In a fat-tailed world, the majority of wealth creation comes from a tiny number of extraordinary months. Trend following is the only process systematically designed to stay on board when they arrive
Start with $100,000 in the S&P 500. After 26 years it becomes $747,697. Remove the best 32 months and it becomes $66,939. The same exercise on the TF Index takes $667,058 down to $62,333. In both cases, the best 10% of months account for approximately 91% of terminal wealth. The right tail is not a bonus. It is the return.
Episode 6 showed that the cut protects the compounding engine by truncating the left tail. It demonstrated that doubling the losses while leaving every winner intact destroys 98% of terminal wealth, that the depth of losses rather than their frequency is what determines geometric outcomes, and that the trend following process curates losses with convex efficiency. The cut is the shield.
But a shield alone does not create wealth. Something must drive the compounding forward. Something must produce the positive returns that accumulate across decades. If the cut controls the left tail, the question is: what creates the right tail?
The answer is the second operation of the trend following process: letting winners run. And the data reveals something that even experienced investors find startling. In both equities and trend following, terminal wealth is dominated by a small number of extraordinary months. The difference is not whether outliers matter. It is whether the process is designed to capture them.
The Dominance of Outliers
The belief that investment returns accumulate gradually, each month contributing a small increment to the growing total, is one of the most persistent illusions in finance. It is also one of the most dangerous. In a multiplicative world where returns compound rather than add, the distribution of monthly contributions to terminal wealth is profoundly unequal. A handful of months do most of the work. The rest are noise.
Hendrik Bessembinder documented this phenomenon in equity markets with a finding that reshaped academic finance: approximately 4% of stocks accounted for all net wealth creation in the US equity market over its history. The remaining 96% collectively matched Treasury bills. The entire equity premium, the reason anyone owns stocks at all, was concentrated in a tiny fraction of the universe. The same pattern appears when the lens shifts from individual stocks to time periods.
The numbers are stark. For the S&P 500, removing the best 32 months out of 313 reduces terminal wealth from $747,697 to $66,939, a 91% destruction. For the TF Index, $667,058 collapses to $62,333, also a 91% destruction. For Berkshire Hathaway, $1,410,933 collapses to $59,622, a 96% destruction. In every case, the vast majority of terminal wealth is created by a small minority of months. This is not a feature unique to trend following. It is a feature of compounding itself in a fat-tailed world. Returns are not distributed evenly across time. They cluster. They concentrate. They arrive in bursts that are disproportionate to anything the remaining months produce.
The conventional framing of the missing best days argument, which warns investors not to miss the market’s best days by being out of equities, is correct in its observation but wrong in its conclusion. The observation is that a small number of extreme positive periods drive the majority of long-term returns. The conclusion drawn is that investors should therefore remain fully invested at all times, accepting whatever drawdowns occur in order to be present for the outlier months. But this conclusion ignores the other side of the distribution. The same fat tails that produce extraordinary positive months also produce extraordinary negative months. Being present for the best months while also being present for the worst is not a strategy. It is a statement of helplessness.
The question is not whether outlier months drive returns. They do. The question is whether a process can be designed that captures the positive outliers while avoiding the negative ones. The answer is the asymmetric profile that trend following produces.
The Shape That Compounds
The raw dependence on outlier months is similar across all three benchmarks. But the shape of those outlier months is fundamentally different. This is where the geometric argument separates trend following from conventional equity exposure.
For the S&P 500, the best month was +12.82% and the worst was −16.79%. The best month was smaller than the worst month in absolute terms, producing a best-to-worst ratio of 0.76. The right tail is shorter than the left tail. The distribution is negatively skewed.
For the TF Index, the best month was +11.95% and the worst was −8.55%. The best month was nearly 40% larger than the worst month in absolute terms, producing a best-to-worst ratio of 1.40. The right tail extends further than the left tail. The distribution is positively skewed.
This ratio, seemingly a minor statistical detail, is the difference between a process that compounds wealth and a process that compounds risk. When the right tail is shorter than the left, the extreme events that drive terminal wealth are smaller than the extreme events that destroy it. The outliers work against the compounding engine. When the right tail is longer than the left, the extreme events that drive terminal wealth are larger than the extreme events that destroy it. The outliers work for the compounding engine.
Chart 18: Monthly return distributions. The S&P 500 (left) has negative skew: the left tail extends further than the right. The TF Index (right) has positive skew: the right tail extends further. In a multiplicative world, this asymmetry determines the geometry of compounding.
The tail structure extends beyond the single best and worst months. At the 5th and 95th percentiles, the S&P 500’s tails are roughly symmetric: −7.83% and +7.61%, a ratio of 0.97. But the TF Index shows clear asymmetry at the same thresholds: −5.17% and +7.27%, a ratio of 1.41. The right tail consistently extends further than the left. This is not a feature of one exceptional month. It is a structural property of the entire distribution.
The win-to-loss size ratio tells the same story from a different angle. The S&P 500’s average winning month is +3.22% and its average losing month is −3.90%, producing a ratio of 0.83. Average wins are smaller than average losses. The TF Index’s average winning month is +3.36% and its average losing month is −2.64%, producing a ratio of 1.28. Average wins are 28% larger than average losses. The process pays more when it wins than it costs when it loses.
The S&P 500 loses more when it loses than it gains when it gains. Trend following gains more when it gains than it loses when it loses. In a multiplicative world, this asymmetry is the architecture of compounding.
Where the Right Tail Lives
The TF Index’s ten best months reveal the mechanism that creates positive skew. These are the months that, collectively, carry a disproportionate share of terminal wealth.
Of the TF Index’s ten best months, seven occurred when the S&P 500 was negative. The process’s largest positive outliers are disproportionately concentrated in precisely the months when equity markets are falling. This is not coincidence. It is the mechanical output of a process that responds to sustained directional price movement regardless of direction.
When equity markets crash, the crash itself is a trend. It is a sustained directional move driven by the same feedback mechanisms described in Episode 5: herding, forced liquidation, margin calls, algorithmic amplification, liquidity withdrawal. The trend following process, having cut its long equity positions as the deterioration began, may have established short positions that profit from the continuation of the decline. The cut removed the process from the left tail of the equity distribution. The trend ride placed it in the right tail of the trend distribution. The same month that destroys equity portfolios creates the trend follower’s largest gains.
This is why the TF Index’s right tail is longer than its left tail. The process is structurally designed to produce its largest gains during the most extreme market environments. It does not merely survive crises. It harvests them. The right tail of the trend following distribution is built, in large part, from the left tail of the equity distribution.
The Mechanism: Two Operations, One Geometry
The positive skew of the trend following distribution is not an accident, not luck, and not the product of individual manager brilliance. It is the mechanical output of two operations acting together.
The first operation truncates the left tail. This is the cut, examined in Episode 6. When a position moves against the process by a predefined amount, the position is closed. The loss is taken. The damage is bounded. No single trade, and no accumulation of losses in a crisis period, can push the portfolio into the catastrophic region of the recovery curve. The left tail is capped. The worst month of the TF Index was −8.55%. The worst month of the S&P 500 was −16.79%. The cut is the mechanism that prevents the left tail from extending.
The second operation extends the right tail. This is the trend ride. When a position moves in the direction of the trend, the process holds. It does not take profits at a predefined target. It does not exit because the gain feels large enough. It trails a stop behind the advancing price and lets the market determine how far the move extends. If the trend continues for weeks, the position is held for weeks. If it continues for months, the position is held for months. The right tail is not capped. It is open-ended.
The combination of these two operations produces the distribution that geometric compounding rewards. The left tail is bounded. The right tail is unbounded. Losses are capped by the process. Gains are determined by the market. The process cannot lose more than its exit rule allows on any single trade. But it can gain as much as the trend delivers. This structural asymmetry, this designed mismatch between the bounded left and the unbounded right, is the source of the positive skew.
A conventional equity investor has the opposite profile. There is no exit rule that caps losses. A stock held through a 50% drawdown inflicts the full 50%. But there is no mechanism that specifically extends gains either, because the investor holds the same position regardless of how far price moves. The distribution is symmetric or, more commonly in practice, negatively skewed because crashes tend to be sharper and more violent than rallies. The equity investor’s best month is smaller than the worst month. The trend follower’s best month is larger than the worst month. The difference is not selection. It is design.
The cut bounds the left tail. The trend ride unbounds the right. Together they produce the positively skewed distribution that compounding rewards. This is not a description of what happened. It is a description of what the process was designed to produce.
The Population Evidence: Skewness Across 41 Managers
If positive skew were a property of a few exceptional managers, it would be an anecdote. It is not an anecdote. It is a population-level characteristic of the process.
Of the 41 trend following managers in the NilssonHedge database with track records exceeding 20 years, 37 have positive skew. That is 90%. The S&P 500 has a skewness of −0.51. The TF Index has a skewness of +0.23. Across 41 independent implementations, spanning different firms, different markets, different timeframes, different position-sizing rules, and different risk management approaches, the same statistical fingerprint appears: the right tail extends further than the left.
Chart 6: Return skewness for 41 trend followers with 20+ year track records. 37 of 41 (90%) exhibit positive skew. The S&P 500’s negative skew (−0.51) is marked for reference. Positive skew is a structural output of the process, not a feature of individual brilliance.
The four managers with negative skew are not evidence against the thesis. They represent the natural variation in a population of 41 implementations. Some managers trade shorter timeframes, producing more whipsaw and less trend capture. Some use tighter stops that reduce the right tail along with the left. But 90% is a supermajority. It is a population-level finding that does not depend on the selection of any particular manager.
This is the difference between a process and a person. If one manager produces positive skew, the explanation might be skill, luck, or a specific market environment. When 37 of 41 managers produce positive skew, the explanation is structural. The process of cutting losses and letting winners run, implemented independently by dozens of firms across decades, reliably produces a positively skewed return distribution. It is a property of the mechanism, not the operator.
Why Positive Skew Compounds
The connection between positive skew and geometric wealth is direct. Episode 2 established that geometric returns are approximately equal to arithmetic returns minus one-half the variance. Variance penalises both upside and downside dispersion equally. But the geometric cost of downside dispersion is convex, meaning each additional unit of loss costs exponentially more, while the geometric benefit of upside dispersion is concave, meaning each additional unit of gain contributes proportionally less.
A positively skewed distribution concentrates its dispersion in the right tail, where the geometric penalty is lower, and truncates its dispersion in the left tail, where the geometric penalty is higher. Two distributions with identical arithmetic means and identical variances will produce different terminal wealth if one is positively skewed and the other is negatively skewed. The positively skewed distribution will compound more wealth because its variance is concentrated where it does less geometric damage.
This is why the TF Index, with a lower CAGR than the S&P 500, produces a higher MAR ratio. The shape of its returns is more geometrically efficient. Each unit of return is converted into terminal wealth with less geometric drag. The S&P 500 earns a higher arithmetic return but pays a higher geometric tax because its negative skew concentrates dispersion in the left tail, precisely where the convex recovery curve inflicts the most damage.
The trend following process does not merely produce positive returns. It produces returns with the correct shape for compounding. The cut ensures that the left tail does not extend into the convex destruction zone. The trend ride ensures that the right tail extends as far as the market allows. The combination produces a distribution that is structurally aligned with the mathematics of geometric growth.
The Patience Cost
If the right tail is where terminal wealth lives, and if the process is designed to extend the right tail, then the critical question for any investor is: what does it cost to wait for the right tail to arrive?
The answer is the experience of the other roughly 90% of months. The months that are not in the best decile. The months that produce small gains, small losses, flat returns, whipsaws, and the slow grind of a process that appears to be doing nothing. The trend follower’s experience, month by month, is dominated by mediocrity. Losing months outnumber winning months in absolute frequency (45% losing versus 55% winning). Many winning months are modest. The process can spend months or even years without producing an outlier. During these periods, the equity markets may be rallying steadily, the S&P 500 compounding month after month, and the trend following process looking like an expensive mistake.
This is the patience cost. It is real, it is painful, and it is the price of admission to the right tail. The outlier months that carry 91% of terminal wealth do not announce their arrival. They cannot be predicted. They cannot be timed. They can only be captured by a process that is running continuously, absorbing the cost of the ordinary months in order to be positioned when the extraordinary months arrive.
The investor who abandons the process during a quiet period, who switches to a strategy with a higher batting average or a smoother equity curve, is making a bet that the right tail will not arrive during their absence. This is, statistically, a terrible bet. The best months are concentrated, infrequent, and unpredictable. Missing even a few of them is catastrophic. Removing the best 10 months from the TF Index’s 313-month history destroys 60% of terminal wealth. The cost of absence is not linear. It is geometric.
The right tail rewards patience and punishes timing. Removing 10 months from 313, just 3% of the total, destroys 60% of terminal wealth. The process must be running when the outlier arrives. There is no second chance.
The Disposition Effect Revisited
Episode 6 showed that the disposition effect causes investors to hold losers too long, extending the left tail. The mirror image is equally damaging: the disposition effect causes investors to sell winners too early, truncating the right tail.
The urge to take profits is powerful. A position shows a 5% gain, and the investor wants to lock it in. An 8% gain feels even more urgent. A 10% gain triggers near-irresistible pressure to sell. The reasoning is intuitive: a gain is not a gain until it is realised. The market could reverse. The profit could vanish. Better to take what is on offer than risk giving it back.
But this reasoning is arithmetic, not geometric. It treats each gain as isolated rather than as part of a compounding sequence. A 10% gain that is taken eliminates the possibility that the same position could have delivered 20%, or 30%, or the 50% move that reshapes the equity curve for a decade. The investor who takes the 10% gain has not locked in a profit. They have capped the right tail of their distribution. They have voluntarily converted what might have been an outlier into an ordinary month.
The trend following process reverses this impulse systematically. The trailing stop does not exit a winning position at a predefined target. It exits only when the trend reverses by enough to indicate that the directional move has ended. Until that reversal occurs, the position is held. The process endures the discomfort of watching open profits fluctuate, watching the market pull back and recover, watching the equity curve wobble, because it knows that the largest gains come from the positions that are held the longest through the strongest trends. Selling early feels prudent. Holding feels reckless. The geometric outcome favours holding.
The Running Ledger
Our $100,000 continues. This episode adds the right-tail metrics that reveal the source of each benchmark’s geometric profile.
The S&P 500 has negative skew and a best-to-worst ratio below 1. Its extreme months work against compounding. Berkshire and the TF Index both have positive skew and best-to-worst ratios above 1. Their extreme months work for compounding. The TF Index achieves this with the smallest maximum drawdown and the highest MAR ratio of the three. The shape of its returns is the most geometrically efficient in the comparison.
The Bridge
The cut truncates the left tail. The trend ride extends the right tail. Together they produce a positively skewed distribution in which the outlier months that dominate terminal wealth are disproportionately positive. This is the geometric architecture of the trend following process: it earns its return not through steady accumulation but through asymmetric exposure to the extreme months that carry the most compounding value.
But the right tail of the trend following distribution has a further property that conventional portfolio theory cannot explain. The TF Index’s best months are concentrated in exactly the periods when equity markets suffer their deepest losses. Seven of its ten best months occurred when the S&P 500 was negative. This means the process is not merely producing positive outliers. It is producing positive outliers precisely when the rest of the portfolio is being destroyed.
Episode 8 will examine this property in full: crisis alpha, the tendency of the trend following process to deliver its strongest performance during the periods of maximum geometric danger. It will show that the negative correlation between trend following and equities during crises is not a coincidence but a structural consequence of the feedback architecture described in Episode 5. And it will demonstrate that the portfolio implications of this property are, by any geometric measure, extraordinary.
Data and Sources
All performance data from the NilssonHedge Trend Following Performance Database (January 2000 to January 2026). All returns are net of management and performance fees. 313 monthly observations for the S&P 500 Total Return, TTU TF Index, and Berkshire Hathaway. Best-month removal counterfactuals computed by setting the n largest monthly returns to zero and recalculating the cumulative product. Skewness calculated as the third standardised moment of monthly returns. 41 managers met the 20-year (240-month) minimum track record threshold. Bessembinder reference: “Do Stocks Outperform Treasury Bills?” (2018), Journal of Financial Economics. Disposition effect reference: Shefrin and Statman (1985). The best-to-worst ratio divides the absolute value of the best single month by the absolute value of the worst single month; ratios above 1 indicate the right tail extends further than the left. Win-to-loss ratio divides the average positive monthly return by the absolute value of the average negative monthly return.
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.