Trade Signing And The Lee-Ready Rule
Most historical trade data doesn't record whether a trade was buyer-initiated or seller-initiated. The Lee-Ready rule infers it from the trade price's position relative to the quotes, and gets it right most, but not all, of the time.
Prerequisites: Order Book Mechanics, Bid-Ask Bounce
A trade prints at 100.01. Was it a buyer aggressively hitting the ask, or a seller aggressively hitting the bid? For strategies that need to know order flow direction — is this stock seeing net buying or net selling pressure — this matters enormously, but a huge amount of historical and even some live trade data simply doesn't tag which side initiated. The Lee-Ready rule is the standard way to infer it after the fact, using only the trade price and the prevailing quotes.
The rule has two steps, applied in order. First, the quote test: if the trade price is above the midpoint of the bid and ask, classify it a buy; if below the midpoint, classify it a sell. Second, if the trade prints exactly at the midpoint (so the quote test is a tie), fall back to the tick test: compare this trade's price to the previous trade's price — if it's higher, call it a buy (an "uptick"); if lower, call it a sell (a "downtick"); if unchanged, look further back to the last price that did differ.
Worked example. Quotes are 100.00 bid / 100.04 ask, midpoint 100.02. A trade prints at 100.03. Since 100.03 > 100.02, the quote test classifies it a buy — someone paid up, close to the ask, to get filled. Next trade prints at 100.01. Since 100.01 < 100.02, the quote test classifies it a sell.
Second example, the tie case. Quotes are 100.00 bid / 100.04 ask, midpoint exactly 100.02. A trade prints at 100.02 — dead on the midpoint (this happens when a large hidden or midpoint-pegged order executes). The quote test can't decide. Fall back to the tick test: the previous trade printed at 100.01. Since 100.02 > 100.01, this is an uptick, so the trade is classified a buy.
What this means in practice
Order-flow-imbalance signals, PIN estimation, and most academic microstructure research that predates exchange-provided liquidity flags all rely on Lee-Ready or one of its variants (the Ellis-Michaely-O'Hara rule, the Chakrabarty-Li-Nguyen-Van Ness rule) to sign trades. Accuracy against ground truth is typically cited around 80–90% for liquid names on standard equity data, dropping meaningfully around fast markets, quote flickering, and trades that occur off the primary exchange's quote.
Lee-Ready: trade above the quote midpoint is a buy, below is a sell; exactly at the midpoint, use whether the price ticked up or down from the last differing trade.
The rule was designed for the quoting conventions and reporting lags of 1990s NYSE data. In modern fragmented, sub-millisecond markets, the "prevailing quote" at the moment of a trade is ambiguous — which venue's quote, with what lag — and misalignment between trade timestamps and quote timestamps is a bigger source of misclassification than the rule's logic itself. Always check what lag convention (if any) is applied before trusting Lee-Ready output on modern data.
Signed trades from this rule are the raw input to Order-Flow Imbalance and to PIN: The Probability Of Informed Trading; see Cleaning High-Frequency Tick Data for the quote-alignment issues that determine how reliable the classification actually is.
Related concepts
Practice in interviews
Further reading
- Lee & Ready (1991), Inferring Trade Direction from Intraday Data