NFL Stack Builder
Correlated stacks for same-game parlays
A great prop isn't always a great parlay. We pair this week's picks into their highest-correlation combinations - QB+WR, game-pass stacks, blowout-flip RB+opp-WR - then compute the joint probability the right way. Independent multiplication under-counts correlation, often by 15-25pp. The numbers below use a binary-correlation joint distribution so the joint edge reflects what actually happens on game day.
How a stack score is built
- 1. Same-game picks get paired and classified - QB+WR (ρ ≈ 0.80), QB+TE (ρ ≈ 0.65), game-pass stack (ρ ≈ 0.48), RB+opp-WR script flip (ρ ≈ 0.55), and others.
- 2. Joint probability =
P(A)·P(B) + ρ·√(P(A)(1−P(A))·P(B)(1−P(B))), clamped to the achievable corridor. Independent multiplication (ρ = 0) systematically under-counts correlated stacks. - 3. Joint edge = joint model probability − joint vig-removed implied probability. Combined parlay odds = decimal product of the two legs.
- 4. Stack score layers a game-script boost: shootouts boost game-pass + bring-back stacks; blowouts boost RB+opp-WR script flips; grinds suppress both.
Correlation values are calibrated from historical same-game outcomes (2022-24) and represent the population average. Individual game variance can be larger; treat the joint edge as directional, not exact.