The Vault

THE FRACTALS OF FINANCE | Research Series Phase 2 | Episode 3 of 9

The Trend-Equity Paradox

Everyone in the industry knows that trend-following provides crisis alpha. Nobody has explained why. The coupled spectrum does.

We identified the twenty worst months for the S&P 500 across four decades. These are the months that destroy portfolios, force margin calls, trigger redemptions, and end careers.

In sixteen of those twenty months, a diversified trend-following strategy delivered a positive return. The average equity loss was negative 7.4 percent. The average trend gain was positive 1.1 percent. The correlation between trend and equity across those twenty months was negative 0.82.

This property is well known in the industry. The managed futures community has documented it for years under the label of crisis alpha. The finding itself is not new.

What is new is the explanation. The permanently coupled spectrum we revealed in Episodes 1 and 2 is the mechanism that creates crisis alpha. And without the coupling, in a truly random and independent market, the crisis alpha would not exist.

The Correlation That Isn’t Constant

The strategy used here is intentionally simple: go long each of the sixty-eight contracts when price is above its 200-day moving average, short when below, and scale each position by the inverse of its recent volatility targeting ten percent annualised risk. This is not how a dedicated outlier hunter constructs a portfolio, and it makes no claim to be. It is chosen because it is transparent, reproducible, and free of optimisation. The goal is to isolate the structural finding, not to showcase a strategy. If the coupling mechanism produces crisis alpha in a deliberately plain construction, it will produce it more powerfully in one specifically designed to harvest it.

Over the full sample, the strategy delivered a MAR ratio of 0.41: annualised return of 2.9 percent against a maximum drawdown of 6.9 percent. The S&P 500 delivered a MAR of 0.09 over the same period: higher return at 4.5 percent, but with a maximum drawdown of 49.4 percent that overwhelms the return advantage. Per unit of worst-case pain, trend delivers more than four times what equities deliver.

The full-sample correlation between them is negative 0.21. Meaningfully negative, but nowhere near the negative 0.82 we opened with, or the negative 0.68 that appears in bear markets. It tells you that trend tilts toward diversification on average. It does not tell you when or how much.

Figure 3.1 Rolling one-year correlation between diversified trend and S&P 500. The correlation oscillates between positive 0.67 and negative 0.66. Bear market periods coincide systematically with the deepest negative readings.

The rolling window reveals a trace that oscillates between positive 0.67 and negative 0.66, spending roughly half its time on each side. But the relationship is not random. Look at the pink background shading marking bear market periods. Every bear market coincides with a plunge in the correlation. Every recovery sees it return toward zero or positive.

The full sample mixes two opposing states: mild positive correlation during good times, deep negative correlation during bad. The average is negative 0.21. The average hides everything.

The Staircase

To make the regime dependence precise, we sorted every month in the forty-year sample into one of four buckets based on the equity drawdown at that point in time. The result is monotonic.

Figure 3.2 Trend-equity correlation conditioned on equity drawdown regime. The correlation flips sign as equities fall and deepens with severity. A perfect staircase.

In calm periods, when equities are within two percent of their peak, the correlation is positive 0.12. In mild drawdowns, positive 0.09. Trend and equity move gently together when the world is fine. Then equities start falling. In drawdowns of five to fifteen percent, the correlation flips to negative 0.40. In crises, when equities are down more than fifteen percent from peak, negative 0.44.

The pattern is a staircase. Each step down in equity performance corresponds to a step down in correlation. The worse equities perform, the harder trend works against them. This is not a statistical artefact. It is a structural property that persists across the entire sample.

The Twenty Months That Matter

The staircase becomes a cliff at the extremes.

Figure 3.3 Monthly returns scatter. The twenty worst equity months cluster in the upper-left quadrant: equities falling, trend rising. The worse the equity loss, the stronger the trend gain.

The twenty worst equity months lost an average of 7.4 percent each. In sixteen of the twenty, trend was positive. The green dots cluster in the upper-left quadrant: the crisis alpha zone. But look more closely at the gradient within the chart. The months where trend failed are clustered among the least severe of the twenty. The deepest equity months, losses of twelve to fifteen percent, all produced the strongest positive trend returns.

The worse the equity loss, the stronger the trend gain. This is exactly what the coupling predicts: deeper, more sustained moves mean the system has shifted further toward the trending end of the spectrum, producing stronger positive feedback, which trend-following harvests by design.

The Coupling Creates the Crisis Alpha

Every major managed futures firm has documented the crisis alpha property. What none of them has explained is why it exists.

The standard explanation is mechanical: when equities fall, trend strategies go short equities and long safe-haven assets, producing returns that oppose the equity drawdown. This is true but incomplete. It describes what trend does. It does not explain why the opportunity exists in the first place.

The answer comes from the coupled spectrum.

Episodes 1 and 2 proved that financial markets are not random walks. They operate along a spectrum from trending to oscillating behaviour, driven by the balance between positive and negative feedback. They proved that markets are permanently coupled along this spectrum.

Bear markets are not random downward drifts. They are events where the coupled system shifts toward the trending end of the spectrum. Positive feedback dominates across the universe. Equities fall persistently. Bonds rally persistently. Commodities and currencies reprice directionally. The coupling means that this directional energy floods through dozens of markets simultaneously, not just equities.

Trend-following, a divergent strategy by design, harvests directional persistence. When the coupled system shifts to the trending state, trend captures that persistence across the full universe. The deeper the system moves toward the trending end, the more directional energy is available across more markets, and the more negative the trend-equity correlation becomes.

Figure 3.4 Trend-equity correlation split by market regime. Bull: positive 0.13. Bear: negative 0.68. The full-sample number is just the weighted average.

The split is stark. During bull markets, the correlation is positive 0.13. During bear markets, negative 0.68. The full-sample number of negative 0.21 is just the weighted average of these two states.

This is why the coupling matters. In a truly random market, where returns are independent, there would be no directional persistence. Prices would not trend. Trend-following would capture nothing. The correlation between trend and equity would be zero in all regimes, not just on average. There would be no crisis alpha.

The coupled spectrum is what makes bear markets directional across many markets simultaneously. Trend-following is what harvests that directionality. The crisis alpha is not magic. It is the mechanical consequence of a non-random, permanently coupled market.

The Mechanism in Action

To see the mechanism at full resolution, we zoomed into the Global Financial Crisis.

Figure 3.5 Three-panel GFC case study. The sequence is visible: equity falls persistently, the coupled system shifts to the trending state, trend repositions and harvests, correlation plunges negative.

The three panels tell the story sequentially. In 2006 and early 2007, equities rise gently, trend earns a modest return, and their correlation hovers around positive 0.2. The system is in the noise zone. The coupling is present but the spectral position is balanced.

Then equities begin to fall. The S&P 500 drops persistently through 2008. The coupled system shifts toward the trending end of the spectrum. Positive feedback dominates. Prices do not bounce; they continue to fall. Trend, responding to the directional signal across dozens of markets, repositions short equities and long the safe-haven markets that are trending in the opposite direction. The middle panel shows the strategy accelerating upward as equity collapses downward. The bottom panel shows the correlation plunging deeply negative.

After Lehman, the sequence intensifies. Equities crater. Trend delivers its largest gains. The correlation hits its deepest negative reading. The coupled system is fully in the trending state, and the divergent strategy is fully aligned with it.

Then the recovery begins. Equities start rising again. The coupled system gradually shifts back through the noise zone. Trend repositions long. The correlation returns toward zero and eventually positive. The spectral position resets.

Beyond Crisis: The Bull Market Story

If the crisis alpha is produced by the coupling mechanism, the same mechanism should operate during bull markets. The coupling is permanent. The spectral position shifts. When the system is in the trending state during a bull market, trend-following should still benefit.

The data confirms this.

During bull markets with coupled trending, the strategy earns 5.1 percent annualised with a correlation to equities of positive 0.04. That correlation is effectively zero. The returns are genuinely additive, not merely a diluted version of the equity return. Both sides of the strategy contribute: the long side earns 4.5 percent and the short side earns 1.2 percent.

During bull markets in the noise zone, the strategy earns 3.5 percent with a correlation of positive 0.24. Still positive, but the equity correlation has crept in. The long side dominates.

During bear markets with coupled trending, the strategy earns 8.7 percent with a correlation of negative 0.58. This is classic crisis alpha, and it is the strongest single-cell reading in the matrix.

The only hostile environment is coupled mean-reversion during flat equity markets, where the strategy loses 5.2 percent annualised. This is the state where every directional signal fails simultaneously. The coupled system is oscillating, negative feedback dominates, and a divergent strategy is maximally misaligned.

The reframed picture: crisis alpha is not a special property of crises. It is one expression of permanent coupling. The same mechanism produces additive returns during bull markets (when the system is trending), protective returns during bear markets (when the system is trending in the opposite direction), and losses during coupled oscillation (when the system is mean-reverting). The coupling is always present. The outcome depends on where the coupled system sits on the spectrum, and which direction equities happen to be moving.

What This Means for Buy and Hold

The matrix in the previous section tells a story that goes beyond trend-following. Read it from the perspective of a buy and hold equity investor and a different picture emerges.

Buy and hold is structurally long only. It benefits during one spectral state: coupled trending with equities rising. In every other state it is either stagnant or exposed. During coupled mean reversion it drifts, going nowhere while absorbing volatility. During coupled trending with equities falling it experiences the full force of correlated losses across the portfolio, with no offsetting mechanism and no structural response.

This is not bad luck. It is architecture. The permanent coupling ensures that when the crisis arrives it arrives everywhere simultaneously. The diversification that appeared to work during the oscillatory calm was never switched off. It was simply running in a phase that produced offsetting returns. When the spectral position shifts to coordinated trend, that same coupling produces co-directional losses. The investor who held a diversified equity portfolio believing they were protected discovers that the coupling was always present. Only its expression changed.

The duration of trending states makes this more serious than the standard crisis alpha narrative suggests. The GFC sustained directional moves for over a year. The 2022 inflation shock ran across bonds, equities and currencies simultaneously for months. A buy and hold investor has no mechanism to benefit from or protect against that duration. The longer the trending state persists, the deeper the exposure compounds without relief.

Trend-following has a structural response to every spectral state. It earns during trending states regardless of direction, struggles modestly during mean reversion, and is maximally aligned with the system during the worst equity crises. Critically, the protection scales with the severity and duration of the trending state. The worse and longer the crisis, the more the asymmetric protection compounds. This is not a coincidence. It is the mechanical consequence of a divergent strategy operating in a permanently coupled system that is releasing directional energy across dozens of markets simultaneously.

The question for any investor building a portfolio for the long run is not whether crises will arrive. The coupled spectrum guarantees they will, and that they will be felt everywhere at once. The question is how long they last, and whether the portfolio has a structural mechanism to respond. Buy and hold does not. A trend allocation does.

The Paradox Resolved

The trend-equity paradox is this: the full-sample correlation is modestly negative, and yet trend is one of the most powerful diversifiers available to an equity investor. The paradox dissolves once you understand the coupling.

The correlation is not a constant. It is a variable that depends on the spectral position of the permanently coupled system. In bull markets, trend and equity are weakly aligned: positive 0.13. In bear markets, they are strongly opposed: negative 0.68. In the worst months, the relationship intensifies further. The full-sample statistic averages across these states and destroys the information.

The industry has known about crisis alpha for years. What this episode adds is the causal mechanism. The permanently coupled spectrum creates the directional persistence that bear markets exhibit. Trend-following harvests that directional persistence from the opposite side, across dozens of markets simultaneously. The crisis alpha is not an empirical curiosity. It is the mechanical consequence of a non-random, permanently coupled market.

Without the coupling, there is no crisis alpha. The coupling is not decoration. It is the source. The Fractals of Finance develops the full architecture of this mechanism and its implications for portfolio construction.

Episodes 1 through 3 form Act I of this series. We have revealed the coupled spectrum, mapped it across sixty-eight markets, and explained the most commercially important consequence for portfolio construction. Act II begins with Episode 4, which turns to the formal question: is the structure random? The answer, drawn from variance ratio tests across the full universe, will eliminate the first of our three states entirely.

The coupling is real. The crisis alpha is its consequence. Episode 4 asks whether any of this could have occurred by chance.

Endnotes

Methodology

The twenty worst S&P 500 months by total monthly log return over the 1986–2026 sample. Mean equity return: -7.43%. Mean trend return: +1.10%. Trend positive in 16 of 20 months. Pearson correlation within these 20 observations: -0.82. This extreme negative correlation in tail events is consistent with the structural mechanism described in this episode.
Trend strategy construction: for each of the sixty-eight contracts, a daily signal is generated as the sign of (Close minus 200-day simple moving average). Position sizing uses inverse volatility scaling: target annualised volatility of 10% divided by trailing 63-day realised volatility, capped at 3.0× leverage. Strategy return on day t = signal(t−1) × sizing(t−1) × log_return(t). The diversified trend index is the equal-weighted average across all contracts with available data. No transaction costs, slippage, or financing costs are modelled.
Performance over common sample 2 January 1986 to 30 January 2026 (10,104 trading days). Trend: annualised return 2.9%, maximum drawdown 6.9%, MAR 0.41. S&P 500: annualised return 4.5%, maximum drawdown 49.4%, MAR 0.09. The MAR ratio (return / max drawdown) measures return per unit of worst-case pain and is the primary performance metric in this series. The MAR advantage of approximately 4.5× reflects the dramatically different drawdown profiles: trend’s worst loss is 6.9% versus the S&P’s 49.4%.
Conditional correlation computed on monthly aggregated returns. Months classified by equity drawdown from peak at month-end: calm (drawdown < 2%, n=219, corr=+0.12), mild drawdown (2–5%, n=61, corr=+0.09), drawdown (5–15%, n=80, corr=−0.40), crisis (>15%, n=121, corr=−0.44).
Bull/bear regime correlation: bull markets defined as periods where the trailing 252-day S&P 500 cumulative log return is positive; bear markets where it is zero or negative. Bull: correlation +0.13 (n=367 months). Bear: correlation −0.68 (n=103 months). Full sample: −0.21.
Coupling state × equity regime analysis uses the spectral classification from Episode 2 (coupled trending: frac_trend > 0.58; coupled MR: frac_mr > 0.58) crossed with equity direction (bull: trailing 12m equity return > 5%; bear: < −5%; flat: between). Simple 12-month momentum trend proxy across all available contracts. Bull + coupled trend: +5.1% annualised, equity correlation +0.04. Bull + noise: +3.5%, correlation +0.24. Bull + coupled MR: +5.0%, correlation +0.13. Bear + coupled trend: +8.7%, correlation −0.58. Flat + coupled MR: −5.2%, correlation −0.16.

Data

Same dataset as Episodes 1 and 2: sixty-eight CSI ratio-adjusted continuous futures, September 1984 to January 2026. Equity benchmark: S&P 500 E-mini (ES__B). All returns are daily log returns. Common sample begins 2 January 1986.

Figures

Figure 3.1: Rolling 252-day Pearson correlation, diversified trend vs S&P 500. Blue/orange fill. Bear market shading in pink.
Figure 3.2: Conditional correlation by drawdown regime.
Figure 3.3: Monthly return scatter with worst-20 highlighted.
Figure 3.4: Bull/bear regime correlation split.
Figure 3.5: Three-panel GFC case study.

This research series is drawn from 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.