Dealer Hedge Flow Methodology

Options dealers (market makers) do not want to bet on the direction of the price: they keep their option books delta-hedged by buying or selling the underlying. The option delta changes when the price moves, so the hedge has to be adjusted all the time. The Dealer Hedge Flow answers one question for every price level: how many US dollars of BTC would dealers have to buy or sell to stay hedged if the price moved from where it is now to that level — and would that flow push the move further or hold it back?

Effect	What dealers do	What it means for the move
Dampen	trade against the move: sell into a rally, buy into a fall	the move meets extra supply or demand and tends to slow down
Amplify	trade with the move: buy into a rally, sell into a fall	the move gets extra fuel and tends to accelerate

The source is the Deribit BTC options market. The profile is recomputed every hour; history starts on 2026-03-16. It is available in Lab → Dashboards → Blocklens Special → Dealer Hedge Flow, through the API and the MCP server (§8).

Not a trading signal

The Dealer Hedge Flow is a descriptive model of dealer positioning built from public trades. It is not a reported dealer book, a price target or a trading signal. See Limitations.

---

1. What it shows

1.1 Hedge flow and its effect

For a price level $P$ the hedge flow $H(P)$ is the total USD amount dealers would trade while the price travels from the current price $S$ to $P$, with their option positions held as they are now:

• $H > 0$ — dealers buy BTC on the way to $P$; $H < 0$ — dealers sell.
• It is one sum in dollars for the whole path, not a value per strike and not «per 1%».

Whether that flow helps or resists the move depends on the direction of the move, so the profile also shows the impact: the hedge flow signed by the direction of the move.

• impact > 0 — amplify: dealers trade in the direction of the move (buy on the way up, sell on the way down).
• impact < 0 — dampen: dealers trade against the move.

Dealers who are long gamma (on balance they hold options bought from takers) sell as the price rises and buy as it falls — they dampen moves. Dealers who are short gamma chase the price — they amplify moves. The profile shows where the book switches between the two, and how many dollars are at stake at each level.

1.2 How to read the dashboard

• Profile — one bar per price level on a fixed $500 grid within ±30% of the current price. The bar is the impact in USD: one colour for amplify, another for dampen; its length is the size of the flow if the price went there.
• Heatmap — the hourly profiles side by side: x — time, y — price, colour — the impact that the profile of that hour gave for that price. The BTC price and the zero gamma level are drawn on top. It is a history of the model, not a record of trades that actually happened.
• Tiles — net GEX, the zero gamma level, and the hedge flow if the price moved ±5% and ±10% (§3.6).
• Strike contributions — net GEX split by strike into calls and puts: which strikes the dealer gamma sits on.

---

2. Dealer position

2.1 From the full trade tape

Deribit does not publish who holds which option. We rebuild the dealer position for every option (every expiry, strike and type) from the full Deribit trade tape since the option was listed. Each trade carries the side of the taker — the participant who sent the market order. The dealer is assumed to be the counterparty of every taker trade:

$$\text{inv}_i = \sum_{\text{trades in option } i} \left(\text{taker sell} - \text{taker buy}\right)$$

• $\text{inv}_i > 0$ — dealers are long option $i$ (takers sold it to them), $\text{inv}_i < 0$ — dealers are short.
• Amounts are in BTC (one Deribit BTC option is on 1 BTC).
• An option that was never traded has inventory 0 but stays on the board, so the strike axis and the call/put split remain complete.

As a sanity check, the total absolute dealer inventory comes to roughly half of the open interest (45–59% on the boards we checked) — plausible, since part of the open interest sits between non-dealer participants.

2.2 No look-ahead

Each hourly profile uses the position as of that hour: only trades with a timestamp up to the moment of the snapshot are summed. History is computed in one forward pass over the tape, adding each day's trades in time order and computing every hourly board right after the trades that precede it. A historical profile therefore never knows about trades made after it.

2.3 Assumptions

Assumption	Meaning	Effect if it does not hold
Every taker trades against dealers	the maker side of each trade is a hedging dealer	some makers are not dealers (or do not hedge); the real flow is smaller in size, the sign structure usually survives
The position is fixed at the snapshot	moving the price along the path does not open or close options	over a long path new trades change the book; the profile is a «what if right now» picture
Sticky-strike volatility	each option keeps its implied volatility and its time to expiry while the price moves	in reality the smile shifts with the price; far levels are less certain than near ones
Deribit BTC options only	other venues and OTC positions are not included	the flow of other venues is missing from the sum
Dealers hedge continuously in the underlying	the hedge is adjusted along the whole path	in practice hedging is discrete and partly done with futures

---

3. Formulas

3.1 Gamma from the expiry forward

Each option is priced with Black-Scholes at zero rate on the forward of its expiry, which is how Deribit itself prices BTC options. For option $i$ with strike $K_i$, implied volatility $\sigma_i$ (Deribit mark IV) and time to expiry $T_i$ (years of 365 days to 08:00 UTC of the expiry day):

$$c_i = \frac{\text{underlying}_i}{\text{index}}, \qquad d_{1,2}(F) = \frac{\ln(F / K_i) \pm \sigma_i^2 T_i / 2}{\sigma_i \sqrt{T_i}}, \qquad \Gamma_{BS}(F) = \frac{\varphi(d_1(F))}{F\, \sigma_i \sqrt{T_i}}$$

• $\text{underlying}_i$ is the forward price of the option's expiry and $\text{index}$ the BTC index price in the snapshot; the ratio $c_i$ (the basis of the expiry) is held fixed as the index moves, so at an index price $P$ the forward is $c_i P$.
• The gamma with respect to the index is $\Gamma_i(P) = c_i \, \Gamma_{BS}(c_i P)$. Calls and puts have the same gamma.
• $\varphi$ and $N$ are the standard normal density and distribution function. Expired options and options without an implied volatility are skipped.

3.2 GEX at a price

Gamma exposure at a price $P$ is the USD amount dealers trade for a 1% move around $P$:

$$\text{gex}(P) = \sum_i \text{inv}_i \, \Gamma_i(P) \, \frac{P^2}{100}$$

Net GEX is this curve at the current price, $\text{gex}(S)$. Positive net GEX — dealers are net long gamma (dampening regime); negative — net short gamma (amplifying regime).

3.3 Hedge flow H(P)

When the price moves by $dp$, the delta of the dealer book changes by $\sum_i \text{inv}_i \Gamma_i(p)\, dp$ BTC, and dealers trade the opposite amount to stay hedged — $p$ dollars per BTC. Adding this up along the path from $S$ to $P$:

$$H(P) = -\int_S^P \sum_i \text{inv}_i \, \Gamma_i(p) \, p \; dp$$

Equivalently $H'(P) = -100 \cdot \text{gex}(P) / P$: the hedge flow grows fastest where GEX is large, and its slope changes sign exactly where GEX does.

3.4 Closed form

With zero rate and sticky-strike volatility the integral has an exact solution (because $K \varphi(d_2) = F \varphi(d_1)$):

$$H(P) = -\sum_i \text{inv}_i \, \frac{K_i}{c_i} \left[ N\big(d_2(c_i P)\big) - N\big(d_2(c_i S)\big) \right]$$

No price grid or numerical integration is involved: the value at every level is exact, also for options a few hours from expiry whose gamma is narrower than the $500 grid step.

3.5 Impact

$$\text{impact}(P) = H(P) \cdot \operatorname{sign}(P - S)$$

• Above the price ($P > S$) dealers buying ($H > 0$) push a rally further — impact $> 0$, amplify; dealers selling ($H < 0$) lean against it — impact $< 0$, dampen.
• Below the price ($P < S$) it is the other way round: selling into a fall amplifies, buying into it dampens.

The profile bars and the heatmap colour are the impact in USD.

3.6 Summary values

Value	Definition
Net GEX	$\text{gex}(S)$, USD per 1% move
Zero gamma	the sign change of $\text{gex}(P)$ nearest to $S$ within $S \pm 50\%$; empty if the curve keeps one sign there
Flips	all sign changes of $\text{gex}(P)$ within $S \pm 50\%$, ascending
Flow at ±5% / ±10%	$H$ exactly at $S \times 1.05$, $0.95$, $1.10$, $0.90$ (not at the nearest grid level)
Largest amplify / dampen zone	the $500 level band (from the neighbour nearer to $S$ to the level) with the largest amplifying / dampening marginal flow

• Sign changes are found by scanning $\text{gex}(P)$ with a $100 step and refining each crossing by bisection to $1. Two crossings closer than $100 to each other can be missed.
• The zero gamma level is searched in ±50% of the price, wider than the ±30% profile, because for BTC it often lies 25–30% away from the price. It can therefore be outside the drawn profile: the tile shows its value, the line on the chart is then clipped at the edge.
• The largest zones are located by the marginal flow of a band, not by $H$ itself: the cumulative $H$ usually peaks at the edge of the range.

3.7 Worked example

The hourly board of 2026-10-24 06:00 UTC, BTC index price $80,661:

Value	Result	Reading
Net GEX	+$32.3M per 1%	dealers net long gamma
Price +5% ($84,694)	$H$ = −$214M	dealers sell into the rally — dampen
Price −5% ($76,628)	$H$ = +$299M	dealers buy into the fall — dampen
Price +10% ($88,727)	$H$ = −$324M	dampen
Price −10% ($72,595)	$H$ = +$661M	dampen
Zero gamma	$59,068 (−26.8%)	the book turns short gamma only after a deep fall

A rough check with the net GEX alone, as if GEX stayed at its current value all the way: −100 × $32.3M × ln(1.05) ≈ −$158M for +5%. The exact flow is larger in both directions because GEX is not constant along the path: on this board it is higher a few percent away from the price than at the price itself.

---

4. How it differs from classic per-strike GEX

The usual GEX chart is a snapshot: «dollars per 1% move right now», split by strike. The Dealer Hedge Flow is built differently:

	Classic per-strike GEX	Dealer Hedge Flow
Question	how much gamma sits on each strike now	how many dollars dealers trade if the price goes to a level
Shape	a snapshot at the current price	a path: gamma is recomputed at every price between now and the level
Result	one value per strike, in USD per 1%	one sum in USD per price level, with amplify / dampen
Dealer position	often guessed from open interest (e.g. «dealers are long calls and short puts»)	rebuilt from the full trade tape for every option
Regime threshold	often the sign change between neighbouring strikes	the zero gamma level of the recomputed GEX curve

The strike view is still available as the strike contributions panel: it shows where the dealer gamma sits; the hedge flow shows what it does when the price moves.

---

5. Legacy values

The API also returns a few older GEX values for continuity: legacy_* fields in the series and a legacy block in the profile. They are not used in the Dealer Hedge Flow. The same block holds the strikes with the largest positive and the most negative GEX (support and resistance strikes); they are computed on Deribit's gamma at the snapshot, and strikes whose total GEX is below $1,000 in absolute value are left out.

5.1 Why Deribit's published gamma is not used

Deribit publishes a gamma for every option, but rounded to 5 decimal places. For BTC the gamma of an at-the-money option is itself of the order of $10^{-4}$–$10^{-5}$, so the rounding step $10^{-5}$ is an error of 5–50% per option, and dealer positions of opposite sign across strikes keep these errors from cancelling. A legacy total GEX built on the rounded gamma (and on the index instead of the forward) differs from the exact Black-Scholes value: on 106 hourly boards of 2026-10-20…24 the ratio of exact to legacy net GEX had a median of 1.02, with 10% of boards below 0.83 and 10% above 1.15. We publish only the exact value as net GEX; the legacy total stays among the legacy values.

5.2 Why the legacy flip point is not a gamma flip

The legacy flip point (legacy_flip_point) is the sign change of per-strike GEX between the two strikes around the current price. It almost always lies within a fraction of a percent of the price and says nothing about where the dealer book as a whole changes regime. The real threshold is the zero gamma level: the price at which the whole GEX curve, recomputed at that price, crosses zero (§3.6).

---

6. Data

• Source: Deribit BTC options — the full trade tape (with taker side) and an hourly snapshot of the options board (strikes, expiries, mark IV, forward and index price).
• Schedule: a new profile every hour, computed right after the hourly snapshot and published at about :20 UTC. A snapshot that fails the board completeness check is skipped: it stays a gap, not a zero.
• History: from 2026-03-16, every hourly snapshot. The positions behind the first profiles already include all trades since each option was listed, so there is no cold start.
• Gap: 2026-06-08 … 2026-07-07 — a collection outage; there are no snapshots and therefore no profiles for these days. The trade tape has no gap, so positions after the outage are complete.
• Daily values (net GEX, zero gamma, flow at ±5% / ±10% in the Metrics Reference and in the daily API series) are the 00:00 UTC snapshot of each day (the first snapshot of the day if 00:00 is missing) — away from the daily expiry at 08:00 UTC. A published day does not change.
• Expiry hour: in the hour before the daily expiry (the 07:00 UTC snapshot) options with about one hour left have a very narrow gamma spike at their strikes. Several flips and a zero gamma level next to the price are possible on that snapshot; this is consistent with the model and is kept as is.

---

7. Tiers

Tier	Profile	History (heatmap, series)	Strike contributions	Export
Free	the latest hourly profile	last 30 days, daily step (1d)	—	—
Pro, Enterprise	any hourly snapshot since 2026-03-16	full history since 2026-03-16, steps 1h / 6h / 1d	yes	yes

The daily metrics in the Metrics Reference follow the usual tier rules of the metrics catalog (Rate Limits & Tiers).

---

8. API and MCP

Endpoint	What it returns
GET /v1/options/gex	daily series: BTC price, net GEX, zero gamma, flow at ±5% / ±10%, and legacy values
GET /v1/options/gex/hourly	the same for every hourly snapshot (Pro)
GET /v1/options/gex/profile	one snapshot (latest by default, or date / ts): levels, hedge flow, impact, GEX per level, zero gamma, flips, ±5% / ±10% flow, strike contributions
GET /v1/options/gex/heatmap	time × price matrix of impact for from / to and resolution = 1h, 6h or 1d, with price and zero gamma overlays

• All amounts are in USD; hedge flow and impact use the signs of §1.1. The 1h heatmap covers at most 90 days per request.
• The same URL returns different history for different tiers (§7). Parameters, response fields, limits and errors: Dealer Hedge Flow API; API keys: Authentication.
• The MCP server has three tools on top of these endpoints: get_gex_daily (daily series), get_gex_profile (a compact summary of one profile: regime, zero gamma, flow for ±1…30% moves, largest amplify and dampen levels) and get_gex_heatmap (history of the profile, coarsened to fit an AI agent's context). The daily values are also available as catalog metrics through get_metric.

---

9. Limitations

• It is a model. The dealer side of every taker trade is an assumption (§2.3); real dealers also trade with each other, hedge with futures and keep positions on other venues.
• «What if right now». The position, implied volatility and time to expiry are frozen at the snapshot. Far levels (±20–30%) describe the current book, not the book the market will have when the price gets there.
• One venue. Only Deribit BTC options are included.
• Hedging is not the only flow. The profile measures one source of forced trading; spot and futures liquidity, liquidations and discretionary flow can be much larger on any given day.

The Dealer Hedge Flow is a descriptive model, not a price target or a trading signal.