Expectancy, Edge, and Why Most Traders Think About This Wrong
The formula is correct. The world it assumes does not exist. And the difference between those two facts determines whether a programme compounds or collapses.
A pendulum, observed at a single moment, reveals almost nothing about its behaviour over time.
You see a position and a velocity. You do not see the arc. You do not see the regularity, the pattern of return, the structural consistency that only becomes visible across many swings. The single observation looks like a fact. The long series of observations is the only thing that tells you what the system actually is.
Most traders measure their edge the way someone measures a pendulum by taking one photograph. They count wins and losses. They calculate averages. They produce a number that feels like a verdict on the system’s quality, and they manage the programme in response to that number.
The number is not wrong. It is incomplete in a way that matters more than it being wrong, because an incomplete measurement applied with confidence produces exactly the kind of systematic error that compounds quietly for years before announcing itself.
This Foundation is about what edge actually is, how the standard measurement misrepresents it, and why the Outlier Hunter’s relationship with expectancy is built on a different and more honest foundation.
The Formula and Its World
Expectancy is defined as the average outcome per trade: win rate multiplied by average win, minus loss rate multiplied by average loss. If the result is positive, the system has a positive expected value. If it is negative, the system is a long-run loser. The formula is clean, logical, and widely taught as the fundamental test of whether a trading approach is worth pursuing.
The problem is not with the arithmetic. The problem is with the world the formula assumes. Expectancy belongs to ergodic systems: environments where the average outcome across many parallel universes converges reliably with the average outcome across a single long timeline. Roll a fair die enough times and your results will converge toward the expected value of 3.5. The ensemble average and the time average are the same thing. The formula works.
Markets are not ergodic. Your trading programme is not ergodic. There is one path through time, and the sequence of outcomes along that path matters in ways the expectancy formula cannot capture. A loss early in the programme’s life, when capital is limited and the drawdown is deepest, does more damage than the same loss later. Two programmes with identical average trade outcomes but different sequences of those outcomes end up at different places, sometimes vastly different places, after a decade of operation.
Expectancy calculates the average across a world of parallel outcomes. It says nothing about the single path you actually walk. In a non-ergodic system, that omission is not a minor technical qualification. It is the difference between a measurement that reflects reality and one that flatters a dangerous illusion.
Why Win Rate Is the Wrong Metric
Of all the ways traders misread their edge, the most pervasive is the focus on win rate. What percentage of trades are profitable? This feels like a natural measure of quality. More wins than losses seems like evidence of skill. A system that is right most of the time feels like a system worth trusting.
The Outlier Hunter’s system is right on fewer than half its trades. Deliberately. By design.
A well-constructed systematic trend following programme wins on roughly thirty-five to forty percent of its trades. The majority of positions are opened, fail to develop into trends, and are closed for small losses. This is not a flaw in the system. It is the direct mechanical consequence of two rules that define the approach: cut losses quickly, and let winners run until the evidence that the trend has ended is clear enough to act on.
Cutting losses quickly produces many small losses. Letting winners run produces occasional large gains. The arithmetic of the system is: many small debits against a small number of large credits. The credits must be substantially larger than the debits to produce a positive geometric return over time. In a well-functioning Outlier Hunting programme, they are. The average winning trade is a multiple of the average losing trade, and that asymmetry is the entire source of the edge.
A trader who measures this system by win rate and concludes it is broken after a sequence of losses has measured the wrong thing. A sequence of losing trades is not evidence that the edge has deteriorated. It is the normal experience of operating a system designed to accept many small losses in exchange for the occasional large gain. The signal that something is genuinely wrong is not the frequency of losses. It is the size distribution of winners: if the occasional large gains have stopped arriving, or have shrunk to the size of the average loss, the system’s structural edge is worth investigating. Win rate tells you nothing about this.
“The system that is right thirty-five percent of the time and makes five times the average loss on each win is more valuable than one that is right sixty-five percent of the time and makes one-point-two times. This is not obvious until you run the numbers across a thousand trades.”
Where the Outlier Hunter's Edge Actually Lives
The edge in Outlier Hunting is not distributed evenly across trades. It is concentrated in a small number of extraordinary outcomes.
In a fat-tailed return distribution, the few events in the right tail of the distribution produce outcomes that are not merely larger than average but categorically different in scale. A single sustained trend in crude oil, or a sustained move in a major currency pair, or a year in which fixed income trends across geographies simultaneously: any one of these events, captured by a programme that was positioned and holding, produces returns that exceed the accumulated result of hundreds of ordinary trades.
This concentration is not a feature of good years. It is the structural property of the return distribution across all years. Remove the ten best trades from a twenty-year trend following track record and the long-run performance collapses. The median trade produces a small loss. The mean trade produces a small gain. The geometric return is driven almost entirely by the outliers at the far right of the distribution.
The implication for how the programme is managed is significant. The question is never whether the current sequence of losses indicates the edge has gone. The edge is structural, arising from the feedback dynamics of complex adaptive markets that create and sustain trends. It does not disappear because the last thirty trades were losers. The question is whether the programme remains positioned correctly to capture the outliers when they arrive: broadly diversified, correctly sized, with exit rules that hold winning trades through the noise long enough for the fat-tail moves to develop fully.
Expectancy, as a formula, smooths the outliers into the average. It treats the trade that returns twenty times its risk as a data point weighted equally with the trade that returns one-point-one times its risk. In doing so, it obscures the architecture of the edge entirely. The Outlier Hunter does not manage the programme by its average trade outcome. They manage it by whether the conditions for outlier capture remain intact.
The Correct Measure: Geometric Return
The measurement that reflects what actually happens to capital over time is the geometric return: the annualised rate at which the programme compounds across the full period of operation, accounting for both the direction and the volatility of the return stream.
The geometric return is always lower than the arithmetic average return, and the gap between them grows with the volatility of the return stream. A programme averaging fifteen percent per year with thirty percent annualised volatility compounds at a materially lower geometric rate than one averaging the same fifteen percent with fifteen percent volatility. The second programme accumulates more capital over a decade even though both programmes produced the same average annual return. The difference is variance drag, the compounding cost of volatility described in Foundation 3.
This relationship is real, and the Outlier Hunter takes it seriously. Variance drag exists. Adverse return sequences cost real capital, and that cost compounds against the programme year on year. But the geometric return formula, in its standard form, treats all variance as equivalent. Upside variance and downside variance are summed into a single number and penalised at the same rate. This treatment is correct for symmetric return distributions. It is incorrect for the kind of distribution an Outlier Hunting programme actually produces.
The Outlier Hunter’s return distribution is structurally asymmetric. Left-tail outcomes are kept small by mechanical stops on initial risk, sized from closed balance equity in the manner described in Foundation 2. Right-tail outcomes are pursued aggressively using unrealised equity, the open profit between the high-water mark and the trailing stop. The two tails are not symmetric. They are managed by entirely different rules and produce entirely different consequences for the compounding base.
This asymmetry has a direct consequence that the standard geometric return formula does not capture. Variance drag is a real cost on the left side of the distribution, where adverse sequences erode realised capital. It is not a cost in the same sense on the right side, where the variance is unrealised profit moving toward an exit that will lock in a step up in the closed equity base. The same statistical variance number, decomposed into its left-tail and right-tail components, has very different consequences for the rate at which capital actually compounds.
The Outlier Hunter accepts that compounding drag is a genuine cost on adverse returns. The programme more than compensates for that cost through the positive convexity of its right tail. Each captured outlier is a step up in closed balance equity, a discrete jump in the compounding base that no symmetric variance calculation can produce. Across enough years and enough markets, the accumulation of those step ups produces a geometric return that exceeds what the variance-only formula predicts for a programme of the same average return and the same total variance.
The result is a different equity geometry than the smooth, low-variance line that conventional managed futures programmes pursue. The Outlier Hunter’s equity curve is stepped. Closed balance equity rises in discrete moves as winning trades close above the previous high-water mark. Between those moves, total equity is volatile, sometimes substantially so, as open positions develop and retrace. But closed equity, the figure that compounds, only moves upward in the meaningful sense, because losses on losing trades are kept small by initial risk control and gains on winning trades are realised only when the trailing stop is struck above prior closed equity.
Each step is a captured outlier. The flat or volatile portions between steps are the periods when unrealised equity is being deployed in pursuit of the next step. The staircase is the visible signature of positive skew. The straight line is what symmetric programmes produce when they succeed and what conventional managed futures programmes are designed to approximate. The staircase is what a programme produces when it pursues asymmetric outcomes and protects the realised base while doing so.
This is why the Outlier Hunter does not respond to a high-volatility period by reducing position sizes to suppress variance. Volatility on its own is not the relevant signal. The relevant signal is whether closed balance equity is being threatened. If the volatility is unrealised profit retracing toward a trailing stop, the programme holds. If the volatility is closed equity being eroded by a sequence of realised losses, the Cut Back Rule engages, as Foundation 2 described, scaling position sizes down proportionally until closed equity recovers.
The geometric return is the honest scorecard. It reflects what the programme actually delivered to capital over time, not what it would have delivered in a world without compounding drag. But the scorecard is most accurate when read against the right distribution. Evaluating an Outlier Hunting programme against the assumptions that produce a smooth equity curve will systematically understate what the programme is doing, because the formula penalises variance that is doing useful work alongside variance that is doing damage. The honest measurement is the one that recognises the asymmetry and reads the geometry that asymmetry produces.
Absorbing States and the Limits of Positive Expectancy
There is one further dimension of the expectancy problem that no amount of good average-trade mathematics can resolve: the absorbing state.
An absorbing state is a condition from which exit is impossible. In trading, the absorbing state is ruin: the loss of sufficient capital to continue operating the programme meaningfully. Once entered, no subsequent run of positive results can undo it. The game is over.
A system with positive expectancy can still lead a trader into an absorbing state. The expectancy formula assumes an infinite number of trials. It assumes the programme is always alive to take the next trade. It does not account for the possibility that a sequence of losses, perfectly consistent with the system’s statistical properties, depletes the capital to the point where the programme cannot continue.
This is precisely why position sizing, covered in Foundation 2, is the foundational constraint that must be satisfied before any discussion of expectancy is meaningful. A programme with excellent expected value per trade but excessive position sizes will encounter its absorbing state during the first significant adverse sequence. A programme with modest expected value per trade but conservative position sizing will survive that sequence, and the next one, and the one after that, until the outlier that justifies the entire approach finally arrives.
Survival is not a separate objective from performance. It is the precondition of performance. Expectancy is irrelevant to a programme that does not survive long enough for its expected value to assert itself.
“Measure the system's edge by what it does to capital across the full path, not by what the average trade produces in isolation. The path is everything. The average is a fiction.”
Edge Is a Property of the System in Its Environment
One further point about edge that the expectancy frame consistently misses: edge is not a fixed property of a strategy. It is a property of a strategy in a specific market environment.
A systematic trend following programme has edge in trending markets and generates losses in ranging, mean-reverting ones. This is not a flaw to be corrected. It is the definition of the approach. The programme deliberately accepts losses in non-trending conditions in exchange for being positioned to capture the outsized gains when sustained directional moves develop.
The long-run edge depends on trending conditions being sufficiently frequent and sufficiently large to more than compensate for the losses accumulated while waiting for them. The historical evidence across more than two centuries of markets and across every major asset class confirms that this condition is met. Trends occur, they persist longer and extend further than a random walk would produce, and the fat-tailed distribution of returns means the large moves, when they arrive, generate returns that dwarf the accumulated losses preceding them.
The structural edge arises from the feedback dynamics of complex adaptive markets. The way participant behaviour creates and amplifies directional moves. The way crowding and momentum reinforce each other. The way information asymmetry and behavioural anchoring sustain trends beyond what fundamental value analysis would predict. This mechanism is not an inefficiency to be arbitraged away. It is the process by which markets discover value through alternating cycles of overshoot and correction. You cannot arbitrage away price discovery. You can only participate in it.
The Edge Has Not Died, It Has Relocated
The most common objection to systematic trend following in the past decade is that the edge has decayed. Observed returns across the simplest divergent strategies have declined from their peak in the 1980s and 1990s. By the mid-2010s, conventional wisdom held that systematic capital had crowded in, that the era of easy trends was over, and that simple trend following was a strategy in terminal decline.
The objection is empirically wrong, and resolving it carefully matters for how the Outlier Hunter understands edge.
The site’s own research, drawing on forty-one years of daily data across sixty-eight global futures contracts, tested the decline narrative directly. The underlying feedback mechanism that produces trends, measured as the amplitude of oscillation between positive and negative feedback regimes across decades, has not decayed. The most recent decade in the sample shows oscillation amplitude statistically indistinguishable from, and slightly higher than, the first decade in the sample. After four decades of explosive growth in systematic capital and increasingly sophisticated competition, the structure that produces trends remains intact. The data refutes the decay hypothesis.
What has happened instead is more nuanced. The decline in simple trend following returns is the joint product of four separable forces, three of which are reversible or implementation-correctable, and only one of which is genuinely structural.
The first and dominant force is regime suppression. The variance ratio, the statistic that measures whether returns are trending or oscillating, varies by macro environment. The 2009 to 2020 period of quantitative easing and zero interest rates compressed the variance ratio across the entire universe. Trends were starved of the macro fuel they require to develop. When the macro environment delivered genuine raw material again, during the 2020 to 2022 inflation, every divergent strategy in the test recovered to pre-QE quality. This force is cyclical, not structural, and it can reverse with macro conditions.
The second force is short-side impairment. Central bank intervention, passive index flows, and systematic dip-buying have structurally compressed the standalone alpha of trading shorts. This is the only one of the four forces that is genuinely structural and likely permanent. But, as Foundation 3 established, symmetric models overstate the damage. Asymmetric calibration, with shorter lookbacks for short signals, recovers measurable improvement at the portfolio level. And short-side portfolio utility remains intact regardless of standalone alpha: the negative bear-market correlation, the convex tail payoff, the regime robustness that prevents catastrophic outcomes during sustained declines.
The third force is the horizon gradient. Within any regime, shorter lookback horizons deliver less edge than longer ones. The twenty-day breakout strategies are dead in every macro regime tested. The three-hundred-day strategies recovered substantially post-2020. The edge has redistributed from shorter to longer horizons, and a programme running an ensemble of horizons captures the diversification benefit of cross-horizon independence, which has been increasing rather than decreasing over time.
The fourth force is the asymmetric calibration finding from Foundation 3. Bull and bear markets operate at different frequencies. A symmetric model is permanently misaligned with one side. Asymmetric calibration, with long lookbacks of two hundred to three hundred days paired with short lookbacks of ten to thirty days, captures the geometry that the symmetric model cannot. This is a frequency mismatch, not a missing edge.
The four forces interacted to produce the appearance of monotonic structural decline in simple trend following. They were not the same kind of phenomenon. One was cyclical macro suppression. One was structural-but-overstated by symmetric measurement. One was a redistribution. One was a frequency mismatch corrected by asymmetric calibration. Decade-level performance summaries conflated all four into a single narrative of decay. The data tells a more honest story.
The implication for the Outlier Hunter is twofold.
First, the structural edge is durable. It rests on the feedback dynamics of complex adaptive markets, which have demonstrably persisted across four decades of adaptive pressure and across every major asset class tested. The case for divergent strategies is stronger and more nuanced than the conventional decline narrative suggests, not weaker. The reader who internalises this is positioned to ignore the cyclical noise of any given period and focus on whether the structural conditions that produce edge are in place.
Second, simple symmetric trend following is dying, even though the structural edge is not. Capturing the edge requires reading the structure correctly. Regime-aware exposure, horizon-diversified ensembles, asymmetric calibration of long and short signals, and structural rather than statistical diversification across markets. The Outlier Hunter’s edge therefore depends on implementations that respect the structure, not on the persistence of any particular set of historical parameters.
For the empirical case underlying this section, including the four-decade amplitude analysis, the regime decomposition, the short-side decomposition, the horizon gradient findings, and the asymmetric calibration grid search, see The Fractals of Finance research series, particularly the episodes on the paradox, the escalator and the elevator, and the verdict.
What Comes Next
The Outlier Hunter does not need to predict when the next trend will arrive. The programme only needs to be present, solvent, and correctly positioned when it does. That is what the first three Foundations were building toward, and what this Foundation has tied together at the level of edge itself. Position sizing keeps the programme alive. Diversification keeps the programme present. The understanding of edge developed here keeps the programme calibrated to the structure that actually produces returns rather than to the surface metrics that conceal it.
Foundation 5 addresses the environment in which trends form. Most traders treat noise as the obstacle the system must overcome. The Outlier Hunter’s programme is built on a different and more accurate understanding: noise is not what the trend must escape. Noise is the medium in which the trend is born. And the market condition that most traders read as safety, the quiet, low-volatility environment, is the one the Outlier Hunter regards with the most caution.
READ DEEPER
→ Expectancy vs Survival: Why the Outlier Hunter Thinks Differently
→ Beyond Expectancy: Protecting the Path of the Outlier Hunter
→ The Hidden Costs of Certainty: Why the Best Trades Are Uncomfortable
→ Who Holds the Risk: Why Markets Survive and Traders Do Not
→ The Fractals of Finance: The Paradox
→ The Fractals of Finance: The Escalator and the Elevator
→ The Fractals of Finance: The Verdict
Previous: Foundation 3: What Diversification Actually Does (and What It Cannot Do) | Next: Foundation 5: The Role of Noise: Why Quiet Markets Are Dangerous
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.
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.Lorem ipsum dolor sit amet consectetur adipiscing elit dolor