Direct answer
Information about Rate Of Change (ROC) can be verified by separating stable mechanics (the definition and calculation) from variable conditions (the data source, time interval, and any data preprocessing). Use a source hierarchy: verify the indicator definition first, then verify the calculation on your own spreadsheet or script using explicitly stated assumptions, and finally verify that different providers use equivalent inputs and formatting.
Mechanism and definition
Rate Of Change is a way to express how much a value changes over a chosen lookback period. “Rate” here is typically about change per unit of time steps, but in many technical indicator contexts it is implemented as a change relative to an earlier value (for example, a percentage change between two points).
To verify ROC information, you need three pieces that should not be assumed implicitly:
- What series is used (for example, a closing price series or another defined input).
- What time interval and lookback are used (the number of bars or the time difference).
- What exact formula variant is claimed (difference vs percentage difference; whether values are aligned by bar close, etc.).
If an article or provider states “ROC(14)” you should treat “14” as a material assumption: it defines which earlier data point the calculation references. If a source says ROC is “percent” but does not show the formula, verification requires you to locate the formula or to test it by reproducing results.
Evidence and reproducible example (no live data)
Use a small, hypothetical series so you can reproduce the calculation step by step.
Assumptions for this example:
- You have a time series of values by bar close: V[0], V[1], V[2], …
- You use a lookback of N = 3 bars.
- You test the common percentage-change form: ROC = (V[t] − V[t−N]) / V[t−N] × 100.
Verification steps:
- Choose a fixed t where both V[t] and V[t−N] exist in your sample.
- Compute the numerator: V[t] − V[t−N].
- Compute the denominator: V[t−N].
- Divide and multiply by 100 to match a “percent” claim.
- Check the sign convention: if the current value is higher than the lookback value, ROC should be positive; otherwise negative.
Independent cross-check method:
- If a provider’s ROC output is available in your environment, apply the same N and the same input definition to your own calculation and confirm the numbers match within rounding. If they do not match, the mismatch usually indicates a different formula variant, different bar alignment (for example, using intrabar values instead of bar close), or different handling of the first N bars (where ROC may be undefined).
Limitations and failure modes
Even with correct mechanics, ROC can differ across sources for non-obvious reasons:
- Data handling and alignment: Missing bars, differing time zone handling, or bar-close vs intrabar definitions can change which two values are used.
- Scaling and formula variants: Some sources use absolute difference instead of percentage change, or they may omit the ×100 scaling.
- Noise sensitivity: ROC amplifies short-term fluctuations because it depends on differences over a fixed lookback.
- Edge cases: When V[t−N] is zero (or near zero), percentage ROC becomes undefined or extremely large, depending on how the source handles division by zero.
Finally, historical relationships do not establish future behavior: even if ROC outputs appear to “work” in one period, you cannot treat that as a reliable predictor without assuming stability conditions that may not hold.
Verification or next question
If you want stronger verification, ask two concrete questions: (1) “Which exact ROC formula variant and input series definition are you using?” and (2) “How are the time interval, bar alignment, and missing-data rules handled?” Then reproduce the calculation yourself with explicit assumptions and compare results.
If two sources still disagree after you standardize the formula and inputs, treat the disagreement as evidence that their variable conditions differ (for example, time alignment or preprocessing), not as proof that ROC is inherently inconsistent.