Skip to content
Financial Context Wiki

Dealer Gamma Exposure: Units, Inventory Assumptions, and Hedge Direction

Estimate dealer gamma exposure with explicit position signs and units, separate open interest from dealer inventory, and test hedge-flow claims without treating GEX as a price signal.

Updated

For educational purposes only; not investment advice. Investing may result in loss.

Direct answer

Dealer gamma exposure (GEX) is a modelled estimate of how the aggregate option Delta attributed to dealers changes when an underlying price moves. If a dealer is locally Delta-neutral, positive position Gamma implies selling more underlying after a rise and buying after a fall; negative position Gamma implies buying after a rise and selling after a fall.

Public open interest does not identify dealer inventory. Every open contract has one long and one short side, while public OI does not reveal account type, opening intraday flow, stock or futures hedges, other options, or OTC offsets. A dashboard must therefore combine observable contracts and model Greeks with unobservable dealer-position assumptions. Its output is a conditional sensitivity, not an order forecast, support level, or volatility guarantee.

Build an auditable estimate

Let q_i be signed dealer option contracts for series i, positive when dealers are net long and negative when they are net short; M_i is the compatible premium or deliverable multiplier; Γ_i = ∂Δ_i/∂S is Gamma per underlying price unit. Use one timestamp and one coherent model coordinate.

  1. Lock every claim: underlying or futures contract, option root, call or put, strike, expiration, exercise style, settlement, currency, multiplier, deliverable, signed quantity, and data timestamp.
  2. Lock the model state: spot or forward, volatility surface, rates, dividends, borrow, time to exact expiration, and the Gamma convention. Do not net outputs produced with incompatible coordinates or units.
  3. Separate observed OI from assumed inventory. A transparent estimator can write q_hat_i = a_i × OI_i, where a_i ∈ [-1, +1] is an explicit assumed signed dealer share, not an observed fact.
  4. Calculate share-equivalent position Gamma as G_share = Σ(q_hat_i × M_i × Γ_i). Its unit is Delta-equivalent underlying units per one underlying price unit.
  5. For a small move ΔS, estimate dealer option-Delta change as ΔDelta_options ≈ G_share × ΔS. A locally Delta-neutral hedge changes by ΔH ≈ -G_share × ΔS underlying units.
  6. If a dollar-notional convention is useful, define it rather than naming it merely GEX: G_dollar_1% = 0.01 × S² × G_share. It is the spot value of the local hedge change for a 1% move, not dollars that must trade.
  7. Reprice across spot, time and surface scenarios, locate any zero of G_share(S) only inside that model, then compare inferred flows with later position data, futures or stock volume, liquidity and realized hedging evidence.

The hedge-direction statement assumes the dealer actually targets Delta neutrality and uses the named underlying instrument. Hedge bands, inventory netting, customer facilitation, transaction costs, options or futures hedges, discretion and market impact can change both timing and size. For a finite move, full repricing or ∫ Γ(S) dS controls; initial Gamma times a large move is only a local approximation.

Worked examples

  • One series, opposite assumptions. Let S = $100, Γ = 0.02 Delta per $1, OI = 10,000, M = 100, and assume a = -0.60. Then q_hat = -6,000, G_share = -6,000 × 100 × 0.02 = -12,000 shares per $1, and G_dollar_1% = 0.01 × 100² × (-12,000) = -$1,200,000. For a rise of $2, ΔH ≈ -(-12,000) × 2 = +24,000 shares, a local buy estimate. If a = +0.60, the same public OI instead implies -24,000 shares of hedge change. OI cannot select the sign.
  • A gamma-flip root moves with inputs. At S = $100, suppose dealers are estimated long 4,000 contracts with Γ_1 = 0.018 and short 6,000 contracts with Γ_2 = 0.012, each with M = 100. Contributions are +7,200 and -7,200 shares per $1, so G_share = 0. At S = $105, full repricing gives Γ_1 = 0.010 and Γ_2 = 0.017; then G_share = +4,000 - 10,200 = -6,200 shares per $1. A further $1 rise locally implies buying 6,200 shares. The zero at 100 was a root of that snapshot, not a permanent price wall.
  • Volume is not current dealer inventory. Suppose yesterday’s OI = 50,000, today’s volume is 80,000, and next published OI is 52,000. With a = -0.40, Γ = 0.004, M = 100, and S = 5,000, the old estimate is G_share = -8,000 units per point and G_dollar_1% = -$2,000,000,000; the next-OI estimate is -8,320 units per point and -$2,080,000,000. Today’s 80,000 contracts cannot be substituted for signed dealer quantity, and neither notional says two billion dollars actually traded.
  • Local sensitivity versus full repricing. At S = $500, an estimated portfolio has G_share = +3,000 shares per $1. A 1% rise is $5, so the local hedge estimate is -15,000 shares, worth $7,500,000 at the starting spot. Full repricing over 500 → 505 instead finds option Delta increased by 12,000 shares, equivalent to average Gamma 2,400 shares per $1; the offsetting sale is 12,000 shares, worth $6,060,000 at 505. A dealer using a 50% hedge adjustment would sell only 6,000 shares. Sensitivity, full-reprice hedge need and actual execution are three different records.

Research checklist and limitations

  • A wrong root, underlying, expiration, multiplier or adjusted deliverable invalidates the aggregation.
  • Call and put Gamma may be comparable only under the same model coordinates and exact contract inputs.
  • Signed dealer quantity is generally inferred; public OI supplies no dealer or customer identity.
  • Every OI contract contains both a long and short side, so OI itself has no Gamma sign.
  • OI is processed after opening and closing activity and cannot describe same-session inventory in real time.
  • Volume counts trades, not outstanding signed positions, and cannot replace OI or dealer inventory.
  • Trade-classification rules can mislabel customer direction, opening versus closing, spreads and transfers.
  • Stock, ETF, futures, index-option, cross-expiry and OTC offsets may be absent or double-counted.
  • Spot, forward and futures Gamma require compatible hedge instruments, price units and contract point values.
  • Per-share, per-index-point, per-contract and dollar-notional units can differ by factors of M, S or 0.01 × S.
  • Vendors may reverse the sign, report customer rather than dealer Gamma, or display absolute exposure.
  • Volatility surfaces, dividends, borrow, rates and exact timestamps change model Gamma and root locations.
  • A single-point Greek is local; gaps, skew repricing and large moves require full portfolio repricing.
  • Near expiration, high local Gamma can change sharply while contracts close, exercise or settle.
  • A modelled gamma flip can move or disappear and is not contractual support or resistance.
  • Delta-neutrality targets, hedge bands and rebalance frequency are dealer-specific and unobservable.
  • Hedging may use futures, ETFs, options or internal inventory rather than immediate underlying trades.
  • Bid-ask spreads, liquidity withdrawal, halts and market impact can dominate the theoretical hedge direction.
  • News, systematic funds, closing auctions and other participants can outweigh dealer hedge flows.
  • Backtests face revised OI, stale snapshots, survivorship, parameter fitting and circular price-impact attribution.

Common misconceptions

  • “Open interest reveals dealer positions.” It counts paired outstanding contracts, not account identity or hedge inventory.
  • “Positive GEX guarantees low volatility.” Countertrend hedging is conditional and may be small, delayed, offset or overwhelmed.
  • “Negative GEX predicts a decline.” Short-Gamma hedging is procyclical in either direction, not inherently bearish.
  • “A gamma wall or flip is a fixed price level.” It is a moving root of selected position and model assumptions.
  • “Dollar GEX is the amount dealers must trade.” It is a scaled local sensitivity; full repricing and actual execution can differ materially.

Authoritative sources

Navigation

Search the wiki...