Markets Are Not Random: The Fractals of Finance
Nine episodes. Sixty-eight markets. Forty-one years of data. The most fundamental assumption in modern finance, tested directly. The answer is no.
Watch the Episode
SERIES OVERVIEW
This Dispatch walks through The Fractals of Finance: a nine-part series published on ATS Trading Solutions. Watch the video above for the guided overview, then follow the links below to read the full series.
Most of the mathematical architecture of modern finance rests on one premise. Not a guess. Not a simplification adopted for convenience. A foundational assumption on which the efficient market hypothesis, the random walk, the capital asset pricing model, Black-Scholes, and Value at Risk are all built.
That premise is this: financial markets are random. Each day’s return is independent of the last. Price movements follow the bell curve. Extreme events are vanishingly rare. Risk is stable and measurable. The past is irrelevant.
The Fractals of Finance tested that premise directly. Not with theory. Not with opinion. With data. Sixty-eight futures markets. Eight asset classes. Six continents. Forty-one years of daily returns. Nine episodes that build a cumulative, empirical case from the ground up.
The premise fails. In every market. Every asset class. Every continent. Every decade.
“Markets are not random. They have never been random. Feedback is the mechanism. Seven independent lines of evidence, across 68 markets, eight asset classes, and forty-one years of data, point to the same answer. That is not a rhetorical claim. That is what the data shows.”
A note on scope before we get into the episodes. The Fractals of Finance measures volatility memory and feedback structure. It is not a claim that tomorrow’s price direction is trivially predictable from today’s. Directional persistence in daily returns is episodic and regime-dependent. What the series documents is the structural organisation of the feedback system itself: the deep memory in volatility magnitude, the universality of the fat-tailed fingerprint, and the causal role of feedback. That is a more robust and more consequential finding than any simple directional signal.
What This Series Covers
The Fractals of Finance is a nine-part research series published on ATS Trading Solutions. It sets out to test the most fundamental assumption in quantitative finance and builds a cumulative empirical case across nine episodes, each one adding one layer, until the full architecture of the case is visible. The argument is not subtle. Neither are the implications.
Episode 0 | Something Is Wrong with the Model
Before the series begins, Episode Zero frames the problem. For over a century, modern finance rested on the random walk hypothesis: price changes are independent, the past tells you nothing about the future, and the bell curve describes the distribution of outcomes. But the evidence kept disagreeing. On a single day in October 2008, the S&P 500 fell by a magnitude the standard model said should not occur once in the lifetime of the universe. In April 2020, crude oil traded at negative thirty-seven dollars per barrel. Under the Gaussian framework, the probability assigned to that event was effectively zero. The model treated it as impossible. Episode Zero asks what the data actually says when you let it speak for itself.
Episode 1 | The Random Walk Is Dead
Part One tests the independence assumption directly. The test is precise: it measures autocorrelation of raw daily returns separately from autocorrelation of absolute returns. Across sixty-eight futures markets, raw return autocorrelation sits near zero on average. The long-run directional average is low, reflecting the cancellation of opposing directional regimes rather than the absence of structure. But absolute return autocorrelation tells a completely different story. The average across all sixty-eight markets is 0.353 at lag one, and it remains positive and statistically significant for over a full trading year. Volatility magnitude is remembered. That memory is universal. Not one market, across any asset class or continent, failed to show it.
Episode 2 | The Nile River’s Secret
Part Two asks how deep the memory goes. The tool is the Rescaled Range method, developed by British hydrologist Harold Edwin Hurst while studying the long-range behaviour of the Nile River across centuries of flood records. Hurst’s method produces a single number between 0 and 1. A Hurst exponent of 0.5 means no memory. Above 0.5 means persistence. The Nile’s exponent was 0.77. Applied to sixty-eight futures markets, the average Hurst exponent is 0.866. Higher than the Nile that inspired the method. The minimum value in the entire dataset, the Swiss Franc, reaches 0.777. Not one market, across any asset class on any continent, behaves like a random walk.
Episode 3 | The Impossible Keeps Happening
Part Three opens the tails. The bell curve says a five-sigma day should occur once every 14,000 years. The dataset contains 2,151 of them. On April 21, 2020, crude oil settled at negative thirty-seven dollars per barrel: effectively a fifty-sigma event under the Gaussian framework. The S&P 500 produced an 11.8-sigma daily move on October 13, 2008. At five sigma, observed extreme events occur 5,791 times more often than the Gaussian distribution predicts. At six and seven sigma, where the Gaussian expected count is effectively zero, over a thousand events were observed at each level. The bell curve does not merely underestimate extremes. It renders them invisible.
Part Four lays all three signatures, memory, persistence, and fat tails, side by side across all sixty-eight markets simultaneously. The question is whether markets that carry more memory also have fatter tails. They do, and the pattern does not respect asset class boundaries. Equities sit beside grains. Bonds overlap with softs. Energy mixes with metals. The dots on the scatter plot occupy a single region regardless of what they trade. When the same phenomenon appears in every market, in every asset class, across every continent, no market-specific explanation survives. The mechanism must be structural.
Part Five moves from observation to controlled experiment. Correlation is not causation, and four episodes of consistent findings do not establish that feedback creates the fingerprint. To isolate the mechanism, the series builds an agent-based model containing two populations present in every real market: divergent agents (trend followers) who condition their behaviour on recent price, creating positive feedback; and convergent agents (value investors and mean reverters) who trade against recent moves, creating negative feedback. The null world contains no divergent agents at all. Every signature vanishes. Thirty percent divergent agents are then introduced. Every signature returns immediately. Within the controlled design, this directly isolates feedback as the mechanism. Remove divergent agents and the fingerprint disappears. Restore them and it returns.
Part Six builds a dial, sweeping the proportion of divergent agents from zero to ninety-five percent in fine increments and measuring every signature at each step. The result is not a gradient. It is a phase transition. Water does not become gradually more solid as temperature falls. Financial markets follow the same principle. All three signatures, fat tails, persistence, and memory, remain near their random baselines for divergent fractions below twenty-five percent. Then, at approximately twenty-five to thirty percent, they erupt simultaneously. Not sequentially. Not gradually. All at once. Real markets sit permanently above this threshold.
Episode 7 | The Permanent Signature
Part Seven tests universality across time. A five-year measurement window is rolled across forty years of data from 1984 to 2026, computing the fingerprint at every quarter. Five natural experiments are examined: Black Monday in 1987, the transition to electronic trading in 2000, the Global Financial Crisis in 2008, the Flash Crash and rise of high-frequency trading in 2010, and the COVID crash in 2020. In forty years of data, the cross-market median Hurst exponent has fluctuated between approximately 0.64 and 0.82. It has never approached 0.5. Not during any crisis. Not across any structural shift. The fingerprint survived every transformation markets have undergone in four decades.
Part Eight assembles the full case. Seven independent lines of evidence, unified into a single framework. Memory: ACF averages 0.353 at lag one, persistent for a full trading year. Persistence: Hurst averages 0.866, higher than the Nile. Fat Tails: 5,791 times more five-sigma events than the Gaussian allows. Universality: all three signatures form a single fingerprint that does not respect asset class, continent, or regulatory boundary. Causation: removing feedback removes every signature; restoring it returns them. Phase Transition: the market reorganises at a critical threshold of approximately twenty-five to thirty percent divergent participation. Permanence: the fingerprint has persisted across forty years, five structural crises, and a complete technological revolution. Each pillar was established independently. Together they form a complete structural explanation for why markets behave the way they do.
Part Nine delivers the verdict. For the specific question this series set out to answer, whether financial markets exhibit the statistical properties the random walk hypothesis predicts, the answer is no. The Gaussian model is systematically wrong on tail probabilities. The independence assumption is wrong on volatility structure. Seven independent lines of evidence across sixty-eight markets, eight asset classes, six continents, and forty-one years of data all point in the same direction. Within the scope this series investigated, the conclusion is not tentative: markets are not random. The mechanism is feedback. The case is closed.
Key Topics Covered
→ Why the autocorrelation of absolute returns reveals persistent volatility memory that raw return autocorrelation does not, and what the difference means for how you understand market structure
→ How the Hurst exponent, a tool built to measure the long memory of the Nile River, reveals that every futures market on earth carries deeper memory than the river that inspired the method
→ Why five-sigma events occur 5,791 times more often than the bell curve predicts, and what that means for every risk model calibrated to Gaussian assumptions
→ How the same statistical fingerprint appears in soybeans, sovereign bonds, crude oil, and currencies, making asset-class-specific explanations structurally impossible
→ Why a controlled agent-based simulation proves feedback is the cause, not merely a correlate: remove divergent agents and the fingerprint vanishes; restore them and it returns
→ The phase transition at approximately twenty-five to thirty percent divergent participation, below which markets are random and above which every signature erupts simultaneously
→ Why the fingerprint has persisted unchanged across forty years, five structural crises, and the complete transition from pit trading to algorithmic execution
Six Things to Take Away
The scope note applies here as it does throughout the series. These takeaways concern volatility memory, tail structure, and feedback mechanics. They are not claims about directional predictability from one day to the next. Directional persistence is episodic and regime-dependent. What is robust and universal is the structural organisation of the feedback system itself.
- Markets carry persistent memory in volatility. The autocorrelation of absolute returns averages 0.353 at lag one across all sixty-eight markets and remains positive and significant for over a full trading year. Volatility magnitude is remembered with exceptional persistence. The near-zero long-run directional average reflects the cancellation of opposing regimes, not the absence of structure.
- The bell curve is wrong. Five-sigma events are not once-in-14,000-year anomalies. They occur every year, in every market, on every continent. At five sigma, observed events are 5,791 times more frequent than the Gaussian model predicts. Any risk framework calibrated to Gaussian assumptions is systematically underestimating the tails.
- The fingerprint is universal. Soybeans, sovereign bonds, crude oil, and currencies all share the same statistical DNA: deep memory, persistent fat tails, and Hurst exponents averaging 0.866. The dots on the scatter plot do not cluster by asset class. One phenomenon, not eight separate anomalies with separate explanations.
- Feedback is the mechanism. In the controlled simulation, removing divergent agents removes every signature. Restoring them returns every signature. The experimental design directly isolates feedback as the cause within the model. This is not a correlation finding dressed up as causation. It is isolation of mechanism under controlled conditions.
- The transition is a phase change. Markets flip from random to structured at a critical density of feedback, approximately twenty-five to thirty percent divergent participation. Below the threshold: noise. Above it: everything we observe in real markets. The transition is not gradual. All three signatures erupt simultaneously at the same critical point.
- The structure is permanent. Forty years. Five crises. Pit trading to algorithmic execution. The cross-market median Hurst exponent has never approached 0.5. The fingerprint has not drifted, not trended, not faded. Build your process on what has always been true, because the evidence says it will keep being true.
Read the Full Series
Every episode is published in full on ATS Trading Solutions. Nine articles, each one building directly on the last. The argument is cumulative. Read them in order.
Ep 0 Something Is Wrong with the Model
The motivating evidence: why the standard mathematical framework of modern finance fails to explain what actually happens in markets.
Testing independence across sixty-eight markets and watching the assumption fail everywhere, in every asset class, without exception.
Hurst exponents, long-range dependence, and the depth of market memory: why the average is 0.866, higher than the Nile itself.
Ep 3 The Impossible Keeps Happening
Fat tails, sigma events, and the catastrophic systematic failure of the bell curve across 647,922 trading days.
Universality across sixty-eight markets: one phenomenon, not eight separate anomalies, and why that distinction changes everything.
Controlled proof that feedback creates the fingerprint: remove divergent agents and it vanishes; restore them and it returns.
Phase transitions and the critical threshold at which the random walk shatters, and why real markets are permanently on the structured side of it.
Forty years, five structural crises, and a fingerprint that has never disappeared, decayed, or been arbitraged away.
Seven pillars unified: the complete structural explanation for why markets behave the way they do.
The case is closed. Markets are not random. The mechanism is feedback. Seven independent lines of evidence, one conclusion.
Read the full series: https://atstradingsolutions.com/the-fractals-of-finance-research-series-the-complete-case/
ABOUT DISPATCHES FROM THE OUTPOST Dispatches from The Outpost is the video series from ATS Trading Solutions where Rich Brennan walks through our published research, deep dives on specific topics, and challenges the conventional wisdom that holds most traders back. Each Dispatch is accompanied by a full written summary, key takeaways, and links to the original research. Watch the video, read the series, go as deep as you want. → Subscribe on YouTube │ → Browse all Dispatches │ → atstradingsolutions.com |
Want the theoretical foundation for why trend following works?
The Fractals of Finance: Determinism, Adaptation and the Geometry of Markets
The book explores the full architecture of feedback, fat tails, and fractal structure in financial markets, and what it means for how we trade, invest, and understand risk.
Available now on Amazon in paperback, hardcover, and Kindle.
Want a practical field manual for trading trends and capturing outliers?
The Aussie Turtles Trend Following Guide: A Field Manual for Hunting Outliers adapts the timeless principles of the original Turtle traders into a systematic, rules-based approach for modern markets. Co-authored with Adam Havryliv.
Available now on Amazon in paperback, hardcover, and Kindle.