Direct Answer
Approximate entropy (ApEn), introduced by Steve Pincus, quantifies the regularity of a time series. Low values mean patterns of a given length tend to be followed by similar patterns, and high values mean the series is less predictable. Values near zero indicate a highly regular series, and larger values indicate more irregularity.
Approximate Entropy: Formula, Meaning, Signals, Examples and How to Use It
How Is Approximate Entropy Calculated?
Approximate entropy (ApEn), introduced by Steve Pincus, quantifies the regularity of a time series. Low values mean patterns of a given length tend to be followed by similar patterns, and high values mean the series is less predictable.
Formula: ApEn(m, r, N) = Phi_m(r) - Phi_(m+1)(r), where Phi_m(r) is the average of ln C_i^m(r) and C_i^m(r) is the share of length-m templates within tolerance r of template i, including itself
Inputs:
- Series values
- Embedding length m (commonly 2)
- Tolerance r (commonly 0.2 times the standard deviation)
- Window length N
Worked example. With Phi_m of -1.10 and Phi_(m+1) of -1.35, the approximate entropy is 0.25. The inputs are illustrative numbers, not market data.
Reading it. Values near zero indicate a highly regular series, and larger values indicate more irregularity. It is only comparable between series computed with the same m, r and length.
Limits. Counting each template as matching itself biases the estimate, which is why sample entropy was proposed as a refinement. It is unreliable on short series.
What is Approximate Entropy?
Approximate entropy (ApEn), introduced by Steve Pincus, quantifies the regularity of a time series. Low values mean patterns of a given length tend to be followed by similar patterns, and high values mean the series is less predictable.
How to use Approximate Entropy
Values near zero indicate a highly regular series, and larger values indicate more irregularity. It is only comparable between series computed with the same m, r and length. A reading alone is never a buy or sell signal; it needs context from price structure, volume, volatility and the broader market regime.
Strengths
- Its output is a single non-negative number per series, so the regularity of different windows or assets can be ranked.
- It was introduced for relatively short, noisy records and uses no model of the process behind the data.
- Reproducible when the formula and data convention are fixed.
Weaknesses and limitations
- Embedding length m and tolerance r must be fixed in advance, and common choices are conventions rather than derived values.
- Series with different standard deviations need r scaled to each, otherwise one series is compared with a relatively tighter tolerance.
- Correlated indicators can create false confidence by repeating the same underlying information.
Editorial and risk note
This page is educational content, not individualized investment advice. It should not imply guaranteed prediction, accuracy or outperformance.
Frequently Asked Questions
Is Approximate Entropy a buy or sell signal?
No. Values near zero indicate a highly regular series, and larger values indicate more irregularity. Counting each template as matching itself biases the estimate, which is why sample entropy was proposed as a refinement. A trade built on it still needs a trigger, an invalidation rule, a position size and realistic execution assumptions.
What is the best setting for Approximate Entropy?
Approximate Entropy takes these inputs: series values; embedding length m (commonly 2); tolerance r (commonly 0.2 times the standard deviation). There is no universal best setting for the tunable ones, so test them across instruments, regimes and out-of-sample periods and favor stable parameter regions over one historical winner.
Can Approximate Entropy be used by itself?
Counting each template as matching itself biases the estimate, which is why sample entropy was proposed as a refinement. It is unreliable on short series. For that reason it is read next to measures of direction, participation, valuation, liquidity and risk, not alone.
Does Approximate Entropy work on every timeframe?
The calculation (ApEn(m, r, N) = Phi_m(r) - Phi_(m+1)(r), where Phi_m(r) is the average of ln C_i^m(r) and C_i^m(r) is the share of length-m templates within tolerance r of template i, including itself) can be run on any bar size, but behavior is not portable: a reading on a five-minute chart describes a very different market window from the same reading on a daily chart. Validate the exact timeframe before relying on it.
Why does Approximate Entropy give false signals?
Embedding length m and tolerance r must be fixed in advance, and common choices are conventions rather than derived values. Series with different standard deviations need r scaled to each, otherwise one series is compared with a relatively tighter tolerance. Lag, noise and regime changes add to this, because market participants react to new information after the reading is calculated.
How should Approximate Entropy be backtested?
Rebuild Approximate Entropy from series values using point-in-time data, write the rules before testing, include delisted securities where relevant, model realistic fills and costs, reserve a validation sample and compare with a simpler baseline.