ReachRichInsights › Tick Data & Microstructure: Large-Order Classification & Lee-Ready

Tick Data & Microstructure: Large-Order Classification & Lee-Ready

Beyond daily and minute K-line, tick-by-trade data carries the most micro-economic information: per-trade time, price, volume, and direction. This post covers A-share tick structure, active buy/sell inference, and typical applications.

Data structure

A-share tick records typically contain: - time — matching time (millisecond precision) - price — trade price - volume — trade volume - side — buy/sell flag (some sources omit, need inference) - order_no / trade_no — order/trade IDs (for linking to order book)

Active buy/sell: Lee-Ready algorithm

When the source doesn't provide side (many A-share tick feeds have side=0), use the classic Lee-Ready (1991) algorithm:

  1. Price > midprice → active buy (uptick / "outside" trade)
  2. Price < midprice → active sell (downtick / "inside" trade)
  3. Price = midprice → use tick test (compare to previous tick)

Lee-Ready isn't 100% accurate but is statistically robust at scale.

Large-order classification

Category Size range (shares)
Retail < 4,000
Medium 4,000 - 20,000
Large 20,000 - 100,000
Super-large ≥ 100,000

Thresholds adjust by stock price (high-price stocks have lower thresholds). The "main capital flow" indicator is essentially "large + super-large net inflow minus outflow."

Applications

Data consistency is critical — all signals (backtest + live) should use the same tick feed.