What data is needed to assess WMA
To assess a Weighted Moving Average (WMA), you need four groups of information: (1) the concept and calculation definition, (2) the input time series and how it is sourced, (3) the timeliness and alignment of the data with your timestamps, and (4) quality checks that confirm the inputs are consistent and usable. Without these, two people can compute “the same WMA” differently and still be correct according to their own conventions.
Mechanism and definition: what WMA calculations require
A WMA is a moving average where more recent observations typically receive higher weight than older ones. To compute it, you must specify the data series (for example, a price series), the window length (number of periods), and the weighting scheme (the exact rule for weights).
Material inputs:
- The underlying time series you average (commonly a sequence of prices). Decide whether you are using closing prices, mid prices, or another field.
- Window length (N): how many past periods are included.
- Weight rule: how weights are assigned across the N periods (for example, a linear scheme where weights increase for newer observations). “Weighted” can mean many specific conventions, so the rule must be explicit.
- Calculation convention: whether weights are normalized (so they sum to 1) and how rounding or decimals are handled.
Stable mechanics vs variable conditions:
- Stable mechanics: once N, the weight rule, and the series are fixed, the computation is deterministic.
- Variable conditions: results vary with the market regime, the selected timeframe, the data source’s field definitions, and any data processing steps (such as how timestamps are recorded).
Evidence or example: what to check before trusting a WMA
A practical “assessment” is not only producing a number; it is verifying that the number is derived from clearly stated, consistent inputs.
What you should verify (independent checks):
- Provenance of the input series: Is the price data from the same instrument definition across the entire history you use? If the provider splits/rolls contracts or changes instrument definitions, the time series may not be comparable.
- Timeliness and time alignment: Are all observations aligned to the same timezone and bar boundaries? Inconsistent timestamps can shift which values fall inside the window.
- Consistency of the field: If one workflow uses “close” and another uses “last” or “mid,” the WMA values will differ even with identical N and weights.
- Handling missing or irregular data: If some periods are missing, you must know whether they are omitted, forward-filled, or treated differently. Each approach changes the effective calculation.
- Parameter disclosure: Record N and the weight rule exactly. A “shorter window” and a “different weighting” are not minor tweaks; they change the responsiveness of the average.
Example assumptions you must state:
- If you compute a WMA on a 10-period window using a specific weight rule, then you assume you have a complete 10-period sequence for each computed point. If your dataset has gaps, the example no longer matches reality.
Limitations and risks: where WMA assessment can fail
At least one important limitation is that a WMA is sensitive to its chosen inputs. If you change the window length, the weighting rule, or the price field, the WMA can change materially. This sensitivity is not an error; it is an expected feature.
Common failure modes:
- Parameter mismatch: Someone may use a different weight convention (or normalization) and obtain a different curve while still claiming it is a WMA.
- Data quality issues: Outliers, missing values, or inconsistent instrument definitions can distort the weighted average.
- Interpretation risk: A moving average reflects a smoothing transformation of past data; historical relationships do not establish how it will behave in the future.
Broader uncertainty: Even with correct calculations, outcomes depend on market conditions, costs and execution mechanics, and the way signals are interpreted. Because of this, WMA values should be treated as computed summaries of past data, not as standalone predictions.