What is CMO (and why the data matters)?
CMO (Chande Momentum Oscillator) is an oscillator that compares summed “up” and “down” price movements over a chosen lookback period. The key idea is simple: if recent movements are mostly upward, the oscillator tends to be higher; if they are mostly downward, it tends to be lower. Because the oscillator is built from specific price changes, the data needed to assess CMO must include both the underlying time series and the exact rules used to turn that series into up/down moves.
What data is needed to assess CMO?
To assess CMO accurately, collect the following inputs and documentation.
1) The underlying time series
- Price series used for the calculation (commonly the closing price). Decide whether you use close, typical price, or another consistent definition.
- Exact instrument and market context (same symbol, same venue, and consistent session handling). Different sources can produce different price histories even for the “same” instrument.
2) The lookback window (the “period”)
- The window length used to aggregate up and down movements (for example, a fixed number of bars). CMO values depend directly on this period.
- Whether the period is expressed in bars of the same timeframe (e.g., 1-hour bars) or calendar time.
3) Timestamping and resampling rules
- The timeframe of the input data (e.g., 5-minute, 1-hour, daily).
- How you handle missing bars: interpolation, skipping, or forward-filling can materially change the computed up/down moves.
- Timezone alignment and session boundaries. For markets with discontinuities, bar construction rules affect the first “available” movement in a window.
4) Up/down movement definition (calculation mechanics)
CMO requires separating movements into upward and downward components. You need the exact rule your method uses, including:
- How price change is computed (e.g., difference between consecutive closes).
- How to treat zero change (count as neither up nor down, or assign to one side). Different implementations may treat edge cases differently.
- Whether absolute differences are used in the summed terms.
5) Provenance and reproducibility information
- Data provider identity and dataset version (a “live” feed can change historically after corporate actions or backfills).
- Any preprocessing steps: cleaning outliers, adjusting for splits/dividends (if relevant), or using adjusted prices.
- The software or formula specification that defines the CMO computation, so someone else can replicate the same numbers from the same inputs.
How does the needed data work in practice?
CMO is assessed by computing it from a rolling window of past price changes. Concretely, a reproducible assessment requires:
- Choosing the price definition (e.g., close-to-close differences).
- Selecting the lookback window length.
- Converting each consecutive price change into an “up” component or a “down” component using a clearly stated rule.
- Summing those components over the window and applying the oscillator’s normalization.
If any of the above differs—such as using a different timeframe, a different price type, a different period, or a different zero-change rule—the resulting CMO series will not match. That is why data provenance and calculation mechanics are part of “what data is needed,” not just the raw prices.
Evidence, examples, and how to compare CMO fairly
A practical check is to compare two CMO calculations only when they share the same inputs and mechanics:
- Same instrument and same timeframe.
- Same period length.
- Same price field (e.g., close).
- Same rule for zero changes and missing data.
If you cannot guarantee those conditions, comparisons can be misleading even when both series are labeled “CMO.” For example, a method that treats missing bars differently can alter the up/down sums in the first few windows after gaps.
Limitations and risks (what can go wrong)
Material limitation: regime dependence
Oscillators like CMO describe momentum-like behavior over a recent window. Market regimes can change, so historical relationships between CMO levels and subsequent outcomes may not hold later.
Failure mode: inconsistent implementation
Small differences in mechanics—such as how negative movements are summed, or how zero-change bars are categorized—can change the magnitude and direction of the oscillator.
Data integrity risks
- Missing or misaligned timestamps can distort consecutive price differences.
- Backfilled or revised datasets can change historical values.