Market Structure

Across 1,092 sessions 2022-2026, NVDA decoupled: on >+10bp 10Y yield days, QQQ -0.25% and SMH -0.21% but NVDA +0.30%. The textbook 'yields up = tech down' rule is wrong about the leader.

The 'compression precedes breakout' pattern doesn't survive 8,688 sessions of data. Tightest days had a 6% breakout rate; widest had 26%. Momentum begets momentum. How our desk actually uses compression — and what we trade instead.

BABA's 60-day correlation to KWEB swings from 0.45 to 0.85 — single-name catalysts decouple from the basket more often than traders assume. The edge isn't a one-line ADR rule; it's identifying which signal is in control at the moment of execution.

SPY+QQQ opening-range breakouts hit target-before-stop 23% in contango vs 11% in backwardation across ~50 sessions. The term structure is a go/slow-down filter: inverted curves are fade days, not breakout days.

SPY 1-minute volume averages 251K shares at 09:30 and 277K at 15:55 — but only 57K at 12:30. Volume and range concentrate at open and close, and that's exactly where manual route control matters most. Smart routers work the middle.

Mid-size morning drives in US equities reverse during lunch 61–67% of the time; strong drives consolidate instead. The asymmetry is driven by dealer gamma, not coincidence — and the right play depends on which side you're on.

A 91-event study of post-earnings reactions in AAPL, MSFT, GOOGL, AMZN, META, NVDA and TSLA from 2023 to 2026. Classical PEAD inverts at the extremes — strong reactions fade, weak reactions bounce.

The framework's long-vs-short asymmetry has inverted in the current regime. 20-session data on 9 high-beta names: long ORB 31% baseline, short ORB 47% baseline, 50% with full confirmation. Short ORB is now the high-probability trade.

Global Weighted Order Flow Timing — SPY 1-min vol runs 2.66 bp/min at 09:30 (Europe still active), drops to 1.36 bp at 14:30 (no overseas), bounces to 1.70 bp into close. Session overlaps amplify US tape speed. Trade the overlap windows, sleep through the dead zone.

VIX term structure: contango (VIX3M > VIX) observed on 94% of sessions 2022-2026; backwardation on the 6% that mark stress regimes. Trend continuation works in contango; mean reversion outperforms in backwardation. Same setups, opposite expected values.

Basket Impact Ratio — single-name divergence from sector ETF predicts next-day return non-linearly. NVDA/AMD/AVGO vs SMH and META/AAPL vs XLK show mild lead (+1 to +3%) prints +0.15% next-day median; strong lead (>+3%) only +0.03% — exhausted.

Liquidity packets — bursts of participation that reveal accumulation or exhaustion before the candle confirms. CVD slope, refresh behavior at price levels, and trade-size distribution separate accumulation packets from exhaustion packets. The chart looks similar; the order flow doesn't.

The 09:45 dead zone — 30 minutes after the first opening sweep where a breakout proves acceptance or gets faded. Current regime favors downside breaks holding and upside breaks failing. Trade the failed retest, not the original break.