Graveyard / Options positioning / US stocks
Trading dealer gamma exposure
Does dealer gamma exposure (GEX) tell you whether SPY will trend or stay pinned?
The idea
When options dealers are short gamma they amplify moves, and when they are long gamma they dampen them. Use the sign of the previous day's GEX to choose between following the first-hour push and fading the deviation from VWAP.
How we tested it
| Data | SqueezeMetrics daily GEX with SPY and QQQ 1-minute data. Momentum: follow a first-hour push of at least 0.20% when GEX was negative and signed volume agreed. Pinning: fade a 0.75 standard deviation VWAP gap when GEX was above its 252-day median. Switch: both rules combined by GEX sign. |
|---|---|
| Period | 2016 to 2026. |
| Costs | 3 bp per round trip. |
| Holdout | In-sample and out-of-sample for the switch. |
| Control | Random-day and coin-flip assignment controls, plus a no-GEX fade. |
Result
| Momentum, SPY | 177 trades, −3.0 bp, profit factor 0.92, Sharpe −0.13 |
|---|---|
| Momentum, QQQ | +2.2 bp, profit factor 1.05, ex-top-5 −11.8 bp |
| Pinning, SPY | 501 trades, −4.0 bp, profit factor 0.73, Sharpe −0.90 |
| Switch, SPY | −4.5 bp, Sharpe −0.89, max drawdown −42% |
| Random-day p, momentum | 0.78 (SPY) / 0.60 (QQQ) |
| Coin-flip assignment p, switch | 0.85 / 0.93 |
Why it failed
Negative-GEX days did have 2.4 times the range, but the first-hour push reversed instead of continuing (correlation −0.045), the opposite of the story. The pinning fade lost even without GEX. A QQQ switch variant with in-sample Sharpe 1.03 fell to −0.32 out of sample.
What would make it worth retesting
Only with per-strike flip-level GEX from paid options open-interest data.
Test id: gexmom-spy-event-v1, gexpin-spy-event-v1, gexswitch-spy-event-v1. Past results of a backtest say nothing certain about the future.