Every published gamma-exposure construction — including the one Cboe's own research uses — prices gamma the same way: Black-Scholes, one implied vol per contract, inverted from the quote mid. The literature offers upgrades: fit an arbitrage-free smile (SVI), or replace the lognormal with a Gram-Charlier expansion whose skew and kurtosis supposedly matter most at 0DTE. No paper measures what switching the model does to the number a terminal actually ships. So we did: the production machinery runs verbatim — same book, same dealer weights, same root finder — with only the vol input swapped, over 14 sessions including the two worst flip days in our archive.
What moves:
| swap | zero-gamma flip (median) | net gamma at spot |
|---|---|---|
| SVI slice fit | 2.5 pts | 4.4% |
| Gram-Charlier | 8.3 pts | 13% |
| higher moments alone | 3.9 pts | ~10% |
| one-tick mid ambiguity (worst case) | 0.27 pts | 0.9% |
Three readings. First, the model systematic is real: a defensible-but-different vol model moves the flip 2–3 points and aggregate gamma 5–13%. Nobody quotes that error bar; ours is now in the terminal tooltip. Second, the upgrades are not upgrades: the SVI refit makes frame-to-frame flip stability worse on 10 of 14 sessions (on 2024-08-05 the p95 frame jump goes 69 → 380 points — fit-parameter chatter is its own noise source), and half its displacement is just collapsing call and put IVs into one smile. Third, the Gram-Charlier story inverts at the aggregate: its corrections are smallest in the last 30 minutes — exactly where the theory says Black-Scholes is most wrong — because near expiry the book's gamma concentrates at the money, where every model that prices the straddle agrees.
There is no minute-level exchange truth for gamma, so none of this says which model is right — it says how much the answer depends on the choice. Per-contract Black-Scholes stays.
What changed instead: the flip's real defect was never the vol model but a discontinuous root selector, so as of this week the terminal's flip is a smoothed, tracked estimator (p95 frame jump 17.6 → 4.5 points) and ships with a crossing standard error drawn as a band. A root deserves an error bar, not a different model.
Full study: How much does the vol model move GEX?
