A working notebook — not a finished system — on how to tell what kind of market you're in, and what that means for scalping gamma inside a range. This is the thinking behind the Kite: the strategy we flew, then retired when the regime turned. The edge was never the scalp. It was knowing which regime we were in.
Price action is regime-dependent. The exact same tactic — fade the edges of a range and harvest gamma — is a money printer in one regime and a shredder in another. So the whole game is three steps: (1) classify the regime from features you can actually compute, (2) size the range-scalp to your confidence in that regime, and (3) stand down — or flip — the moment the features say the regime broke.
Dealers hand us the tell, because they must stay delta neutral. In a positive-gamma exposure (GEX) regime their dealer hedging sells rips and buys dips — it manufactures the very range we scalp. In negative gamma the same hedging amplifies moves and the range dissolves. Read the dealer, read the regime.
| Regime | Fingerprint | What it means for the range |
|---|---|---|
| Pinned range (positive gamma) | Spot bracketed by put wall and call wall; GEX positive & large; realized vol < implied; returns mean-reverting | The Kite's habitat. Fade the edges, harvest gamma, iron condor-style structures line up. |
| Trend (negative gamma) | Below the gamma flip level; GEX negative; dealers buy strength / sell weakness | Ranges break; fading is a bleed. Kite off. Trade with the move or stay flat. |
| Volatility expansion | IV rising, term structure backwardated, relative volume (RVOL) spiking, cross-asset correlations → 1 | Both range and naive trend fail. Defined-risk only, or flat. |
| Chop (low signal) | Low GEX magnitude; RV ≈ IV; weak autocorrelation; no dominant force | No edge to press. Minimum size or pass. |
Four families of features, from the ones with the most causal force (dealer positioning) to the ones that confirm (flow).
| Signal | Compute | Range-scalp reading |
|---|---|---|
| GEX (sign, $) | modeled net dealer gamma | + large ⇒ pin & scalp; − ⇒ stand down |
| Distance to flip (σ) | (Spot − Flip) / E[move] | comfortably above ⇒ range; near/below ⇒ danger |
| VRP | IV − RV | high & RV low ⇒ options rich, price sticky ⇒ favorable |
| Variance ratio VR(q) | Var(r_q)/(q·Var(r₁)) | < 1 reversion ⇒ scalp; > 1 trend ⇒ off |
| Autocorrelation ρ(1) | corr(rₜ, rₜ₋₁) | negative ⇒ range; positive ⇒ trend |
| OU half-life | ln2 / λ | short ⇒ tight fast scalps; long ⇒ widen or pass |
| Bollinger bandwidth | (UB − LB)/MB | low & flat ⇒ range; expanding ⇒ regime shift |
| RVOL | volume ÷ baseline | ≈ 1 ⇒ scalp; spike ⇒ exit |
| VIX term slope | front ÷ back | contango ⇒ benign; backwardation ⇒ off |
No single feature is proof. The way to use them is to normalize each into a z-score against its own recent history, sign it so "range-favorable" is positive, weight each by how much you trust it, and sum to one range-favorability score in roughly [−1, +1]:
The Kite worked because we only flew it in the pinned-range regime: steady gamma income fading the walls while dealers did the heavy lifting. It got turned off because we let it keep flying after the regime changed — GEX flipped negative, RVOL spiked, the variance ratio crossed above 1, and the fades that used to print started to bleed.
The postmortem was not "the scalp is broken." It was "we didn't respect the regime filter." That is the entire lesson of this page: the model isn't the entry. The model is the switch that says fly or fold.
Educational notes only — not investment advice, not a description of a live trading system, and nothing here is a recommendation. Options involve substantial risk.
Gamma scalping harvests the small, repeated round-trips a stock makes inside a range: you fade the edges — sell strength, buy weakness — and let dealer hedging pull price back to the middle. It only works when the market is in a pinned, positive-gamma regime; in a trending or expanding-volatility regime the same fades bleed.
You read the regime from features you can compute: the sign and size of dealer gamma exposure (GEX) and the distance to the gamma flip, realized vs implied volatility (the variance risk premium), a variance ratio or return autocorrelation to tell mean-reversion from trend, and relative volume as a tripwire. No single number is proof — you weight them into a "range-favorability" read and size to your confidence.
The Kite scalped a range and worked while the market stayed in a pinned, positive-gamma regime. It was turned off because it kept flying after the regime changed — GEX flipped negative, relative volume spiked, the variance ratio crossed above 1 — and the fades that used to print began to bleed. The lesson wasn’t that the scalp broke; it was that the regime filter has to be respected.
If forced to pick one: dealer gamma — its sign and your distance to the gamma flip. Positive, large gamma with price comfortably above the flip is the range’s habitat; crossing toward or below the flip is the danger zone. Pair it with relative volume as the fastest tripwire for a regime change.