Get Gex Gradient Signal

Gamma asymmetry around spot — above- vs below-spot dealer gamma skew.

Answers: is dealer gamma stacked above or below current price, and how does that bias the next move? Decomposes per-strike gamma exposure into four zones (above / below spot, ATM, wings) and scores the asymmetry, flipping sign and damping under long-gamma regimes.

• Heavy above-spot concentration + short gamma → any up-move unwinds → bullish. Below-spot mirror is bearish.
• Under long-gamma the same concentration acts as resistance instead of fuel, so the signal is inverted and damped.

Logic highlights (src/signals/basic/gex_gradient.py):

• Classify each strike in gex_by_strike by (strike − spot) / spot: above/below, plus ATM (≤ ±1.5%) and wing (≥ ±4%) tags.
• asymmetry = (above_abs − below_abs) / (above_abs + below_abs) ∈ [-1, +1].
• raw = asymmetry if net_gex < 0, else −asymmetry × 0.40 (SIGNAL_GEX_GRADIENT_LONG_GAMMA_DAMPING).
• confidence = max(0.25, 1 − wing_fraction) — wings pin, kill directional edge.
• score = clip(raw × confidence, [-1, 1]).
• Abstains (score=0) if total gamma < SIGNAL_GEX_GRADIENT_MIN_GAMMA (default 5e7 — thin-OI guard).

Params:

• symbol (default SPY). Returns 404 if no row exists yet.

Returns:

{
  "underlying": "SPY", "timestamp": "...",
  "clamped_score": 0.38, "score": 38.0, "direction": "bullish",
  "weighted_score": 0.0, "weight": 0.0,
  "above_spot_gamma_abs": 72500000.0,
  "below_spot_gamma_abs": 31200000.0,
  "asymmetry": 0.3986,
  "wing_fraction": 0.21,
  "context_values": {
    "source": "gex_by_strike",
    "above_spot_gamma_abs": 72500000.0, "below_spot_gamma_abs": 31200000.0,
    "atm_gamma_abs": 45800000.0, "wing_gamma_abs": 21900000.0,
    "above_spot_gamma_signed": 54300000.0, "below_spot_gamma_signed": -18700000.0,
    "wing_fraction": 0.21, "asymmetry": 0.3986, "strike_count": 42
  },
  "score_history": [{"score": 22.0, "timestamp": "..."}, "...up to 90"]
}

• score / clamped_score — [-100, +100] / [-1, +1].
• above_spot_gamma_abs / below_spot_gamma_abs — |Σ gamma| for each side; USD-scaled. null if source="unavailable".
• asymmetry — [-1, +1]; pre-regime adjustment. Matches score sign under net_gex < 0, flipped under net_gex > 0.
• wing_fraction — [0, 1]; share of total |gamma| at wing strikes (> ±4% OTM). Read wing_window_reached first. When it is false the ingested chain never reached ±4% (production defaults reach ±1.3% on SPX to ±3.0% on SPY), so wing_fraction is 0 because the bucket was unmeasurable, not because the wings were empty — and the confidence damper below is inert.
• wing_window_reached / max_strike_distance_pct — whether the wing bucket was reachable at all, and how far the widest surveyed strike sat from spot.
• context_values.atm_gamma_abs / wing_gamma_abs — absolute gamma in each bucket.
• context_values.above_spot_gamma_signed / below_spot_gamma_signed — signed sums.
• context_values.strike_count — int; strikes surveyed. Thin data → widen thresholds.
• context_values.source — "gex_by_strike" | "unavailable".

Trader interpretation.

• score > +50 with net_gex < 0 → classic short-gamma upside setup; dealers will chase. Favor calls.
• score < −50 with net_gex < 0 → short-gamma downside setup; dealers accelerate flush.
• Strong score with net_gex > 0 → structural resistance in that direction; fade instead of follow.
• wing_fraction > 0.5 → confidence already reduced by code; treat score as weak even if magnitude is high.
• strike_count < 10 → sparse data; widen threshold.

Page design. Four-zone horizontal strike map centered on spot: above- and below-spot buckets as mirrored bars, intensity by |signed gamma|. Separate ATM / wing donut slices showing concentration share. asymmetry gauge (−1 → +1). Regime chip (net_gex > 0 vs < 0) since the same asymmetry has opposite implications in each regime.