Same-game parlays hold up reasonably well at two legs, get shaky at three, and by four or five legs the pricing stops meaning much. That's the finding that keeps surfacing when you model SGP margins leg by leg: the book's edge compounds quietly rather than spiking, and the point where a bettor can no longer tell whether they're getting a fair price sits right around the third leg. It's not a scandal. It's just how correlation pricing works when the underlying model is guessing.

The margin isn't linear, and that's the whole problem

On a straight three-leg parlay across unrelated markets, the hold is predictable. Multiply the vig, account for the fact that books shade each leg slightly, and you land somewhere in the 12–18% range depending on the sport and the operator. Same-game changes the math because the legs aren't independent. A quarterback's passing yards and his team's total points move together. A striker's shots on target and his anytime goalscorer price are the same event wearing different clothes.

Books handle this by applying correlation adjustments, which is a polite term for "we think these legs are worth less combined than the naive multiplication suggests, so we're cutting the payout." When that adjustment is accurate, the SGP is priced fairly and the margin stays close to a standard parlay. When it's wrong — and it's wrong often at low-liquidity markets — the bettor is either getting a small gift or, far more commonly, paying a hidden tax they can't see.

Three legs is where the model loses its footing

The reason clarity collapses at three rather than two is combinatorial. Two legs give you one correlation to estimate. Three legs give you three pairwise correlations plus a three-way interaction. Four legs: six pairwise, four three-way, one four-way. The number of things the pricing model has to get right grows faster than the number of legs, and the data to calibrate each one gets thinner as you go.

Run the numbers on a typical NFL Sunday. A two-leg SGP on a quarterback's passing yards and his team's moneyline might carry a hold around 8–11%. Push to three legs by adding a receiver's catch total and you're often looking at 15–22%, because the model now has to price the QB-to-WR dependency on top of everything else. Add a fourth leg from a different drive script and the hold can drift past 25% — at which point you're not really betting a market, you're betting against a spreadsheet you can't see.

Where operators quietly disagree

Not every book prices the same three-leg SGP identically. On a mid-tier soccer fixture, the spread between the best and worst SGP price on identical legs has been observed at over 40% in payout terms. That's not a rounding error. That's two operators disagreeing about the same correlation, and one of them being meaningfully wrong.

What this means for anyone actually betting these

If you're going to play SGPs, the honest framing is that they're a product designed around entertainment, not edge. The two-leg version is close to a normal parlay with a small correlation tax. The three-leg version is where you should start asking what you're paying for the convenience of a single ticket. Beyond that, you're mostly buying lottery-ticket variance with a margin that no publicly available model can verify.

The open question is whether books will ever publish their correlation matrices, the way exchanges publish order books. Some sharp-friendly operators have moved toward showing implied SGP pricing alongside the naive multiplication, which lets bettors see the adjustment in real time. If that becomes standard, the three-leg cliff stops being a mystery and starts being a number you can compare. Until then, treat any SGP past three legs as a bet you can't price — and size it accordingly.