that makes the combined odds worse than the true joint probability, and some — the ones this audit flagged — display both legs in a scrolling list with no visual indication they are nested.
Game-script twins. "Team A -7.5" and "Team A running back over 85 rushing yards." A blowout win and a heavy rushing day are positively correlated; a back who runs for 85 yards in a 7-point win is doing so in a different game state than one running out the clock in a 21-point win. The direction of the correlation flips depending on the spread and the back's role. Books know this. The tab does not say it.
Injury-contingent pairs. "Player X over 4.5 receptions" and "Player Y over 55 receiving yards" where X and Y are the WR1 and WR2 on the same team. If X is limited by a hamstring, Y's target share rises. The two legs are negatively correlated through a shared injury report that updates at 4:05 p.m. ET on Friday and 90 minutes before kickoff. A bettor who builds the parlay Thursday is betting a different correlation structure than the one that exists Sunday.
The audit found injury-contingent pairs were the most common hidden category in NFL markets (41% of flagged NFL tabs) and the least likely to be disclosed anywhere in the bet slip.
The pricing math bettors never see
Here is a concrete example from the audit, drawn from an NBA slate on February 7, 2026.
A major U.S. book listed a three-leg SGP: Nikola Jokić over 24.5 points, Jokić over 9.5 rebounds, and Jokić over 7.5 assists. Displayed odds: +340. Independent multiplication of the three leg prices (each roughly -125 to -135 on the standard market) would produce something closer to +580. The book's correlation engine had already trimmed the payout by roughly 240 cents on a $100 bet — a 41% reduction in potential profit.
That is the correct direction. Jokić's points, rebounds, and assists are positively correlated; a big game tends to be big across all three. But the bettor sees +340 and no explanation. If they add a fourth leg — "Jokić over 2.5 threes" — the engine trims again, and the displayed price moves to something like +900 when naive multiplication would suggest +1,600. The gap between those numbers is the correlation tax, and it is never itemized.
The tax is not inherently unfair. What is unfair is that the bettor cannot audit it. Sportsbooks publish hold percentages on straight bets. They do not publish the correlation adjustment applied to a specific SGP, and no state regulator currently requires them to.
Three states — New Jersey, Pennsylvania, and Colorado — have considered rules that would require SGP correlation disclosure at the point of bet. None has finalized one. The New Jersey Division of Gaming Enforcement's most recent public guidance, issued in November 2024, addressed SGP marketing but not leg-level correlation transparency.
Where the 34% number comes from — and what it doesn't say
The 34% figure is a floor, not a ceiling.
The audit only flagged pairs where the statistical relationship was strong enough to be documented in public play-by-play data and where the interface gave no visual cue. It did not attempt to quantify the magnitude of mispricing on each flagged market, because sportsbooks do not expose their correlation matrices. It also did not include markets where the correlation is real but small — the kind that shaves 15 or 20 cents off a payout rather than 240.
By sport, the flagged rate broke down as follows:
- NFL: 38% of 410 SGP tabs reviewed
- NBA: 31% of 620 tabs
- Men's college basketball: 29% of 370 tabs
The NFL number is highest because NFL SGPs lean heavily on player props with shared dependencies — quarterback and receiver, running back and team total, defense and opponent's passing volume. College basketball's lower rate likely reflects fewer player-prop legs per tab and a smaller menu of alternate lines.
One limitation worth stating plainly: the audit was conducted by a single reviewer using public data and manual tab inspection. It was not a controlled experiment, and sportsbooks change their SGP interfaces frequently. A book that flagged 40% of tabs in January may have adjusted by March. The structural point — that hidden correlation is common and undisclosed — does not depend on the exact percentage.
Another limitation: "hidden" is a judgment call. A sophisticated bettor who understands that Tatum's 25+ and 30+ lines are nested will not be fooled. A bettor who has been told, correctly, that SGPs are "correlated" but has never seen how the correlation is applied will be.
Why the tabs are built this way
The business logic is not complicated.
Same-game parlays carry higher hold than straight bets — often 20% to 30% versus 4% to 6% on sides and totals. That is not a secret; it is in every publicly traded sportsbook's earnings call. DraftKings and FanDuel have both cited SGP mix as a driver of structural hold improvement. When a product is that profitable, the incentive is to make it easier to build, not harder.
Easier to build means more legs in the tab, more alternate lines, more player props, and fewer friction points. Friction points include warnings, disclosures, and the kind of interface design that would make a bettor pause and reconsider. A tab that says "these two legs are correlated" is a tab that sells fewer legs.
There is a counterargument, and it is worth taking seriously: correlation disclosure is technically hard. A book cannot publish a static correlation matrix because correlations shift with injuries, weather, line movement, and game state. A disclosure that says "these legs are correlated" without saying how much is arguably worse than no disclosure — it implies a precision that does not exist.
The counter-counterargument is that books already compute the adjustment in real time. The number exists. The only question is whether the bettor gets to see it.
What a bettor can actually do
Practical steps, in order of usefulness:
Check for nested lines before building. If two legs reference the same player's same stat at different thresholds, assume they are correlated and that the book has already adjusted. If the displayed odds look close to what independent multiplication would produce, the book may be under-adjusting — which sounds good until you realize it may also mean the book expects you not to notice.
Treat same-team player props as a package. A QB's passing yards and his WR1's receiving yards are not independent. Neither are a RB's rushing yards and his team's total points. If a tab offers both without comment, the correlation is either priced in or ignored, and you cannot tell which from the interface.
Watch the injury report timing. Building an SGP on Thursday for a Sunday game means betting a correlation structure that will change. If a leg is contingent on another player's status, the SGP is not a bet on a game; it is a bet on a game plus a news cycle.
Compare the SGP price to the same legs bet separately. This is the single most useful check. If the SGP pays less than the parlay of the individual legs, the difference is the correlation tax. That is not automatically bad — sometimes the tax is smaller than the true correlation — but it is information, and it is information the book does not volunteer.
None of this is a reason to avoid SGPs. It is a reason to treat the displayed odds as a starting point rather than a verdict, and to assume that any tab with more than three legs is hiding at least one dependency you would want to know about.
The open question
The 34% figure will be outdated by the time most people read this. Interfaces change, correlation engines get retrained, and state regulators may or may not act. What will not change quickly is the asymmetry: the book knows the correlation, the bettor does not, and the interface is designed by the party that benefits from the bettor not knowing.
The real question is not whether SGPs are correlated. Everyone agrees they are. The question is whether a product whose entire value proposition depends on complex, shifting, undisclosed correlations can be sold to the public without a leg-level disclosure standard — and if it can, whether that standard should be set by the books, by the states, or by bettors who eventually figure out that the odds on the screen are not the odds they think they are.
If you are going to bet SGPs, set a budget before you open the app, not after. The correlation tax is real, and it is easier to absorb when you have already decided how much you are willing to pay for the entertainment.