Most parlays are -EV. Some aren't.

Each leg of a parlay carries the book's margin. A single bet at -110 has ~4.5% vig. A 3-leg parlay at -110 per leg compounds that — the implied probability of cashing all three is meaningfully lower than the “true” joint probability. Books love parlays because the structure favors the house.

Add the fact that most casual parlay bettors stack uncorrelated long shots (Lakers ML + Yankees ML + an over) and you have a structurally bad product. Whet doesn't dispute this — it's the math.

A correlated parlay leverages the fact that two outcomes are not independent. Tatum 30+ points + Celtics team-total over 115 are correlated (high-pace high-Tatum nights drive both). If the book prices the legs as if they were independent, the true joint probability is higher than the parlay payout implies — making it +EV.

Whet estimates how correlated the legs are — same game, same player, shared game-script — and flags the parlays where that correlation is strong enough to move the math. Most still fail the +EV bar; the few that clear it are the parlays the scanner surfaces.

A +EV parlay is still volatile — the variance scales with the number of legs. Even if the math says +EV, the optimal Kelly fraction is lower than the equivalent single-bet Kelly. Whet's smart sizing applies a variance penalty to parlay stakes that compounds with leg count.

The result: a 3-leg +EV parlay might be staked at half the Kelly fraction of an equivalent single bet. The math agrees on the edge, just not on the stake.