- The GEX |netGex| profile published BEFORE the US RTH open overlaps the Volume Profile of the RTH session that follows at a median Histogram Intersection of 57.4%.
- The Point-of-Control of the two distributions (GEX max-magnitude strike vs VP max-volume strike) is a median 15 points apart on ES, with 25% of days within 10 points.
- The result is consistent across three independent cross-check metrics (Bhattacharyya 84%, Cosine 64%, Wasserstein 8.4pt). Sample n = 132 trading days, Feb–Aug 2026.
- Beyond where volume concentrates, the field also leans which way: in a decided gamma regime the charm magnet and the overnight loading area point to the side of the open the volume POC lands on ~65% of the time when they agree — but only outside the transition zone, where the read breaks down (§3).
01 Universe & method
For each trading day with a published PRE_OPEN snapshot in the TLADe archive (= GEX profile broadcast roughly 60 minutes before US RTH open), we compare the shape of two histograms over the strike-price axis:
- The GEX |netGex| profile (= mass of net gamma at each listed strike, taken in absolute) extracted from the snapshot's
plotData. - The Volume Profile built from 5-minute IB RTH candles of the same day (09:30–16:00 ET), volume distributed uniformly across the ticks each bar spans.
Both profiles are re-binned onto a common 5-point grid over the strict RTH range, then normalized to a probability distribution (Σ = 1). Restricting to the RTH range (no buffer) is important: outside the day's actual trading range, the GEX profile carries structural mass at popular round strikes (5800, 6000, 7000) that has no corresponding intraday flow.
| dimension | value |
|---|---|
| Trading days (sample n) | 132 |
| Period covered | 2026-02-09 → 2026-08-14 |
| Bin width | 5.0 ES points |
| Restriction | strict RTH range (no buffer) |
| GEX source | PRE_OPEN.json plotData.data ("Total Net GEX" trace) |
| VP source | ES 5m RTH candles, uniform tick distribution |
The Magnet (referred to throughout this study as the GEX POC) is defined operationally as:
Magnet ≡ argmaxK ∈ RTH range |netGex(K)|
where netGex(K) = callGEX(K) − putGEX(K) is the signed net-gamma exposure at strike K, read from the "Total Net GEX" trace inside PRE_OPEN.json.plotData, and the argmax is taken on the absolute value over strikes that fall inside the day's eventual RTH high–low range.
Not the same as any single named level shown in the TLADe terminal:
- Call Wall uses only call-side gamma → argmax callGEX.
- Put Wall uses only put-side gamma → argmax putGEX.
- Max Pain minimizes aggregate option-holder payoff at expiry — different objective entirely.
- Zero Gamma (ZG) is the zero crossing of netGex, not its absolute peak.
The Magnet is therefore the dollar-weighted dealer-hedging center of mass: the single strike where the combined call & put gamma exposure peaks in magnitude, regardless of sign. It is the object whose visit / non-visit behaviour the study quantifies.
02 Headline metric & cross-checks
Histogram Intersection is the literal "% of common mass" between the two normalized distributions, point-by-point on the shared 5-point grid:
HI(p, q) = Σi min(pi, qi)
Reads as: "on a typical day, the GEX pre-open already accounts for X% of where the day's volume will accumulate". Range [0, 1]. Two identical distributions give 1.0; two disjoint distributions give 0.0; a subset gives the subset's mass.
| metric | mean | median | p25 | p75 | min — max |
|---|---|---|---|---|---|
| Histogram Intersection | 56.6% | 57.4% | 50.2% | 65.7% | 19.6% — 80.2% |
| Bhattacharyya coefficient | 81.4% | 83.8% | 77.6% | 88.2% | 33% — 95% |
| Cosine similarity | 61.1% | 63.6% | 51.2% | 73.4% | 16% — 92% |
| Wasserstein / EMD (ES pts) | 10.5 | 8.4 | 5.6 | 13.8 | 2.7 — 40.9 |
| POC distance (ES pts) | 22.8 | 15.0 | 10.0 | 30.0 | 0 — 110 |
Three different metrics agree on the same direction: the GEX pre-open profile is not a random sample of the strike space — its shape concentration matches where the day's volume will then concentrate. The Wasserstein distance (= the price-axis "earth mover" cost of transforming one profile into the other) sits around 10 points on a typical day, which is a small displacement relative to the ~50–100 point ES RTH range.
03 Which side: the field's directional lean
The overlay in §2 answers where the day's volume concentrates. A second, independent question is which side — does the volume lean above or below the cash open? Histogram intersection is a shape statistic; it is silent on direction. So we ask the field separately.
The read. At the 09:30 ET open we look at where two directional anchors sit relative to the open: the charm magnet (the short-dated pin) and the loading area — the strikes where dealer gamma has been built overnight, recovered by comparing the pre-open profile to the prior close (a load / unload read per strike). The predicted side is simply the side of the open those anchors fall on; the realized side is where the RTH volume POC ends up.
In the clear regimes (positive- and negative-gamma), each anchor leans the correct way on its own about 60% of the time. When the two anchors agree — which happens on ~75% of clear-regime days — they land on the realized side 65% of the time:
| anchor (conclamato regimes) | predicts VP-POC side | n |
|---|---|---|
| Charm magnet | 60% | 84 |
| Loading area (overnight gamma build) | 60% | 73 |
| Both anchors agree (~75% of days) | 65% | 51 |
Structure is universal; direction needs a clear regime. The two findings behave oppositely under the gamma regime, and the contrast is the point:
| property of the field | regime-dependent? | behaviour |
|---|---|---|
| Overlay — where volume concentrates | no | ~58% common mass in every regime, transition included |
| Direction — which side of the open | yes | 65% in clear regimes; breaks down in the transition zone |
Read physically: the shape of the field — where the walls are — is a structural fact that holds whatever the regime. The direction the volume then leans is only legible when the gamma regime is unambiguous; in the transition zone the same anchors carry no reliable directional information. This is an observation about the field's information content, not a prescription — the field says where in every regime, and which way only when the regime is settled.
04 Gallery
Each card below is a one-day snapshot showing three panels: GEX |netGex| (amber, horizontal) on the left, Volume Profile RTH (cyan, horizontal) in the middle, and the overlay with both POCs marked on the right. Click any card to enlarge.
Color legend at the bottom of each card. Gallery is grouped by month and refreshed monthly with new sessions.
05 Limits of the study
- Sample size 132. Coverage runs Feb 2026 – Aug 2026. The sample is comfortably above the n=15 threshold used in Study #001 and now clears the 100+ day mark, so the effect is well past first-look territory.
- RTH range restriction is essential. Without restricting to the day's actual high–low, the GEX profile spans ~2,600 ES points (popular round strikes), drowning the signal. The restriction is therefore part of the methodology — see §5 for details.
- Bin width is fixed at 5 points. Listed SPX strikes step by 5 in the relevant range; ES tick is 0.25. The 5-point grid matches the GEX granularity. Finer grids would over-fragment the VP into ~40× more bins than the GEX has values, biasing every overlap metric downward.
- Pre-open vs RTH-open is statistically indistinguishable. Re-running the study using the RTH_OPEN snapshot instead of PRE_OPEN gives near-identical metrics (HI ~57% either way). The structure the pre-open profile predicts is not refined by the first hour of trading flow — implying the dealer footprint at pre-open carries the entire information content.
06 Methodology
GEX profile extraction
From each PRE_OPEN snapshot we read plotData.data and select the trace whose name contains "Total Net GEX". Strikes (the trace's y array) and net-GEX values (x) are paired one-to-one. The absolute value of net-GEX is taken — sign carries directionality but not "presence", and we are measuring where the mass IS, not which direction it pulls. Strikes are tick-aligned to 0.25.
Volume Profile construction
For each RTH bar (09:30–16:00 ET, 5-minute timeframe) we distribute the bar's volume uniformly across the ES ticks between its low and high. This matches the convention used internally by the TLADe terminal's POC/VAH/VAL computation. Volume per tick = bar_volume / number_of_ticks.
Common-grid rebinning
Both profiles are aggregated onto a fixed 5-point grid. For each bin centre K, we sum:
- All GEX |netGex| values at strikes in [K − 2.5, K + 2.5)
- All VP volume at ticks in [K − 2.5, K + 2.5)
The grid is restricted to the strict RTH range [VPmin, VPmax] of the day, then both arrays are normalized to sum = 1.
Statistics
HI, BC, Cosine, and 1-D Wasserstein computed directly on the two normalized arrays p and q. POC for each profile is the bin centre with the maximum value within the restricted grid (not the globally maximal strike in the un-restricted GEX, which would bias toward round strikes outside the RTH).
Author: TLADe Research · contact
Related on TLADe: the PA + GEX framework guide describes how the Anchored VWAP layer relates to the GEX framework — see also Study #001 for the AVWAP-side investigation.
More research: TLADe Stats index.
This is descriptive research on historical market structure, not financial advice.