The calm voice in the storm: it started with liquidity
Before contagion, before panic, before the headlines: there was a tightening in funding markets, a silent withdrawal from risk. Those who noticed the shift in flows were already preparing for impact.
In today’s high-frequency markets, liquidity can disappear in seconds. One moment, spreads are tight and depth looks healthy; the next, order books thin, and prices tumble with little resistance. Traditional liquidity metrics struggle to capture these rapid shifts — that’s where our new research steps in.
In our latest paper, “Quote Volatility Pressure,” we introduce a machine learning-based illiquidity measure built on a concept we call Quote Volatility (QV) — a metric designed to detect and quantify short-term bid-ask oscillations in real time.
Conventional liquidity measures often rely on trade data or quoted spreads, but these can mask instability when market makers pull back or algorithms reposition simultaneously. Our Quote Volatility metric shifts focus to microsecond-level quote movements, revealing when market depth starts to tremble before visible price moves occur.
Trained on historical episodes of order book thinning and rapid price reversals, the QV-based model learns nonlinear relationships between depth, volume, and volatility — uncovering stress patterns invisible to static measures.
Across multiple global markets, spikes in Quote Volatility emerged as leading indicators of systemic liquidity stress:
🦠 March 2020: The Covid-19 selloff saw QV surges align with the collapse of E-mini S&P 500 liquidity during “Black Monday” and “Black Thursday.”
⚔️ Early 2022: The onset of the war triggered synchronized quote instability across European futures.
💥 August 2024: Japan’s Nikkei crash revealed extreme QV peaks and a 12.4% index fall — the largest in years — with no major news catalyst.
📉 April 2025: The renewed tariffs trade war reignited volatility pressures, marking another sharp QV escalation as global risk sentiment fractured.
In contrast, QV remained stable during normal market periods, highlighting its precision in distinguishing genuine liquidity stress from routine volatility.
The Quote Volatility Pressure (QVP) model leverages supervised learning on labeled episodes of market stress. By analyzing thousands of microprice and quote update patterns, the model captures:
Intraday depth imbalances that precede rapid price swings,
Order book withdrawal patterns during deleveraging, and
Momentum amplification by algorithmic trend-followers.
This enables QVP to serve as a real-time liquidity fragility index, offering predictive signals for execution risk, trading costs, and potential price cascades.
For arbitrageurs and market makers, QVP enhances risk management by signaling when spreads are likely to widen and fills may become costly.
For institutional investors, it provides a live gauge of execution conditions and hidden liquidity costs.
And for regulators and exchanges, QVPM offers an early-warning system for liquidity-driven flash crashes — enabling intervention before disorderly trading unfolds.
The message is clear: modern markets are no longer just about volatility — they’re about liquidity elasticity. When depth disappears, volatility becomes nonlinear, and systemic risk surfaces almost instantly.
By capturing this microstructure dynamic through Quote Volatility Pressure, we take a step toward a more transparent and resilient market ecosystem — one where liquidity stress can be monitored, predicted, and mitigated in real time.
🔗 Read the paper: “Quote Volatility Pressure and Intraday Momentum”
A new machine learning framework for real-time liquidity monitoring in high-frequency markets.