Does gamma survive the close?

Black–Scholes gamma is routinely treated as identifying convexity in daily option returns. We test this identification using 182 million intraday trade pairs and 0.16–2.6 million daily pairs per symbol across five underlyings (2020–2025). The correlation between empirical and theoretical gamma does not improve from intraday to daily horizons—it turns significantly negative ( − 0 . 25 to − 0 . 47 across symbols). The negative correlation becomes more severe as vega exposure grows relative to gamma exposure, and is confirmed by regressions that place no model Greeks on the right-hand side. The evidence is most consistent with substantial vega contamination through negative spot–IV comovement (the leverage-effect channel): because Δ σ is negatively correlated with Δ S , the vega component systematically loads into the gamma term. The 1 2 Γ BS ( Δ S ) 2 term can be computed but does not isolate convexity exposure.