How we process institutional options flow.

Profit Builders consumes the live OPRA options tape over Polygon's WebSocket feed. Trades arrive in real time with sub-millisecond timestamps, exchange identifiers, condition codes, and per-fill size. Every print is written to a normalized PostgreSQL store with the raw condition-code array preserved.

A trade is classified as a sweep when 2+ exchanges fill within a ≤500ms window — the threshold defined in CBOE Rulebook Section 6.11 for Intermarket Sweep Order semantics. Block trades are single-fill prints above the premium floor. Aggregation windows widen for larger sweeps (12s for 15+ fills) so 20+ leg institutional orders aren't split into fragments.

Each tape print carries one or more OPRA condition codes. Profit Builders surfaces the institutional-aggression badges for prints tagged ISO (code 219 — Reg-NMS Intermarket Sweep Order, equity-only by routing structure) and CROSS (code 229 — broker internalization, the closest options analog to dark-pool flow). Multi-leg auto/auction codes (232–239) are decoded and exposed as structure tags rather than gates.

Aggression is computed against the NBBO snapshot at trade time (15-second staleness cutoff). The spread-relative methodology follows Hasbrouck (2004) on effective spread — prints in the upper 35% of the bid-ask spread are tagged AT_ASK, lower 35% AT_BID, middle 30% NEUTRAL. Above-the-ask and below-the-bid prints are flagged separately as aggressive.

For symbols where Polygon does not return live Greeks (typically index options like SPX/SPXW/NDX/RUT/VIX), Profit Builders computes them via Black-Scholes-Merton with a Newton-Raphson IV solver per Corrado & Miller (1996). ETF options apply continuous dividend yield q in the d₁ formula — d₁ = [ln(S/K) + (r − q + σ²/2)T] / (σ√T) — with delta = e^(−qT)·N(d₁), per Haug (2007).

Vol/OI ratios use prior-day open interest from the Polygon EOD snapshot — the institutional standard. Intraday OI estimates are avoided because they introduce sample-path bias and aren't available cleanly through any vendor at the per-contract level.

The GEX heatmap surfaces dealer hedging walls strike-by-strike, expiry-by-expiry. We compute net gamma exposure per cell as γ × OI × spot² — the public SqueezeMetrics dollar-gamma convention — assuming the standard worst-case dealer position (short calls, long puts) when no flow attribution is available.

What this means: wall direction is robust — the strikes our heatmap flags as Max +GEX (call wall) and Max -GEX (put wall) are computed from real Polygon greeks during regular trading hours, restricted to a ±3% band around spot for actionable relevance, and align with the price levels where dealer delta-hedging activity concentrates. Wall magnitudes are convention-dependent— tools that layer in retail-vs-institutional flow inference (e.g., SpotGamma's GEX 2.0) will report different absolute dollar values for the same underlying chain. We use the public methodology because it's reproducible from the Polygon greeks any user can verify; we do not claim parity with proprietary models.

After-hours, when Polygon greeks aren't streaming, individual cells fall back to a Gaussian-shaped gamma estimate centered on ATM. Those cells are visually annotated (dashed amber border) and the heatmap header surfaces an ESTIMATEDpill so users can't mistake synthetic numbers for live greeks.

Every signal that enters the database carries a scorer_version stamp identifying which ruleset produced it. Conviction grading is an evolving internal layer — we will not publicize aggregate win rates that mix grader versions, since the metric is non-stationary when rules change. Forward performance will be reported per-version once a stable ruleset has accumulated a meaningful sample.