Direct answer
Spread by session can be measured by choosing (1) a precise definition of spread, (2) session windows with an explicit time zone, and (3) a repeatable measurement method that records bid and ask at consistent timestamps. The result is a set of spread statistics per session (for example, average spread, median spread, or spread distribution), computed from the same measurement rules.
Mechanism or definition
Spread is typically the difference between the best ask price and the best bid price at a specific moment: spread = ask − bid. To measure spread by session, you first define what a “session” means. Common session partitions are based on market geography and trading hours (for example, Asia/Europe/North America), but whatever split you use must be explicit: start time, end time, and time zone.
Next, define your measurement fields and timestamps. A practical approach is:
- Inputs to record: the best bid and best ask (or the quoted spread, if that is the only available data), plus timestamp, plus instrument.
- Timestamp rule: either measure at fixed intervals (for example, every N seconds) or measure at predefined times within each session.
- Repeatability rule: use the same sampling rule across all days and all sessions.
Then compute per-session statistics such as:
- Average or median spread (central tendency)
- Percentiles (how often spreads are wide)
- Count of valid samples (data completeness)
Evidence or example
Assume you define three sessions and use a single time zone (for example, UTC). For each trading day, you sample the best bid and best ask every 60 seconds during each session window, and compute spread = ask − bid for every sample.
To compare sessions, you do not compare raw numbers from different conditions without controlling assumptions. Instead, you compare statistics computed from the same sampling rule:
- Session A median spread on Day 1…Day 30
- Session B median spread on Day 1…Day 30
- Then summarize across days (for example, median-of-day-medians, or average-of-spread-means)
This structure lets a reader independently verify what you measured: the session boundaries, the timestamp sampling rule, and the spread formula.
Limitations and risks
Several failure modes can make “spread by session” misleading:
- Market condition confounding: liquidity and volatility change over time. Even with identical session labels, one month may be calmer than another.
- Execution vs. quote mismatch: measuring bid/ask quotes does not ensure you could actually transact at those prices, because execution quality can differ.
- Data source differences: two providers may record “best bid/ask” differently or include different quote lifetimes; missing data can bias results toward quieter periods.
- Clock and time zone errors: a small timestamp misalignment can shift samples into the wrong session.
- Costs and route effects: some “spread-like” costs may be visible only after considering commissions, execution venue behavior, or other trading costs.
Verification or next question
To verify spread-by-session claims, check that the measurement method is fully specified: session start/end times with time zone, the spread definition used (ask − bid or quoted spread), and the timestamp sampling rule. Also verify sample counts per session and whether missing data was excluded or handled consistently.
A useful next question is: Are you measuring quote spread or realized execution spread? If you only have quotes, your measurement reflects quoted liquidity conditions, not necessarily the spread you would experience when orders are executed.