Backtest-informed

Spot in Wall Band — your position between the walls, at a glance

TL;DR
  • Spot in Wall Band shows where spot sits between the Put Wall (0%) and the Call Wall (100%) as a single percentage.
  • It is the single most predictive feature we identified for wall behavior — 27% of the model weight in our backtest of 5,992 events.
  • 7 zones from "Below Put Wall" to "Above Call Wall," each with a clear contextual reading.
  • Honest framing: our backtest AUC is 0.64. Walls are context, not signals. Combine with regime, flow, and price action.
  • Works on every ticker on /radar — pure geometry, no ticker-specific data required.

What is Spot in Wall Band?

The dashboard has always shown you the Call Wall, the Put Wall, and the current spot price as three separate numbers. To know where you actually were in the gamma structure, you had to do mental math: "spot is 6432, put wall is 6400, call wall is 6500 — so I'm 32 dollars above the put wall, out of a 100-dollar band — about a third of the way up."

Spot in Wall Band collapses that calculation into a single percentage and a visual bar. 0% means spot is at the put wall, 100% means spot is at the call wall, 50% is exactly halfway. A white diamond marker on a colored gradient shows you the position live, and a short reading underneath tells you what zone you're in and what to look for there.

It's a UI improvement on top of data the dashboard was already showing — but it changes how fast you can read positioning. Most experienced GEX users were doing this math anyway. Now the system does it for you.

The formula: Spot in Wall Band % = (spot − put_wall) ÷ (call_wall − put_wall) × 100. Pure geometry. No ticker-specific calibration, no historical lookup, no per-ticker hold rates baked in.

Why this metric — and why now?

This is the first GEXBoard feature derived directly from a backtest of our own historical data. In May 2026 we ran an internal study using LightGBM (gradient-boosted decision trees) on 5,992 historical wall touch events across SPX, SPY, and QQQ between March and May 2026. We fed the model 37 candidate predictors of whether a wall would hold or break within a 30-minute window after a touch event.

Position-in-band came out as predictor #1 with 27% of the model's total predictive weight — more than the next two features combined (regime-strength interaction at 14% and minutes-since-open at 9%), and more than four times the predictive weight of raw distance-to-wall (which came in 10th at 2.3%).

In plain English: where spot sits inside the band is more informative than how close spot is to any individual wall. That counterintuitive finding is what justified shipping this card.

Honest about what it's not: our out-of-sample AUC on the binary classification (held vs broken in 30 minutes) is 0.64. That's significantly better than random (0.50) and in line with what serious quant-finance literature shows for intraday market-behavior prediction — but it is not a magic predictor. Walls are context. Anything in this space claiming AUC above 0.70 usually has data leakage. We surface 0.64 because that's what the data honestly supports.

The 7 zones and how to read each one

The card classifies your current position into one of seven zones. Each zone has a short reading underneath the number that tells you what's interesting about being there. Memorize the zone names — they map directly to what you should be watching for.

Zone% of bandWhat it means · what to watch
Below Put Wall < 0% Spot broke below the put wall. Structural support has failed. Trend continuation is more likely than reversal until a new put wall forms at a lower strike.
Put Wall Zone 0–20% Spot is testing the put wall. Either bounces off it (regime stabilizes) or breaks down (opens room lower). This is the moment to watch flow + price action closely.
Lower Half 20–45% Spot is biased toward the put wall but not at it. Downside pressure side of the range. In long gamma, mean-reversion toward mid is statistically favored.
Mid-Band 45–55% Spot is balanced between walls. Neither side has gravitational priority. Range-bound bias unless flow or news pushes one direction.
Upper Half 55–80% Spot is biased toward the call wall. Upside resistance side of the range. In long gamma, fade-the-rally back toward mid is statistically favored.
Call Wall Zone 80–100% Spot is testing the call wall. Either rejects from it (regime stabilizes) or breaks out (opens room higher). Same critical moment as the put wall zone — watch flow + PA.
Above Call Wall > 100% Spot broke above the call wall. Structural resistance has failed. Trend continuation is more likely than reversal until a new call wall forms at a higher strike.

How to use it in your workflow

The card is meant as context, not a signal. Three rules of thumb based on what we observed in the backtest:

  1. Combine with regime. A spot at 85% (upper half / call wall zone) means very different things in long gamma vs short gamma. In long gamma the call wall is more likely to hold (mean reversion likely). In short gamma the call wall is more likely to break (trend continuation likely). The dashboard's Net GEX card and dealer regime label tell you which regime you're in.
  2. Combine with flow. Position alone doesn't tell you intent. If spot is at 22% (lower half) and flow is heavy put-buying, dealers are likely being pushed into amplifying the move down. If spot is at 22% but flow is heavy call-buying, dealers are likely starting to defend the level for a bounce.
  3. Confirm with price action. Walls are mechanical levels. Price action on a chart shows you whether the level is actually being respected in real time. Use TradingView (or our TradingView indicator) to confirm whether the structural level is being tested with conviction or being faded.
💡
A typical 0DTE intraday read: spot at 25% of band + long gamma regime + flow buying calls = probable mean reversion toward mid (60%+) over the next 30–90 minutes. Spot at 25% + short gamma regime + flow buying puts = continuation lower, with put wall as the next target. The card gives you one of the four inputs; the dashboard already gives you the other three.

Which tickers does it work on?

Every ticker on /radar. Because the calculation is pure geometry between the existing walls, there's no per-ticker calibration needed. SPX, SPY, QQQ, NVDA, TSLA, AAPL, MSFT, META, AMZN, AMD, IWM, GLD, TLT, DIA, and every other listed ticker show the same percentage scale.

The card switches automatically when you change ticker or DTE bucket. Cross-asset comparison is meaningful: "SPX is at 85% and QQQ is at 12%" is a divergence signal worth investigating.

Limitations and honest disclosures

The backtest sample is 2 months of data (March 17 to May 18, 2026). That's enough to identify the direction of the effect with statistical confidence, but not enough to nail down precise hold rates per ticker, per regime, per time of day. Specifically:

  • Per-ticker hold rates vary meaningfully (SPY 45% / SPX 38% / QQQ 22% for call walls in short gamma regime) — that's why we don't ship per-ticker confidence scores yet.
  • Some bucket combinations (specific time slots × regime × wall side) have small sample sizes (n < 50) — those don't justify hard probability claims.
  • The model AUC is 0.64. That's not 0.90. Don't trade this as a binary signal.
  • 0DTE specifically only has ~1 week of dedicated history (prewarm started May 11), so 0DTE-specific patterns are still being learned.

What's next on the roadmap

Spot in Wall Band is the first backtest-derived feature we've shipped — the lowest-risk, highest-confidence finding from the study. Several stronger features are being held back until we have more data to validate them safely:

  • Counter-regime wall strength badges — early backtest signal suggests walls counter to the dominant regime (call walls in short gamma, put walls in long gamma) hold more often. Numbers vary too much per ticker right now to ship as a visible badge — waiting for 3–6 more months of data to confirm magnitude per ticker.
  • Hold-rate confidence scores per wall — eventually each wall would show a small percentage indicating its historical hold rate in similar conditions. Needs reliable per-ticker, per-regime baselines.
  • Time-of-day overlays — the third strongest predictor in the backtest was minutes-since-open. Open and close exhibit different patterns than mid-session. A "session-phase" indicator may surface eventually.
  • Personal Studio mode — for advanced traders who already have their own framework, a configurable rules engine that evaluates the market against their definitions (not ours). Early-stage research, no commitment yet.

The pattern: ship only what the data clearly supports, frame it honestly, watch how users actually use it, and let signal stability across more months of data dictate what gets shipped next. The backtest pipeline is committed to git and re-runs as the dataset grows.

💬
Your feedback shapes priority. If you're using Spot in Wall Band and notice something — a zone that doesn't match your read, a ticker where it feels off, a context layer you wish was there — reply to any of our emails or hit /support. Concrete feedback from active users moves features up the queue faster than internal speculation.