Direct answer: what to measure
Variable spread is the spread that can change over time for the same instrument. You can measure it by capturing the spread components at defined timestamps and computing spread size statistics from those captured observations. For independent verification, record (1) the exact time of each observation, (2) the input prices used, and (3) the calculation method.
Mechanism or definition: measurable fields
To measure variable spread, choose a consistent data source and record the same fields each time:
- Quote-based measurements (standalone spread)
- Bid quote (B)
- Ask quote (A)
- Timestamp (T) for when the quotes are recorded
- Spread size: S = A − B
This method measures the spread present in the quotes at your chosen times. It does not guarantee what you will realize when you trade because fills depend on execution.
- Execution-based measurements (realized spread) If you have access to your own fills:
- Entry/exit timestamps
- Executed price(s)
- (If available) the prevailing bid/ask at execution time A common realized-spread approach is to compare executed prices to a reference price (often mid) at the same moment. If mid is not available, you must state the alternative reference explicitly.
Evidence or example: a calculation you can replicate
Assume you collect quote observations for one instrument over a session. At each timestamp Tᵢ you record Bᵢ and Aᵢ, then compute Sᵢ = Aᵢ − Bᵢ.
From the set {S₁, S₂, …, Sₙ}, you can compute measurable summaries such as:
- Average spread: mean(S)
- Maximum spread: max(S)
- Dispersion: for example, range(S) = max(S) − min(S)
- Tail behavior: proportion of observations above a chosen threshold (state that threshold)
To compare “variable” behavior across periods, split observations into time windows (for example, a volatile window vs. a calmer window) and compute the same summaries for each window. The validity of the comparison relies on consistent timestamping and the same calculation definition.
Limitations and risks: where measurements can fail
Even with careful logging, variable spread measurements have important limitations:
-
Timestamp and clock mismatch If your timestamps do not reflect the actual moment the bid/ask became available, your computed spread may not correspond to the intended market state.
-
Quote staleness and data gaps In fast markets, quotes can change quickly. Missing observations or delayed data can make variability appear smaller or larger than it truly was.
-
Quote spread vs. realized spread differences Quote-based spread does not automatically equal what you get in execution. Execution can differ due to order handling and liquidity at the moment of fill, so realized spread can diverge from S = A − B.
-
Costs and reporting choices If you compare numbers across providers or accounts, you must distinguish spread size from other costs that may be charged separately. Reporting conventions can vary, so comparisons require shared definitions and assumptions.
Verification or next question: how to validate independently
To verify your measurement method:
- State your chosen definition (quote spread S = A − B, or a realized-spread definition using executions and a defined reference).
- Record the exact timestamps and the exact input prices used for each observation.
- Use at least two time windows and compute the same summaries for each.
- Report assumptions (for example, how you aligned execution time to reference prices, and what you do when data is missing).
Next, consider measuring both quote-based spread and realized spread (if you have execution records) for the same time windows. Any divergence highlights a key limitation: variability in quotes does not fully determine variability in outcomes.
If you want, share what data you have access to (bid/ask quotes only, or also executed fills) and how timestamps are recorded, and the measurement fields can be tailored without changing the core definition of spread variability.