Liquidity Is Not There When You Need It

An order-book snapshot answers a narrow question: what limit orders were displayed when the snapshot was taken? It does not answer how much can be executed after an order reveals direction, consumes queue priority, and causes other participants to update. The distinction is temporal and endogenous. The book changes because the trade attempts to use it.

The central confusion this research tries to dissolve is the gap between displayed liquidity (the resting size you can see) and executable liquidity, the size you can actually trade against at a known price before the rest of the market reacts. They coincide on quiet days, which is exactly why they are so easily confused. They diverge under stress, which is exactly when the distinction matters.

Section 01Two kinds of liquidity

A resting limit order is a contingent commitment governed by price-time priority and the venue's cancellation rules. The familiar "free option" analogy captures the exposure to informed flow, but it is not literal: an order that is reached while live must execute according to the rulebook. Displayed depth is nevertheless provisional because orders ahead of the incoming trade can cancel or reprice before they are reached. Hidden and iceberg interest complicates the inference in the other direction, since executable size may also exceed what is visible.

A large order therefore does more than consume static levels. It changes the information set of every observer and alters the state it was meant to measure. The pre-trade book is a benchmark; the execution path is a response.

Section 02Why it evaporates

Adverse selection is one reason liquidity becomes more expensive under pressure. Glosten and Milgrom showed that a spread can arise even for risk-neutral competitive market makers when some incoming orders are informed. As the conditional probability of informed flow rises, a wider spread or smaller quote can be rational. Inventory risk, volatility, capital constraints, and venue mechanics add further channels, so withdrawal should not be attributed to information asymmetry alone.

Kyle's model gives the companion intuition for impact. In a market with an informed trader hidden among noise, price moves in proportion to net order flow, and the constant of proportionality (the market's depth, often written as Kyle's λ) measures how much each unit of imbalance pushes the price. Liquidity, in this language, is simply a small λ. The uncomfortable corollary is that λ is not fixed: it is largest, meaning depth is smallest, exactly when informed flow is most likely. Layer in queue position (the reality that at a given price you are behind other orders, and may never be filled before the level is cancelled or traded through) and "the size was there" becomes a statement about a world that no longer exists by the time your order arrives.

Displayed depth is a snapshot of options that exist only until someone decides not to honour them. The book you traded into is not the book you saw.

On the limit order book

Section 03The stress case

The clearest illustrations come from venues usually considered among the deepest in the world. On 15 October 2014, the U.S. Treasury market (the benchmark for "risk-free" liquidity) experienced an extraordinary intraday round trip. According to the official Joint Staff Report by the U.S. Treasury, the Federal Reserve Board, the Federal Reserve Bank of New York, the SEC and the CFTC, the yield on the 10-year note swung through one of the largest intraday ranges in decades, with a sharp drop and rebound compressed into a window of minutes, in the absence of any single obvious news catalyst. The report documented a decline in order-book depth and changes in order flow and liquidity provision around the event: a deep market thinning out under its own internal dynamics rather than an external shock.

The lesson generalises beyond any one episode. Liquidity is procyclical: it is supplied generously when volatility is low and the cost of being wrong is small, and it is withdrawn when volatility rises and that cost grows. This is not a malfunction to be regulated away; it is the predictable behaviour of rational liquidity providers protecting themselves from adverse selection. But it means the liquidity in your risk model (usually measured on calm days) is systematically the liquidity you will not have on the day you need to trade.

implementation_shortfall
    = signed_quantity * (realised_fill - decision_price)
      + fees

drivers include:
    spread + delay + impact(size, participation, depth) + opportunity_cost

Section 04What backtests miss

Almgren and Chriss framed the execution problem as a trade-off between two costs that pull in opposite directions: trade quickly and pay market impact, or trade slowly and accept the risk that the price drifts away while you wait. A backtest that fills instantly at the midpoint pays neither, and therefore overstates the edge available to a live strategy. The error is directional rather than random: it is biased toward optimism and is often most consequential for high-turnover, short-horizon strategies.

The liquidity in your risk model is the liquidity you will not have on the day you need to trade.

The central contribution is to treat liquidity as a response function rather than a stock. Executable size depends on order size, urgency, information content, participation rate, and the state of competing liquidity providers. Research that models only average spread and displayed depth estimates the market before the strategy arrives. Execution modelling begins when the strategy is allowed to change that market.

This research is analysis and commentary for general information. It is not investment advice, an offer, or a solicitation, and it contains no price forecasts. Reported events are attributed to their sources; the interpretation is the author's.