Dealers are short $13.1bn of SPY gamma per 1%, on the assumed dealer side
Where are dealers long and short gamma in each market, and is that position moving?
In brief
Where dealers' gamma sits today on the stated convention, and what would change the read- On the stated convention, dealers are short $13.1bn of SPY gamma per 1% move (7th percentile of 252 sessions).→ Exhibit 2
- The largest single cell holds −$2.9bn of SPY's exposure per 1% move.→ Exhibit 1
- SPY's fell $6.4bn per 1% move over five sessions.→ Exhibit 3
- in SPY puts is 1.85 times that in calls.→ Exhibit 2
- After top-quintile gamma days SPY's next-month had a median of 11.3%, against 14.2% after bottom-quintile days.→ Exhibit 4
| What would change this view | Trigger | Today | Decided by |
|---|---|---|---|
| Dealers turn long SPY gamma | net gamma exposure above $0bn per 1% move | $−13.1bn (16 Nov 2026)not crossed | each session's close (next 17 Nov 2026) |
| SPY positioning reaches a one-year extreme | net gamma below $−11.4bn (10th percentile of the year) | $−13.1bn (16 Nov 2026)crossed | each session's close (next 17 Nov 2026) |
The reasoning behind each point
- Net gammaOn the stated convention, dealers are short $13.1bn of SPY gamma per 1% move (7th percentile of 252 sessions).
Net gamma exposure assumes dealers are long the calls and short the puts customers trade: Γ × S² × open interest × 100 × 0.01, in dollars per 1% move. A long position damps moves as dealers hedge; a short one amplifies them — a reading of the convention, not of any dealer's book.
- Where it sitsThe largest single cell holds −$2.9bn of SPY's exposure per 1% move.
It is the 0.15:0.35 bucket expiring in 0-7 days. The heat map shows all 54 cells: delta bucket across, days to expiry down.
- Five sessionsSPY's net gamma fell $6.4bn per 1% move over five sessions.
Most of the change is in the 0.60:0.80 bucket (−2.56bn).
- Open interestOpen interest in SPY puts is 1.85 times that in calls.
11,326,553 puts against 6,126,914 calls on 16 Nov 2026.
- The testAfter top-quintile gamma days SPY's next-month realised volatility had a median of 11.3%, against 14.2% after bottom-quintile days.
98 top-quintile and 109 bottom-quintile sessions with a complete month after them; the months overlap, so the sessions are not independent. A test of the claim that dealers pin the market, not a signal.
Next: Positioning — who is crowded in futures
SPY: largest cell 0.15:0.35, 0-7 days
In which delta and expiry buckets is dealers' gamma concentrated today?Exhibit 1 · SPY dealer gamma by option delta and days to expiry
About this exhibit
Dealers short $13.1bn per 1% on the assumed side (7th percentile)
How large is the net position against its own year, and where did the last five sessions move it?Exhibit 2 · SPY dealer net gamma and SPY's close, each session over the year
Net −$13.13bn per 1% move, the 7th percentile of its year (252 sessions). Open interest: 11,326,553 puts against 6,126,914 calls (1.85×).
About this exhibit
Exhibit 3 · Five-session change in SPY net gamma, by delta bucket
About this exhibit
Median 11.3% after high-gamma days, 14.2% after low
Has been lower in the month after high-gamma days than after low-gamma days?Exhibit 4 · SPY's realised volatility in the month after high- and low-gamma days
479 of 2,921 sessions with a percentile have a complete month after them, the last on 16 Oct 2026. The months overlap, so neighbouring sessions are not independent: a test of the claim that dealers' hedging calms the market, not a signal.
About this exhibit
Appendix
The sign convention, the scaling and the buckets, and what each check holds55 checks, all pass
Where dealers' exposure to gamma sits in each market, on a stated convention: dealers are taken to be long the calls and short the puts customers trade, so the sign is assumed, not observed. The heat map splits the net exposure by how far each contract is from the money and how long it has to run; the history sets it against its own year and the market's close; the five-session change shows where it moved. Beside the state stands its test — how much the market moved in the month after the days of highest and lowest exposure — because the claim that dealers' hedging calms a market is contested, and the page describes the record rather than asserting it.
Net gamma exposure and the five-session change
| Symbol | Meaning | From | Assumed or estimated |
|---|---|---|---|
| the contract's gamma | the options engine | – | |
| open interest, contracts of 100 | the options engine | – | |
| the dealer side: long calls, short puts | assumed | assumed |
A convention, not an observation: the sign of each contract's dealer position is assumed.
State, not a forecast: the dealer side of each contract is assumed, not observed — dealers long the calls and short the puts customers trade. Net gamma exposure = Σ ±Γ × S² × OI × 100 × 0.01 (calls +, puts −), in dollars per 1% move; the heat map splits it into 6 bands of |delta|, calls and puts together, by 9 bands of days to expiry, the last holding every contract beyond 730 days, so its cells add up to the whole book.
| As stated | |
|---|---|
| Dealer side | dealers long calls, short puts |
| Exposure per contract | gamma * spot^2 * open interest * 100 * 0.01 |
| Unit | USD per 1% move |
Delta: 0.00:0.15 · 0.15:0.35 · 0.35:0.50 · 0.50:0.60 · 0.60:0.80 · 0.80:1.00. Days to expiry: 0-7d · 7-21d · 21-35d · 35-60d · 60-90d · 90-180d · 180-365d · 365-730d · >730d.
Percentile: the share of the trailing 252 sessions (a year; at least 63) at or below the day's net exposure, computed here and causal — the options engine's own percentile ranks against the whole history and is not used.
The test beside the state (the author's decision of 3 Oct 2026 (D-2026-10-03-20)): for each session with a percentile, realised volatility over the next 21 sessions (√252 × sample s.d. of daily log closes, from the Performance page's closes), grouped by the percentile — top quintile (≥ 80), bottom (≤ 20), the rest — with the medians and counts. The windows overlap, so neighbouring sessions are not independent; it describes the record and is not a signal.
Days to expiry are calendar days (τ = days ÷ 365). The site's realised volatility is √252 × the sample standard deviation of daily log closes, overnight returns included (the author's decision of 3 Oct 2026 (D-2026-10-03-19)); the options engine states its own as annualised by 365 on trading days, overnight return included. The quotes are the vendor's end-of-day set for the session, stamped 16:00 New York, 16 Nov 2026; the options engine's results for a session are read only when they are complete and were written after the close had settled (the author's decision of 4 Oct 2026 (D-2026-10-04-02)).
Option data: marketdata.app. Every figure derived from it on this page is computed here.
This page follows the author's decision of 4 Oct 2026 (D-2026-10-04-01) and the author's decision of 3 Oct 2026 (D-2026-10-03-21) and the author's decision of 4 Oct 2026 (D-2026-10-04-05) and the author's decision of 4 Oct 2026 (D-2026-10-04-06) and the author's decision of 4 Oct 2026 (D-2026-10-04-08) and the author's decision of 7 Oct 2026 (D-2026-10-07-SKEW1).
How the numbers are made
session_folder ──▶ core.imports: the newest complete folder on or before the run day; every file's SHA-256 against checks.json ──▶ the verified folder the verified folder ──▶ net gamma exposure and its 48 delta × DTE cells on the session, the aggregate held to the sum of its cells; the trailing one-year percentile; the five-session change by delta bucket ──▶ positioning by asset closes + positioning by asset ──▶ realised volatility over the month after each session, grouped by the percentile's top and bottom quintile, with counts ──▶ the conditional test
Sources
- licensedOptions engine session folder: end-of-day open interest and greeks, aggregated to dealer gamma exposure
marketdata.app
- computed hereThe Performance page's daily closes, copied as served and compared with it
Computed here from the closes the Performance page prints (Prices: Yahoo Finance).
Pages this one copies from
| Page | Title | Updated | Data to | Arrays | Values | Source checks |
|---|---|---|---|---|---|---|
| Performance | Cross-asset scoreboard | 2026-11-17 | 16 Nov 2026 | 8 | 0 | 23 pass |
Reproduction checks
| Check | Published | Computed | Tolerance | Result |
|---|---|---|---|---|
| Session folder 2026-11-16 · _SUCCESS: present: the engine writes it last, and a folder without it is not read | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · summary_history.parquet: present: a file the contract names | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · surface_latest.parquet: present: a file the contract names | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · dealer_history.parquet: present: a file the contract names | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · combos.parquet: present: a file the contract names | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · sample_raw.parquet: present: a file the contract names | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · README.md: SHA-256 equals the hash checks.json lists for it | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · combos.parquet: SHA-256 equals the hash checks.json lists for it | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · dealer_history.parquet: SHA-256 equals the hash checks.json lists for it | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · sample_raw.parquet: SHA-256 equals the hash checks.json lists for it | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · skew_history.parquet: SHA-256 equals the hash checks.json lists for it | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · summary_history.parquet: SHA-256 equals the hash checks.json lists for it | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · surface_latest.parquet: SHA-256 equals the hash checks.json lists for it | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · checks.json: session_date is the folder's name | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · checks.json: session_date is a weekday (a US session) | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · checks.json: session_date is on or before the run day 2026-11-17 | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · checks.json: quote_time_utc is at or after the session's 16:00 New York close (the vendor's end-of-day quotes) | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · checks.json: written_utc is at or after the session's 16:00 New York close plus 15 minutes (written once the session had settled) | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · checks.json: written_utc is not after the moment of the check | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · checks.json: quote_time_utc is not after the moment of the check | 1 | 1 | 0 | pass |
| Session folder 2026-11-16 · checks.json: every one of the export's gate checks passed, standing in for the engine's own (6 listed) | 1 | 1 | 0 | pass |
| Contract: the session folder states dealer as the site's contract does | {"buckets": "|delta| bands, calls and puts together, by days-to-expiry bands; a cell with no contracts is written as 0, the sum over no contracts", "delta_buckets": ["0.00:0.15", "0.15:0.35", "0.35:0.50", "0.50:0.60", "0.60:0.80", "0.80:1.00"], "dte_buckets": ["0-7d", "7-21d", "21-35d", "35-60d", "60-90d", "90-180d", "180-365d", "365-730d", ">730d"], "dte_cap": "none: the cells cover the whole book, the last band every contract beyond 730 days", "flow_5d": "net_gex_usd_per_1pct less its value five sessions earlier", "gex": "gamma * spot^2 * open interest * 100 * 0.01", "log_pc": "ln(oi_puts / oi_calls)", "sign": "dealers long calls, short puts", "unit": "USD per 1% move"} | {"buckets": "|delta| bands, calls and puts together, by days-to-expiry bands; a cell with no contracts is written as 0, the sum over no contracts", "delta_buckets": ["0.00:0.15", "0.15:0.35", "0.35:0.50", "0.50:0.60", "0.60:0.80", "0.80:1.00"], "dte_buckets": ["0-7d", "7-21d", "21-35d", "35-60d", "60-90d", "90-180d", "180-365d", "365-730d", ">730d"], "dte_cap": "none: the cells cover the whole book, the last band every contract beyond 730 days", "flow_5d": "net_gex_usd_per_1pct less its value five sessions earlier", "gex": "gamma * spot^2 * open interest * 100 * 0.01", "log_pc": "ln(oi_puts / oi_calls)", "sign": "dealers long calls, short puts", "unit": "USD per 1% move"} | 0 | pass |
| The folder's buckets are the contract's: 6 |delta| bands by 9 days-to-expiry bands, the 54 cells of the heat map (bands that differ) | 0 | 0 | 0 | pass |
| Every row is either an aggregate (no buckets) or a cell (both buckets) (rows with one bucket only) | 0 | 0 | 0 | pass |
| Every date is on or before the session (rows after it) | 0 | 0 | 0 | pass |
| Every symbol and date that ships cells ships all 54 delta × DTE cells once (168 symbol-dates; incomplete or repeated) | 0 | 0 | 0 | pass |
| Every asset ships its cells on the session (assets without) | 0 | 0 | 0 | pass |
| Net gamma exposure: each aggregate equals the sum of its 54 cells (168 symbol-dates; max abs difference, $bn per 1%) | 0 | 0 | 0.000001 | pass |
| Every symbol-date that ships cells ships its aggregate (168 symbol-dates; without) | 0 | 0 | 0 | pass |
| Log P/C is ln(open interest in puts ÷ in calls), recomputed on every aggregate row and empty where either is zero (23866 rows; max abs difference) | 0 | 0 | 1e-9 | pass |
| The five-session change is net gamma exposure less its value five sessions earlier, recomputed (23826 rows; max abs difference, $bn per 1%) | 0 | 0 | 0.000001 | pass |
| Percentiles: each asset's trailing one-year percentile of net gamma exposure, recomputed by a plain loop, equals the published one (23370 values; max abs difference) | 0 | 0 | 1e-9 | pass |
| Percentiles are causal: SPY's computed on its history cut at 11 dates equals the full series there (mismatches) | 0 | 0 | 0 | pass |
| Conditional test, SPY: every session with a percentile and a complete month after it falls in exactly one group (479 sessions; counted otherwise) | 0 | 0 | 0 | pass |
| Conditional test, QQQ: every session with a percentile and a complete month after it falls in exactly one group (479 sessions; counted otherwise) | 0 | 0 | 0 | pass |
| Conditional test, IWM: every session with a percentile and a complete month after it falls in exactly one group (479 sessions; counted otherwise) | 0 | 0 | 0 | pass |
| Conditional test, TLT: every session with a percentile and a complete month after it falls in exactly one group (479 sessions; counted otherwise) | 0 | 0 | 0 | pass |
| Conditional test, IEF: every session with a percentile and a complete month after it falls in exactly one group (479 sessions; counted otherwise) | 0 | 0 | 0 | pass |
| Conditional test, HYG: every session with a percentile and a complete month after it falls in exactly one group (479 sessions; counted otherwise) | 0 | 0 | 0 | pass |
| Conditional test, GLD: every session with a percentile and a complete month after it falls in exactly one group (479 sessions; counted otherwise) | 0 | 0 | 0 | pass |
| Conditional test, USO: every session with a percentile and a complete month after it falls in exactly one group (479 sessions; counted otherwise) | 0 | 0 | 0 | pass |
| SPY: the five-session changes by delta bucket add up to the aggregate's five-session change (max abs difference, $bn per 1%) | 0 | 0 | 0.000001 | pass |
| QQQ: the five-session changes by delta bucket add up to the aggregate's five-session change (max abs difference, $bn per 1%) | 0 | 0 | 0.000001 | pass |
| IWM: the five-session changes by delta bucket add up to the aggregate's five-session change (max abs difference, $bn per 1%) | 0 | 0 | 0.000001 | pass |
| TLT: the five-session changes by delta bucket add up to the aggregate's five-session change (max abs difference, $bn per 1%) | 0 | 0 | 0.000001 | pass |
| IEF: the five-session changes by delta bucket add up to the aggregate's five-session change (max abs difference, $bn per 1%) | 0 | 0 | 0.000001 | pass |
| HYG: the five-session changes by delta bucket add up to the aggregate's five-session change (max abs difference, $bn per 1%) | 0 | 0 | 0.000001 | pass |
| GLD: the five-session changes by delta bucket add up to the aggregate's five-session change (max abs difference, $bn per 1%) | 0 | 0 | 0.000001 | pass |
| USO: the five-session changes by delta bucket add up to the aggregate's five-session change (max abs difference, $bn per 1%) | 0 | 0 | 0.000001 | pass |
| Every takeaway and tile keeps the house grammar and carries no process word (5 takeaways, 5 tiles; faults) | 0 | 0 | 0 | pass |
| The options source line carries the attribution string the session folder states | marketdata.app | marketdata.app | 0 | pass |
| The Performance page: the source page is published with all its checks passing (dated 17 Nov 2026, 23 checks) | 1 | 1 | 0 | pass |
| The Performance page: the site still serves the page this one copied (dated 17 Nov 2026) | 1 | 1 | 0 | pass |
| The Performance page: every copied observation equals the source page, matched by date (4168 values in 8 arrays; mismatches) | 0 | 0 | 0 | pass |
| Decisions: every desk decision this page's specification names is in the decisions file (D-2026-10-03-16, D-2026-10-03-20, D-2026-10-03-21; ids missing) | 0 | 0 | 0 | pass |
- Options engine session folder: end-of-day open interest and greeks, aggregated to dealer gamma exposure — marketdata.app
- The Performance page's daily closes, copied as served and compared with it — Computed here from the closes the Performance page prints (Prices: Yahoo Finance).