Direct answer
To assess a “mean reversion range,” you need data that lets you (1) define what “mean” you are reverting to, (2) quantify how far price typically moves away from that mean, and (3) judge whether the relationship you observe is stable enough to measure. Because markets and data feeds differ, the main work is not finding a single number, but collecting inputs with clear provenance and applying quality checks.
Mechanism and definition: what you are measuring
A mean reversion range is a way to describe the distance of an observed price from a chosen central tendency (the “mean”) using a rule for typical deviation. The key idea is operational: you must state the assumptions behind the “mean” and the “range.” For example, a range could be defined using a rolling window where you compute a central value and then compute a deviation measure (such as an average absolute deviation or another dispersion measure), after which you set upper and lower bounds.
What data is needed for that?
- Price data for the instrument you are studying (at least open/high/low/close; closes are commonly used for central tendency and deviation).
- A defined sampling frequency (for example, daily, 4-hour, hourly) and a defined window length used to compute the mean and deviation.
- A consistent timestamp convention and a method to handle missing bars (because even small gaps can distort rolling statistics).
Evidence and examples: inputs, provenance, timeliness, and checks
1) Inputs you should collect
- Instrument price series: historical price data for the specific currency pair (or the exact series used).
- Timeframe settings: the sampling interval and the rolling window length used in the calculation.
- Definition parameters: the deviation rule (how “range” is computed) and the multiplier or threshold logic, if any.
- Data handling rules: how you treat outliers, missing data, splits/roll adjustments (if applicable), and whether you use bid, ask, mid, or another price construction.
2) Provenance: where the data comes from
To independently verify claims, you should record:
- Provider/source of the data (exchange feed, vendor, platform export, or another dataset) and whether it uses real-time or end-of-day processing.
- How the series is constructed (for instance, whether prices are adjusted, and whether the same field is used consistently across the whole period).
3) Timeliness: what “as of” means
Even without real-time data, you must document the data coverage:
- The start and end dates of the historical sample.
- Whether your assessment changes when you extend the window or move the endpoints.
4) Quality checks (what can break the calculation)
Perform checks that directly target measurement errors and model fragility:
- Missing data and irregular bars: confirm that the series is continuous for the timeframe chosen.
- Outlier sensitivity: verify whether a few extreme moves dominate the deviation measure.
- Regime shift checks: evaluate whether the central tendency and deviation change materially across different subperiods.
- Reproducibility: another analyst should be able to reproduce the same central tendency and bounds using your exact parameters and data.
A workable verification exercise (no predictions)
To verify that your “range” is a stable description, recompute it under slightly altered, documented assumptions:
- Move the rolling window length within a reasonable range.
- Test more than one dispersion rule (for example, an absolute-deviation-based rule vs. a standard-deviation-based rule).
- Compare how much the resulting bounds shift. This does not forecast future behavior; it shows how sensitive your definition is to choices.
Limitations and risks (material failure modes)
Mean reversion range assessments can fail even when calculations are correct. Common limitations include:
- Non-stationarity: the statistical relationship between price and its “mean” can change over time.
- Different microstructure: if you use one price definition (like end-of-day closes) for a concept intended for intraday behavior, the range may not represent the same dynamics.
- Cost and execution effects: spreads, commissions, and slippage can matter because realized movement often differs from theoretical price movements.
- Overfitting to history: tuning window length or deviation parameters on one period may produce a range that fits that past period but not other periods.
Verification or next question: a ready checklist
Use this checklist to ensure the assessment is independently checkable:
- AFV (assumptions, fields, values): clearly state the mean definition, deviation rule, window length, and sampling frequency.