Definition and purpose
Fractal Dimension Index (FDI) is a label used for a fractal-based statistic that aims to quantify the complexity of a time series by looking at how its apparent detail changes when you view it at different scales. In practical forex research, the “input” is typically price-like data (for example, a sequence of observed values over time), and the “output” is a single number intended to summarize the series’ roughness or scale-dependent behavior.
FDI is not the same thing as a trading rule. It does not inherently tell you what the next candle will do. Instead, it can be used to characterize whether a series appears more or less complex across scales—useful for descriptions, comparisons, or for building other models that need a complexity feature.
How Fractal Dimension Index works (simple model)
A simple way to think about the idea is: two sequences can both “move,” but one can look smooth at many zoom levels while the other looks jagged or irregular. Fractal approaches estimate a dimension that reflects that multi-scale irregularity.
A common workflow in indicator-style usage is:
- Choose the time series you want to measure (e.g., a sequence of historical prices or returns).
- Select a calculation approach and parameters (for example, how to split the series into scales/windows and what transformation to apply).
- Estimate the fractal-related statistic that becomes the FDI value.
Because the term “Fractal Dimension Index” can be used across different implementations, the exact computation can vary. The key assumption is that the data contain scale-dependent structure worth estimating, and that the estimation method is appropriate for the sample length and noise level.
Assumptions for any calculation
When people compute FDI, they implicitly assume:
- The historical window is long enough to estimate scale behavior.
- The data frequency and preprocessing (such as whether you use prices versus returns) match the question you’re asking.
- The observed roughness is not dominated entirely by measurement noise or microstructure effects.
Evidence and example (what to verify)
Since outcomes depend on implementation choices, the most independent “evidence” you can check is internal consistency:
- Recompute the index using the same method on the same type of series.
- Change one assumption at a time (like window length or data transformation) and observe how stable the results are.
A simple illustrative example (not a live quote or forecast):
- Suppose you compute FDI on a historical segment where the price series visually appears smooth and another segment where it appears jagged.
- If the method is capturing scale-dependent roughness, you would expect the FDI for the jagged segment to differ from the smooth segment.
However, “expect” here means “might,” not “will.” Estimation can produce counterintuitive values if the sample is short, if the series contains outliers, or if the chosen scales do not meaningfully represent the underlying behavior.
Limitations and failure modes
Several limitations commonly affect fractal-dimension-style indicators:
- Estimation sensitivity: Different algorithms, parameter settings, and preprocessing steps can yield different numeric results from the same underlying data.
- Sample length: With too little history, the multi-scale relationship is poorly estimated, which can make the FDI unstable.
- Noise and microstructure: Very short-term price behavior can reflect trading frictions, spreads, and execution effects rather than the broader complexity the method intends to measure.
- Non-stationarity: If the statistical properties of the series change over time, a single FDI value computed on one window may not represent later windows.
Also, even if you observe historical association between “higher FDI” and a past outcome, that does not establish future predictive power. Relationships can break due to market regime shifts, costs, and changes in how the data are produced.
Verification and next question
To verify FDI claims in any forex context, focus on what can be checked:
- What exact definition and computation method is used (the implementation matters).
- What input data are used (prices vs returns; timeframe; preprocessing).
- What window length and scale selection are assumed.
- How stable the index is under small changes to those assumptions.
If you’re evaluating a specific chart or report, a useful next step is to ask: “Which method defines this Fractal Dimension Index, and what assumptions does it use for scaling and preprocessing?”