# Market Efficiency & Regime Persistence Measures A Statistics reference entry (Complexity & memory) in the LuxAlgo Library: explained, not implemented as a chart indicator. ## What are Market Efficiency & Regime Persistence Measures? Market efficiency and regime persistence measures are statistics that score how directionally efficient price travel is and how long directional regimes tend to last. The best known is Kaufman's efficiency ratio: the absolute net change over a window divided by the sum of the absolute bar-to-bar changes inside it. It is bounded between 0 and 1: near 1, price traveled in nearly a straight line; near 0, it churned and canceled itself out. Monotonicity counts (the fraction of bars closing in the same direction), run-length statistics, and trend-duration profiles belong to the same family. The family has two roots. The practical one is Perry Kaufman, who published the efficiency ratio in Smarter Trading (1995) as the adaptation engine for his adaptive moving average, making indicator speed a function of market character rather than a fixed setting. The statistical one is older: runs tests compare how often a sequence changes direction against what chance predicts, and Harold Edwin Hurst's mid-century work on Nile flood records, carried into finance by Benoit Mandelbrot and Edgar Peters, framed persistence as long-range dependence. Chart tools such as monotonicity indexes and trend-duration forecasts are lightweight, windowed descendants of those ideas. Mechanically, every member answers the same question: over the last N bars, how much movement was signal and how much was noise? The efficiency ratio compares net displacement with path length, a monotonicity count tallies the share of closes in the dominant direction, and the R-squared of a [linear regression](https://www.luxalgo.com/library/concept/linear-regression/) fit asks how well a straight line explains the window. Positive [autocorrelation](https://www.luxalgo.com/library/concept/autocorrelation/) in returns is the direct statistical signature of persistence, [entropy measures](https://www.luxalgo.com/library/concept/entropy-measures/) score how disordered the up-down sequence is, and [fractal dimension](https://www.luxalgo.com/library/concept/fractal-dimension/) grades the jaggedness of the path itself. High efficiency, positive autocorrelation, low entropy, and low fractal dimension all describe the same condition: movement that carries rather than cancels. Their job is regime classification. Trend systems tend to earn in efficient stretches and bleed in churn, so quantifying travel efficiency turns "is this trending?" into a measurable input rather than a feel. They overlap with the Hurst exponent and the Choppiness Index but are generally simpler, faster to compute, and usable on shorter windows. Because raw readings vary with instrument, timeframe, and window, practitioners usually normalize them with a [z-score](https://www.luxalgo.com/library/concept/z-score/) or [percentile rank](https://www.luxalgo.com/library/concept/percentile-rank/) against the instrument's own history rather than trusting universal thresholds. ## How to read efficiency and persistence measures on a chart These are computed statistics rather than visual patterns, so identification means adding the measure to an indicator pane and calibrating what its readings mean on your instrument. 1. Apply an efficiency ratio or monotonicity indicator with a window of roughly 10 to 30 bars, matched to your holding period; shorter windows flag regime changes sooner but flip more often. 2. Sanity-check extremes against the chart: high readings should coincide with visually clean runs and low readings with overlapping, rangebound bars; if not, the window is mis-sized. 3. Set thresholds from history, not defaults: express the current reading as a percentile of the past few hundred bars and define efficient and choppy as the tails of that distribution. 4. Add a persistence view by marking consecutive same-direction closes or plotting a trend-duration profile to see how long directional regimes have historically lasted on this instrument and timeframe. 5. Track transitions rather than levels: a fast rise from a low base is the classic start-of-regime signature, while decay from a high plateau warns the run is aging. ## How traders use it - Gating entries: trend and breakout logic enabled only above an efficiency threshold, with range tactics such as fades and band reversion taking over below it. - Driving adaptation: the efficiency ratio is the engine of KAMA-style adaptive moving averages, which speed up when travel is clean and flatten in chop. - Setting expectations: the historical distribution of trend durations gives base rates for how long regimes have lasted, informing holding periods and trailing decisions without pretending to forecast the next one; [probability cones](https://www.luxalgo.com/library/concept/probability-cones/) extend the same base-rate thinking to price levels. - Segmenting research: splitting backtest results by efficiency regime, often alongside [distribution-of-returns profiling](https://www.luxalgo.com/library/concept/distribution-of-returns-profiling/), to learn whether an edge is regime-specific before trusting aggregate numbers. - Rotating toolsets: when efficiency stays low, shifting from trend baselines toward oscillators and cycle work such as [dominant cycle measurement](https://www.luxalgo.com/library/concept/dominant-cycle-measurement/), which assumes rhythm rather than drift. ## Efficiency and persistence measures vs related statistics - **Autocorrelation** (https://www.luxalgo.com/library/concept/autocorrelation/): Autocorrelation measures serial dependence in returns directly, lag by lag and signed, while efficiency measures compress the whole windowed path into one bounded number. They usually agree; autocorrelation is the finer diagnostic but noisier on short windows. - **Fractal Dimension** (https://www.luxalgo.com/library/concept/fractal-dimension/): Fractal dimension grades the roughness of the path, from near 1 for a straight line toward 2 for pure churn, making it roughly an inverted efficiency reading arrived at through different mathematics. - **Entropy Measures** (https://www.luxalgo.com/library/concept/entropy-measures/): Entropy scores the disorder of the return sequence without caring about net direction, so a market can read moderately entropic while still drifting. Efficiency measures anchor to net displacement, the more direct input for trend-following logic. ## FAQ ### What is the Kaufman efficiency ratio? It is net directional change divided by total path length: the absolute difference between closes N bars apart, divided by the sum of the absolute close-to-close changes over those N bars. A value of 1 means every bar moved the same direction; values near 0 mean the path wandered and mostly canceled itself. It is the adaptation input Kaufman built KAMA around. ### Can persistence measures predict when a trend will end? No. They describe the current character of price movement and the historical distribution of regime lengths, nothing more. A trend that has already run longer than its typical duration is not obligated to stop, and regimes can flip abruptly at any reading. Use them to choose tactics and size exposure, not to time endings. ### What is a good efficiency ratio value for a trending market? There is no universal number. The ratio's distribution is skewed low, and sustained readings above roughly 0.3 to 0.4 already indicate unusually clean travel on many markets. Calibrate against the instrument's own history rather than importing a fixed threshold. ### Is this the same efficiency as the efficient market hypothesis? No. The efficient market hypothesis is an academic claim about information being reflected in prices. These indicators measure travel efficiency, meaning how straight the recent path was. A market can be informationally efficient and still print a directionally efficient trend; the shared word is mostly coincidence. ### How do these measures differ from the Hurst exponent? The Hurst exponent estimates long-range dependence and needs a large sample to stabilize, which suits offline studies. Windowed efficiency and monotonicity measures are cruder but respond within a handful of bars. Many practitioners run Hurst-style analysis in research and a simple efficiency gate live. ### What window length should I use for an efficiency ratio? Match it to your trading horizon: 10 bars classifies roughly two weeks of daily data, while 30 bars smooths regime flips at the cost of later turns. Kaufman's original KAMA work used a 10-bar ratio. Testing a small grid of windows beats optimizing a single value. ## Related concepts - Hurst Exponent: https://www.luxalgo.com/library/concept/hurst-exponent/ - Fractal Dimension: https://www.luxalgo.com/library/concept/fractal-dimension/ - Entropy Measures: https://www.luxalgo.com/library/concept/entropy-measures/ - Change-point Detection: https://www.luxalgo.com/library/concept/change-point-detection/ --- Source: https://www.luxalgo.com/library/concept/market-efficiency-and-regime-persistence-measures/ (LuxAlgo Library, the encyclopedia of trading & technical analysis). Free to use with attribution: https://www.luxalgo.com/library/license/