Across 5,365 sessions on SPY/QQQ/NVDA/TSLA/AMD, exhaustion-close rate climbs from 27% on calm-volume days to 49% at 1.8-2.5x volume. High relative volume is participation, not sponsorship — and order flow shows which we're looking at.

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 1-minute range medians 4.9 bp at the open vs 2.2 bp at the lunch low. Every bp retail routing surrenders compounds during exactly the windows where active traders make money. DMA isn't a status symbol — it's the layer that lets traders choose interaction instead of inheriting the average.

A 100% annualized borrow rate is a 0.27%/day P&L tax. A 5-day hold burns 1.37%; a 21-day hold burns 5.75% — often the entire expected alpha. The complete framework for trading HTB names: borrow math, Reg SHO, term vs open borrow, recall risk, and the desk decision tree.

UPST sits at 32% short of float. BYND at 28%. AMC at 18%. One HTB vendor is a single point of failure on every one of these names. The operational case for four-vendor redundancy: the trade you're missing is the trade with the entire desk's prep behind it.