Direct answer: what data is needed
To assess a Momentum Indicator in a way that another person can independently verify, you need four categories of information: (1) the exact definition of the indicator you are using, (2) the input price data it consumes, (3) the data provenance and timeliness details (where the data came from, how it was sampled, and in what time zone), and (4) the quality and consistency checks applied to that data. Because market data and indicator implementations can differ, the same named “Momentum Indicator” may not produce the same results unless these details match.
Mechanism and definition: inputs required to compute it
A Momentum Indicator is computed from a change in price over a chosen lookback period (often using a “current” value and an earlier value). To assess it accurately, document the following inputs:
- Price source: which series is used as the input (for example, close price, open price, high/low-derived values). If an implementation uses a different price than another, the computed momentum will differ.
- Lookback window and horizon: the number of periods used to measure the change (for example, “N bars”). The numeric lookback is a core parameter.
- Calculation rule: the exact formula (for example, whether the indicator uses a difference, a percentage change, or another transformation). Two formulas that both use “momentum” can behave differently.
- Output convention: the expected direction and scaling (for example, whether values are raw or normalized). This affects interpretation but does not change the underlying need for the inputs above.
Evidence or example: how to specify the data so it can be reproduced
A practical way to make assessment verifiable is to publish a “calculation card” for the indicator. For each run, specify:
- Dataset identity and sampling: the instrument definition (e.g., the trading symbol), the timeframe (e.g., hourly bars), and the number of rows used.
- Time zone and bar boundaries: whether timestamps are aligned to UTC or a local exchange time, and how day/hour boundaries are handled.
- Missing data policy: how missing bars are treated (e.g., dropped, filled, or left as gaps). Different policies change computed changes.
- Parameter values: the lookback window N and the chosen price source.
- Formula text: the exact computation rule.
With these items, an independent verifier can recompute the same indicator on the same dataset and check whether the values match. If values do not match, the mismatch will usually trace back to formula differences, price source differences, sampling differences, or data-quality handling.
Limitations and risks: what can fail even with correct data
Even if you collect all inputs above, there are material limitations to account for:
- Noise and sensitivity: Momentum based on short lookbacks can react to random fluctuations, and the behavior can change as market volatility changes.
- Regime shifts: The relationship between price changes and future outcomes is not stable across all conditions; historical patterns can be non-representative.
- Non-stationary inputs: When trading conditions change (liquidity, microstructure, or execution environment), the same formula may behave differently.
- Implementation inconsistency: Different platforms may interpret “momentum” differently (price source, percentage vs difference, rounding, or bar construction), producing different outputs from the same label.
These failure modes mean you should treat any indicator output as descriptive of the input series and formula, not as a standalone predictor.
Verification or next question: how to check you have the right data
To confirm that your assessment is grounded in verifiable facts, focus on reproducibility and consistency checks:
- Recompute: calculate the Momentum Indicator from the documented formula and parameters on the same input dataset; verify that computed values match the reference implementation.
- Consistency across timeframes: test whether results change as expected when timeframe changes, since sampling affects the lookback period in real time.
- Sensitivity checks: rerun using alternative but clearly defined data-quality handling (for example, different missing-value handling) to see how fragile the output is.
- Documentation audit: ensure every variable that affects the calculation—price source, lookback, formula, time alignment, and missing data policy—is explicitly stated.
If you want to go one step further, your next question should be: which exact momentum formula and price source are you targeting, and can you fully reproduce its values from a single, well-documented dataset?