Positioning · 10 min read

Dealer Gamma, Vanna, Charm — The Positioning Map

Market makers hedge mechanically. When you can estimate what they hold, you can estimate how their hedging will push — pinning in positive gamma, acceleration in negative gamma.

Why dealer positioning matters

Options market makers take the other side of customer flow and then hedge the greeks. That hedging is not discretionary — it is mechanical, and it acts on the underlying. When dealers are long gamma, their hedging leans against price moves (buy dips, sell rips): volatility is dampened and prices tend to pin near heavy strikes. When dealers are short gamma, hedging chases price (sell into weakness, buy into strength): moves accelerate.

GEX(K) = Σᵢ Γᵢ × OIᵢ × multiplier × S²
Gamma exposure aggregated per strike — the standard first-order positioning map.

Beyond GEX

Gamma is one row of the map. Vanna (the sensitivity of delta to volatility) links IV changes to hedge flows: when IV falls, dealers long vanna must adjust deltas, which produces flows correlated with the vol-crush rallies familiar after event days. Charm (delta decay with time) produces the slow, date-driven hedge unwind visible into big expiries. A positioning tensor — strike × expiry × {GEX, DEX, vanna, charm, OI, volume, IV} — is a far richer state description than any single 'gamma flip level'.

One honest caveat: open interest alone does not tell you which side dealers hold. Assuming every open contract is customer-bought (dealer-short) is a common but crude approximation. Transaction-classified data — such as exchange open/close volume summaries that split buys from sells and customers from market makers — materially improves the estimate. This is one of the datasets our research roadmap prioritizes.

The patterns a model can learn

The classic configurations are learnable: negative gamma + spot approaching a large put wall + accelerating put buying + steepening skew has historically preceded downside acceleration more often than baseline. Positive gamma + heavy ATM open interest + declining realized vol has favored pinning and mean reversion. A sequence model over the positioning tensor can learn these — and the many quieter configurations in between — without being told the rules.

Our approach is to let the model output probabilities (breakout, pin, acceleration) with calibration checked out-of-sample, rather than hard trade calls. Positioning is a pressure field, not a prophecy.

OptionsOceans is a live marketplace of independently tracked options opportunities, operated by WOBR AI LIMITED. A systematic research engine validates strategies, matches them to current market states, and tracks every published opportunity forward — wins and losses alike. Educational research only — not a broker and not investment advice.

Educational content only. Nothing here is financial advice. Options involve substantial risk of loss.