Every gamma product on the market draws the call wall, the put wall, and the gamma flip as single-strike lines. We measured 1,090 trading sessions of SPY and SPX dealer positioning to ask whether the data supports that precision. It doesn't. A dealer gamma level is a zone roughly $2–5 wide — and here is exactly how we know.
1. SPX carries roughly 11× SPY's dealer gamma, and the gap has widened every year since 2022. 2. The SPY-only wall and the SPX-only wall sit at genuinely different strikes — median $2–3 apart — and it is not a rounding artifact (we tested that hypothesis; it failed). 3. The two books converge: at $5 price resolution, the SPY-only wall is the combined wall in 96–100% of sessions for everything out to 30 days — in calm, normal, and stressed markets. 4. For the full book including LEAPS, SPY-only is decaying in real time: retention fell from 1.00 to 0.59 as institutional long-dated SPX positioning tripled. 5. The gamma flip is the least precise of the three levels and deserves the widest band on any chart.
A standard objection to any SPY-based gamma tool goes like this: "The real dealer book is in SPX. SPY is the retail sideshow. Levels computed from SPY alone are missing most of the gamma."
The first half of that objection is correct, and we will show precisely how correct. The interesting question is the second half: does missing most of the gamma mean missing the levels? Magnitude and location are different claims. A wall is not set by how much gamma exists — it is set by where gamma clusters. SPY and SPX track the same index, so their gamma might cluster at the same prices even if one book is an order of magnitude larger.
Nobody had measured it, so we did.
For each of 1,090 trading sessions from January 2022 through May 2026, we took a morning snapshot (45 minutes after the open) of the full SPY option chain and the full SPX chain — both the AM-settled monthlies and the PM-settled SPXW weeklies — with open interest and per-contract gamma. Data: AlgoSeek US equity and index options via QuantConnect.
Dollar gamma per contract is the standard construction: Γ × OI × 100 × S² × 0.01 — the dollar hedging flow per 1% move in the underlying. Two methodological points matter more than everything else:
Levels were computed twice per session per expiration bucket (0DTE, 0–7 days, 0–30 days, all expirations): once from SPY alone, once from SPY + SPX + SPXW combined. The gamma flip was located by re-pricing the entire book with Black-Scholes across a ±12% spot sweep and interpolating the zero crossing of net dealer gamma — not by cumulatively summing across strikes, which finds noise.
| Year | SPY gross gamma | SPX gross gamma | SPX / SPY |
|---|---|---|---|
| 2022 | $20.2B | $132.9B | 6.6× |
| 2023 | $30.8B | $202.0B | 6.5× |
| 2024 | $32.5B | $276.1B | 8.5× |
| 2025 | $31.6B | $301.5B | 9.5× |
| 2026 | $32.8B | $364.5B | 11.1× |
Median gross dollar gamma per 1% move, all expirations, by year.
Concede the objection's premise immediately: the SPX book is enormous relative to SPY, and the ratio has nearly doubled in four years. SPY dealer gamma has been flat around $32B since 2023 while SPX has nearly tripled. Institutional positioning is growing in SPX; retail ETF gamma is static. Any analysis that treats the two books as interchangeable in size is wrong and getting wronger.
At $1 resolution, the wall computed from SPY alone and the wall computed from SPX alone sit a median of $2–3 apart at 0DTE. Our first hypothesis was mundane: the SPY/SPX ratio is not exactly 0.1 (dividends drift it — across the sample it ranged such that SPX 7200 mapped anywhere from $0.89 to $3.06 below SPY 720), so SPX's round-number strikes land systematically off SPY's round numbers, and perhaps the "disagreement" is just that arithmetic.
We tested it, and it failed. If grid arithmetic drove the disagreement, the wall gap would track the ratio's deviation from 0.1 — which swung 3.4× across the sample. The correlation across eight bucket/side combinations was approximately zero, with signs flipping (−0.28 to +0.25). The hypothesis is falsified and we report it because negative results are results.
What remains is the interesting explanation: SPY and SPX are different clienteles. Retail and ETF flow concentrates SPY open interest on SPY's round numbers; institutional and structured-product flow concentrates SPX open interest on SPX's round numbers. Same index, different traders, genuinely different strikes — two interleaved gamma grids, permanently offset, that never quite touch.
Zoom out from $1 bins and the two books collapse into one. Median retention of the SPY-only call wall inside the combined SPY+SPX book:
| Expiry bucket | $1 bins | $2 bins | $5 bins | $10 bins |
|---|---|---|---|---|
| 0DTE | 0.67 | 1.00 | 1.00 | 1.00 |
| 0–7 days | 0.49 | 0.91 | 1.00 | 1.00 |
| 0–30 days | 0.37 | 0.75 | 1.00 | 1.00 |
| All expirations | 0.27 | 0.62 | 1.00* | 1.00 |
1,090 sessions pooled. Put-wall results are equivalent. *Pooled full-book value masks a year trend — see Finding 4. Top-3 hit rates at $5 bins: 96–100% for 0–30 DTE.
Convergence is monotonic and clean: 0DTE merges by $2 bins; everything out to 30 days merges by $5. And it is regime-invariant — splitting the sample by VIX (calm <15, normal 15–25, stressed ≥25, which includes the 2022 bear market, the August 2024 volatility spike, and the April 2025 shock) leaves the $5-bin result intact in every regime, with 0DTE at $2 bins dipping only to 0.92 under stress.
One result surprised us: convergence is strongest during triple-witching weeks. Across the 17 quarterly expiration weeks in the sample, full-book retention at $2 bins was 1.00 versus 0.60 in ordinary weeks. Mechanically this makes sense — quarterly expiration concentrates open interest onto the large round strikes both books share — but it means the two books agree most at precisely the moment open interest peaks.
The pooled full-book number above hides the most forward-looking result in the study. Full-book call-wall retention at $5 bins, by year:
| Year | 2022 | 2023 | 2024 | 2025 | 2026 |
|---|---|---|---|---|---|
| Retention @ $5 | 1.00 | 1.00 | 1.00 | 1.00 | 0.59 |
| Retention @ $2 | — | 0.74 | 0.69 | 0.45 | 0.23 |
All-expirations bucket only. Short-dated buckets show no decay: 0DTE and 0–7 day retention at $5 bins is 1.00 in every year, including 2026.
Once LEAPS and multi-month expirations are included, the SPY proxy is failing in real time — a clean monotonic decay that tracks the growth of long-dated institutional SPX positioning (covered-call overlays, structured products, portfolio hedges) that simply has no SPY equivalent. In 2026, a full-book gamma picture built from SPY alone sees roughly half the peak call gamma.
The honest summary cuts both ways: short-dated dealer positioning is fully visible in SPY; long-dated dealer positioning genuinely requires SPX. The two claims come from the same table.
Of the three standard levels, the gamma flip — the spot price at which net dealer gamma changes sign — is the least precise. The SPY-only flip landed within $2 of the full-book flip in only about half to two-thirds of sessions, with a median error near $2 and a 90th-percentile error several dollars wide. The flip inherits model risk the walls don't have (it depends on implied volatility, rate, and dividend inputs, not just open interest), and its location is only pinned between the two strikes that bracket the sign change. It is routinely quoted to the penny. It should not be.
Put the findings together and a simple picture emerges. A dealer gamma level is not a price. It is a zone with an intrinsic width of roughly $2–5 — wider for the flip, wider still at longer tenors — because the "level" is the superposition of two real books held by different traders at strikes that were never designed to align. No amount of additional data narrows it, because the width is a property of the market, not of anyone's measurement.
Charting products in this category — ours included, until this study — have universally drawn these levels as single-strike lines. A line at 725.00 asserts a precision the underlying positioning does not support. The line is not wrong about where; it is wrong about how exactly.
Dark Horse renders gamma levels as zones: the call wall, put wall, and gamma flip are drawn as shaded bands whose width comes from the live concentration of gamma across neighboring strikes, floored at the widths this study measured. A tight band means gamma is stacked on one strike — a level dealers will defend. A wide band means it is smeared across several — a mushy region, not a wall. The width itself is information, and to our knowledge no other product surfaces it.
The full methodology — every level definition, the dollar-gamma construction, the retention metric, the stated limitations, and the nine per-finding datasets behind the tables above — is published as an open repository: github.com/DarkHorseTrading/dealer-gamma-precision. Released under CC BY 4.0 — free to reproduce, cite, or attack the numbers.
Dark Horse provides analytical and educational tools only. Nothing on this page is investment advice, dealer positioning is not a prediction of future price, and past relationships between positioning and price may not persist. © Dark Horse · a 4C Markets LLC brand