Direct answer
Liquidity by session is measurable when you turn “liquidity” into specific, observable metrics and you attach those metrics to clearly defined session time windows. You then record values at agreed timestamps (or aggregates over the session) and compare results using the same rules each time.
Because this is an observational concept, not a single universal indicator, you should expect that measurements differ by data source, execution venue, and market regime. The goal is not to predict prices, but to estimate how easily trading can occur during different parts of the trading day.
Mechanism or definition
A practical way to measure liquidity “by session” is to choose a liquidity metric that can be computed from available market data, then compute it for each session window.
Common measurable fields include:
- Traded activity: total traded volume or notional during a session. This is often easiest to compute, but it does not guarantee tight spreads.
- Price impact proxy: how much mid-price moves after a specified amount of aggressive trading (requires tick or event-level data).
- Quoted spread: bid-ask spread at specific timestamps or averaged over a session.
- Order-book depth (if available): summed size within set distance levels around the mid-price.
Define your session windows in a way that can be reproduced, for example by using a fixed start/end time in a chosen time zone (and documenting that choice). You then compute each metric either:
- At fixed timestamps (e.g., at minute marks), or
- As aggregates (e.g., mean spread over the window, or total volume across the window).
If you compare sessions across days, also define how you aggregate across days (median vs average), because liquidity can be highly variable.
Evidence or example
Consider a simple measurement design using two metrics: average quoted spread and traded volume.
Assumptions you must state:
- You use the same time zone for all session windows.
- Your data source provides consistent definitions for volume and bid/ask.
- You measure spread using the same calculation each time (e.g., bid minus ask converted to a chosen unit).
Procedure (non-trading, measurement-only):
- Pick session windows (for example, “Session A” and “Session B”) with fixed start/end timestamps.
- For each day in your sample, compute the average spread across every timestamp in the window, and compute total volume across the same window.
- Compare sessions using a summary statistic across days (e.g., median difference in spread).
A second design option is to add a price impact proxy. You would specify a fixed “aggressive size” threshold and measure the resulting mid-price change. This can better reflect execution difficulty, but it requires more granular data and careful assumptions about what counts as aggressive trading.
You can also align this measurement with another independent liquidity proxy (like depth near the mid-price, if order-book data exists) to reduce the risk that one metric is misleading for a particular market regime.
Limitations and risks
Several material limitations can cause “liquidity by session” measurements to fail as a stable comparison.
1) Data-source and venue differences Quoted spreads and depth can differ across venues or data feeds. Even if you label sessions identically, the underlying liquidity can be different for the instrument and venue your data represents.
2) Variable market conditions Volatility, news timing, and macro events can compress spreads or widen them independent of time-of-day. Historical relationships do not guarantee future similarity.
3) Costs and execution effects Quoted liquidity is not the same as execution liquidity. Real execution depends on order type, available depth, latency, and fees. Your measurement must clearly state whether it reflects quotes (availability) or realized trading experience (which includes costs).
4) Aggregation choices can hide outliers A session average can mask brief liquidity gaps (for example, during sudden spikes). Using only averages may understate variability.
Failure mode to plan for: you measure “high liquidity” in terms of volume, but spreads and depth are actually poor during the same window for your data source. In that case, your metric selection leads to a misleading interpretation.
Verification or next question
To independently verify your conclusions:
- Recompute the same metrics using a different aggregation rule (e. g. , median vs mean) and check whether the session ranking changes.