Direct answer
Information about performance statistics can be verified by (1) clarifying the definition of each metric, (2) confirming the inputs and time window, (3) reproducing the calculations from raw data when available, and (4) testing how results change when you include costs, fees, or alternate assumptions. Historical results can be checked for internal consistency, but they do not automatically establish future outcomes.
Mechanism and definition
“Performance statistics” are summary numbers that attempt to describe how an activity performed over a period. Examples include returns over time, volatility measures, drawdown measures, and risk-adjusted ratios. Verification starts with definitions:
- Identify the metric formula (how it is computed), not just the label.
- Identify the scope: the exact period, the data source, and whether results are net or gross of costs.
- Identify the basis: whether calculations use price changes, executed trade outcomes, or reported account equity.
Stable mechanics are the parts that do not depend on the market’s future path. Variable conditions include market regime, execution quality, latency, order types, and costs such as spreads, commissions, and financing. When someone reports performance statistics, they are implicitly combining both stable calculation mechanics and variable conditions.
To verify, list the assumptions you must accept before any number can be trusted—for instance, what counts as a “month,” whether weekends or holidays are included, and how cash flows (deposits/withdrawals) were handled.
Evidence and reproducible checks
A reproducible verification workflow can be performed without real-time data:
- Confirm the time window and granularity: record start/end dates and the sampling frequency used for returns (daily, weekly, monthly).
- Recompute a small set of values: pick one metric and one short segment. Use the stated formula and the provided time series, if available.
- Check net-vs-gross handling: verify whether the reported statistics account for costs. If the dataset does not include costs, repeat the calculation under a “no-cost” assumption and note what changes when costs are introduced.
- Validate internal consistency: ensure that drawdown measures align with the reported return path, and that risk measures derived from returns match the same return series.
- Run sensitivity tests: change only one assumption at a time, such as how time zones are treated or whether dividends/financing are included, and observe whether the ranking or conclusions flip.
If raw inputs are not available, you can still verify limited properties: whether the numbers are computed from consistent units, whether they match the claimed time window, and whether multiple reported metrics correspond to the same underlying return series.
Limitations and risks (material failure modes)
Several failure modes can make performance statistics misleading even when calculations are arithmetically correct:
- Missing costs: statistics may exclude spreads, commissions, or financing, overstating net performance.
- Inconsistent reporting windows: changing start dates or filtering out difficult periods can inflate results.
- Survivorship and selection bias: only successful performers may be shown, hiding how typical cases behave.
- Equity adjustment issues: deposits and withdrawals can distort return calculations if not treated correctly.
- Historical non-predictiveness: historical relationships do not guarantee future results; market conditions and execution can change.
Because outcomes vary with market conditions, costs, execution, and jurisdiction, treat verification as a way to confirm internal consistency and assumptions, not as proof of future performance.
Verification checklist and next question to answer
Before using any performance statistics, confirm these points:
- What exact metric definitions and formulas were used?
- What inputs were used (return series or equity series), and were they net of costs?
- What time window and sampling frequency were used?
- Can you reproduce at least one metric from the same inputs with explicit assumptions?
- Which limitations could materially change interpretation (costs, bias, cash-flow handling)?
A useful next question is: Which assumptions must be true for the reported statistics to be comparable to other reported statistics? If you cannot answer that precisely, the numbers may not be meaningfully comparable.