The greatest threat to the trend follower is not a bad month or a bad year. It is the decade-long bull market where the process quietly bleeds while buy-and-hold investors celebrate. The geometry of wealth extracts its price in patience.
The previous eleven episodes have built the case for trend following as a geometric compounding engine: the mathematics, the mechanism, the crisis alpha, the portfolio synergy, the ensemble evidence. The case is strong. But strong cases have a weakness: they can sound like sales pitches. This episode exists to prevent that.
Every geometric benefit documented in this series has a cost. The positive skew comes from a low win rate. The crisis alpha comes from years of underperformance during bull markets. The drawdown containment comes from a process that spends most of its time in drawdown. The negative equity correlation comes from a return stream that moves opposite to the asset class most investors are benchmarked against. The benefits compound. The costs grind. This episode quantifies those costs in detail. The reader who finishes it and still wants to allocate to trend following will do so with open eyes. That is the goal.
The Ergodicity Problem
The physicist Ole Peters posed a deceptively simple question: would you accept a bet that pays +50% or -40% with equal probability? The arithmetic expected value is positive, (+50% plus -40%) divided by two, or +5% per round, and an economist trained in expected value would accept it. But an individual playing this game repeatedly would go broke. A sequence of +50%, -40%, +50%, -40% does not compound to +5% per round. It compounds $1.00 to $1.50 to $0.90 to $1.35 to $0.81. After four rounds the player has lost 19% despite a positive arithmetic expectation.
This is the ergodicity problem. In a system where outcomes are multiplicative, the time average, what a single individual experiences over time, diverges from the ensemble average, what a large group experiences at a single moment. Most of finance optimises for the ensemble average. The individual investor lives the time average. The distinction is not academic. It is the reason the arithmetic mean overstates the compound return, the central insight of Episode 2.
The trend-following process is designed for the time average. By cutting losses (limiting the -40% outcome), maintaining diversification (reducing the correlation between sequential bets), and sizing positions to survive drawdowns, the process approximates the bet size that maximises the geometric growth rate, without relying on the known probability distributions that formal optimisation frameworks require. It does not maximise the expected value of any single round. It maximises the growth rate of the entire sequence. This is why trend following’s arithmetic return is often lower than the S&P 500’s. It is optimising for a different objective: not the average outcome, but the path-dependent outcome that determines actual terminal wealth.
The framework explains a puzzle that has run through the entire series. How can a strategy with a lower arithmetic return produce comparable or superior terminal wealth? Because arithmetic return is an ensemble quantity, the average across many parallel paths, while terminal wealth is a time quantity, the product along a single path. The strategy that maximises the ensemble quantity (the S&P 500, with its higher arithmetic return) does not necessarily maximise the time quantity (terminal wealth), because the time quantity depends on the path. The S&P 500’s path includes a 50.9% drawdown, and that single event, experienced in time rather than averaged across an ensemble, consumes years of compounding. Trend following sacrifices ensemble performance for time performance. And time performance is what the investor actually lives.
If you have absorbed the ergodicity argument, you can already predict the costs that follow. A process optimised for the time average will sacrifice win rate, accepting many small losses to avoid the large losses that are catastrophic in time but merely bad in expectation. It will look weak in fixed rolling windows, because any fixed window is an ensemble sample, not a time sequence, and the process appears weak in the years when no crisis forces the two averages to diverge. It will spend more time below its peak, because it is not chasing the highest available monthly return at every moment. And it will feel worse than it computes. The five costs quantified below are not five separate problems. They are one problem stated five ways: the problem of living in time rather than across an ensemble. Every cost is the ergodicity argument in a different unit of measurement.
The Cost Ledger
The costs of trend following are real, measurable, and persistent. They are not hidden in the data. They are the data. Table 1 sets out the five, each against its S&P 500 comparison over the full 313-month window.
Table 1: The five patience costs of standalone trend following, TF Index versus S&P 500, January 2000 to January 2026, net of fees. Every figure is computed directly from the monthly return series.
Each cost is the same mechanism seen from a different angle. The win rate is low because the cut is doing its job. The grind is the texture of a process that enters and exits many small positions while waiting for a trend to persist. The time in drawdown is the price of staying invested across many markets at once. The rolling underperformance is the price of crisis alpha. And the return shape is the positive skew that Episodes 7 and 11 documented, optimised for mathematics rather than comfort.
Time in Drawdown
Figure 1 shows the drawdown profiles side by side. The contrast is the whole argument in one image. The S&P 500 experiences fewer drawdown episodes but far deeper ones, twice breaching 40% and once reaching 50.9%. The TF Index spends more of its history below its previous peak, but its worst decline never exceeds 21.0%. The geometrically destructive drawdowns, the ones that consume years of compounding, occur only in equities.
Figure 1: Drawdown profiles for the S&P 500 (top) and TF Index (bottom), Jan 2000 to Jan 2026. The S&P has fewer but deeper drawdowns; the TF Index spends more time underwater but never beyond 21%. The 40%-plus drawdowns that compound destructively appear only in equities.
Rolling Underperformance
Figure 2 traces the rolling five-year CAGR of each series. On a rolling three-year basis the TF Index underperforms the S&P 500 in 68.7% of all windows; on a rolling five-year basis it trails in 63.4%. The investor evaluating trend following at a random point is more likely than not to be looking at underperformance. The maximum gap reached 22.2 percentage points in February 2014, when the S&P’s five-year CAGR was surging and the TF Index’s was near zero.
Figure 2: Rolling 5-year CAGR, S&P 500 versus TF Index. Blue shading marks periods where the S&P leads; green where the TF Index leads. The maximum gap of 22.2 percentage points occurs in February 2014. The TF Index leads during and after crises; the S&P leads through extended bulls.
That 22-point maximum is not simply a quiet period for trend following. It is the product of a specific and identifiable regime. The 2010 to 2019 decade was shaped by coordinated central-bank intervention that compressed directional persistence to its weakest readings in decades, by a structural impairment of the short side as passive inflows and policy backstops supported falling markets, and by crowding at standard lookback horizons that pushed the available edge toward longer, less populated frequencies. The patience cost during those ten years was not random underperformance. It was the predictable consequence of an unprecedented policy experiment that deliberately suppressed the raw material trend following harvests. When the experiment ended, the process recovered immediately: the TF Index returned 11.3% annually from 2020 through 2022 while the S&P 500 suffered a 24.6% drawdown. The cost was regime-specific. The recovery confirmed the mechanism was intact.
The Return Distribution
Figure 3 places the two monthly return distributions side by side. The S&P 500 wins more often, 65.2% of months, but its distribution is negatively skewed: the left tail extends further, holding the rare but devastating months. The TF Index wins less often, 55.3% of months, but its distribution is positively skewed: the right tail extends further, holding the occasional large gain that drives terminal wealth. The TF shape is geometrically superior and psychologically harder to endure, because behavioural research consistently finds that losses are felt about twice as keenly as equivalent gains. A process of many small losses and occasional large gains is optimised for mathematics, not for comfort.
Figure 3: Monthly return distributions, S&P 500 (left) and TF Index (right), Jan 2000 to Jan 2026. The S&P’s negative skew puts its long tail on the loss side; the TF Index’s positive skew puts its long tail on the gain side.
The Emotional Accounting
Table 2 translates the statistical costs into the lived experience of the investor between crises.
Table 2: The patience cost translated into lived experience. The geometric benefits, the crisis alpha, the positive skew, the drawdown containment, are invisible during these periods. They are stored in the structure of the return distribution, waiting for the crisis that releases their value.
The investor must hold through the cost to capture the benefit. During the years between crises, the benefits are not absent. They are latent, encoded in the shape of the distribution, and they pay out only when a crisis forces the time average and the ensemble average apart.
Why the Cost Is Worth Paying
The costs are real. The question is whether they are worth the price, and the answer depends on what the investor is buying. There are four things on the receipt.
First, drawdown containment. A maximum drawdown of 21.0% versus 50.9%. The recovery cost of a 21% drawdown is 27%; the recovery cost of a 51% drawdown is 104%. The difference in recovery burden is not arithmetic, it is geometric, and every percentage point of avoided drawdown compounds forward through every subsequent month. Episode 8 showed that 61.5% of the TF Index’s terminal wealth was generated during crisis months. The patience cost is the premium for this geometric insurance.
Second, portfolio synergy. Episode 9 demonstrated that a 60/40 S&P/TF blend produces more terminal wealth than either component alone. The blended portfolio does not experience the full patience cost, because the equity component compounds during bull markets while the TF component provides geometric protection during crises. The investor who evaluates trend following in isolation sees the cost; the investor who evaluates the blend sees the synergy.
Third, ergodic optimality. The process maximises the geometric growth rate of the full sequence, not the arithmetic expectation of any single period. The low win rate, the treading water, the rolling underperformance: these are the cost of a process designed for what actually happens to capital over time rather than what is expected to happen on average. The S&P 500’s higher win rate and higher rolling CAGR feel better, but its GFC drawdown is the price of that comfort, and that price compounds destructively through the entire recovery.
Fourth, and less obvious: a patience cost that is regime-conditional rather than constant. The aggregate statistics, 68.7% of windows underperforming, 82.7% of months in drawdown, are averages across all regimes including the exceptional QE decade. Outside that decade the rolling underperformance is less frequent, the drawdowns shallower relative to equities, and the distribution tighter toward the positive. The 2010 to 2019 decade is the single most hostile environment trend following has faced in four decades, produced by a deliberate and ultimately temporary compression of directional persistence. Investors who treat that decade’s patience cost as the expected baseline are building their expectations from the worst case. The 39-year extended record, which includes the QE decade in full, still shows both DUNN WMA and EMC Classic outcompounding the S&P 500. The cost is real. Its severity is regime-dependent.
The insurance analogy is imprecise but directionally correct. Trend following costs money during calm markets and pays off during crises, like insurance. But it differs in one critical respect. Insurance is a negative expected-value proposition for the buyer: the premiums, over time, exceed the expected payoffs. Trend following, over full market cycles, has a positive geometric expected value. The premiums, the patience cost during bull markets, are more than recovered by the payoffs, the crisis alpha during bear markets, because those payoffs protect the compounding base from the geometrically destructive drawdowns equities suffer. The insurance is not free, but unlike fire insurance it is, over the full cycle, self-financing and then some. The patience cost is not a defect. It is the mechanism.
The Behavioural Trap
The most destructive cost of trend following is not statistical. It is behavioural. Most investors who abandon trend following do so at exactly the wrong time, and the pattern is predictable. An investor allocates after reading about crisis alpha and geometric efficiency. For two or three years the strategy performs reasonably. Then a sustained equity bull market begins. The TF allocation stagnates while the benchmark surges. After three years of underperformance the investor begins to doubt the process; after five years they reduce the allocation; after seven, with the rolling gap at 15 or 20 points, they eliminate it entirely.
The elimination occurs near the end of the bull market, precisely when crisis alpha is about to become most valuable. The investor who held through the 2010 to 2019 decade was rewarded with the 2020 to 2022 period in which the TF Index returned 11.3% annually while the S&P 500 fell into a 24.6% drawdown. The investor who abandoned the allocation in 2018 or 2019 captured the full cost and none of the benefit. This is the behavioural trap. The process demands the one thing markets are designed to punish: patience and discipline while everyone around you is doing something different. The equity investor surrounded by rising prices feels validated; the trend-following investor surrounded by the same rising prices feels punished. Both experience the same environment. Only one is being compensated for the next crisis.
The trap is not limited to individuals. A pension fund that allocates 20% to trend following and watches it underperform for five years faces questions from the board, from consultants, and from beneficiaries. The geometric argument, that the allocation will prove its value in the next crisis, is always prospective; the underperformance is always present. Most institutional allocators who cut their trend-following allocation cite underperformance as the reason. They are, in effect, cancelling the insurance policy because the house has not caught fire recently.
The solution, to the extent one exists, is structural rather than psychological. The investor should not rely on willpower to maintain the allocation through the barren years. They should build it into a rebalancing framework executed mechanically, without discretion. Episode 9’s blended portfolio, rebalanced monthly to target weights, is exactly such a framework. It does not require the investor to decide each quarter whether trend following still works; it requires them to set the allocation once and maintain it through the cycle. The rebalancing also helps: during the underperformance years it buys more trend following at lower prices, and during the crisis payoff it trims at higher prices. The discipline is embedded in the process, not in the investor’s psychology.
The Honest Summary
Trend following costs the investor in five measurable ways: a win rate below 50% on a standalone basis, months of near-zero returns, 83% of history spent in drawdown, rolling underperformance against equities in the majority of windows, and a return distribution that feels worse than it compounds. These costs are structural. They will persist. They are not a problem to be solved. They are the price to be paid.
The question is not whether the costs are real. They are. The question is what they buy. They buy a maximum drawdown of 21% versus 51%. They buy a MAR ratio of 0.359 versus 0.157. They buy a blended portfolio that outperforms both components. They buy crisis alpha that turns the worst months for equities into the best months for the process. They buy positive skew in a world where equities deliver negative skew. And they buy all of this net of fees.
It is also worth restating what the investor in a blended portfolio actually experiences. The costs above describe the standalone TF Index. In the 60/40 S&P/TF blend from Episode 9 they are substantially diluted. The equity component compounds during bull markets, partially offsetting the TF component’s stagnation; the blend’s win rate is higher than the standalone TF Index’s; the rolling underperformance against a pure equity benchmark is narrower; and the lived experience is materially less painful. The costs documented here represent the worst case, the investor who holds 100% trend following. The realistic case, a 40 to 60 percent allocation within a diversified portfolio, is considerably more tolerable.
The geometry of wealth is not free. It demands patience measured in years, discipline measured in decades, and the willingness to look wrong for longer than most investors can tolerate. The investor who cannot pay this price should not allocate to trend following. The investor who can pay it has access to the geometric properties the first eleven episodes documented. The cost is the price of the compounding.
The Bridge
The series has now presented both sides of the geometric argument: the benefits, in Episodes 1 through 11, and the costs, in this episode. The reader has the full picture. The remaining episodes turn from analysis to action. Episode 13 begins Act IV, Building Geometric Wealth. It provides the toolkit for evaluating trend-following programs through the geometric lens developed in this series: which metrics matter, which are misleading, and how to distinguish a geometrically efficient machine from one that merely looks good on an arithmetic basis. The series shifts from understanding the geometry to using it.
Data and Sources
All performance data is drawn from the NilssonHedge Trend Following performance file for the period January 2000 to January 2026 (313 monthly observations), net of management and performance fees. Win rate is the percentage of months with positive returns. Time in drawdown is the percentage of months in which the cumulative return is below its prior peak; the longest drawdown is the greatest number of consecutive months below a prior peak. Rolling CAGR windows are computed on a monthly rolling basis (36 months for three-year, 60 for five-year figures). Maximum drawdown is the largest peak-to-trough decline; MAR is CAGR divided by the absolute maximum drawdown. All figures in this episode were verified directly against the source monthly returns of the S&P 500 Total Return and the TTU TF Index series.
The ergodicity discussion draws on Ole Peters and Murray Gell-Mann, “Evaluating Gambles Using Dynamics” (2016), and Peters, “The Ergodicity Problem in Economics” (Nature Physics, 2019). The growth-optimal bet-sizing reference is J. L. Kelly, “A New Interpretation of Information Rate” (Bell System Technical Journal, 1956). Behavioural loss-aversion findings reference Kahneman and Tversky, “Prospect Theory: An Analysis of Decision Under Risk” (Econometrica, 1979). The structural decomposition of the 2010 to 2019 regime (variance-ratio compression, short-side Sharpe decay, and horizon crowding across 68 futures markets) derives from a separate research study and is not reproducible from the performance database; it should be cited directly to that study when this episode is published.
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.