What a Drawdown Actually Means
A drawdown is not a verdict on the system. It is a structural feature of every honest programme. How you interpret it determines whether you survive long enough to find out what you are actually running.
Stand on a coastline that has not been visited in a hundred years and what you see is not the work of the previous day, the previous week, or the previous quiet decade. You see the cumulative product of long uneventful periods of small surf and a smaller number of storm events that did most of the actual work of shaping the shore. The headlands are storm-cut. The bays are storm-deepened. The sea stacks were left standing because the storms removed everything around them. The visible coastline is a record of two phenomena operating together: long stretches of small change and short concentrated bursts of large change. Neither phenomenon alone explains the shape. Both together do.
The equity curve of an Outlier Hunting programme is shaped the same way. The visible result is the cumulative product of long periods of small accumulation and a smaller number of structural events that did most of the actual work. Some of those structural events are adverse. They are the deep drawdowns that test every conviction the framework was built on. Some of those structural events are favourable. They are the outliers that justify the entire approach. Neither phenomenon alone explains the equity curve. Both together do, and the geometry of how they combine is what this Foundation is about.
A drawdown is not an interruption of the programme. It is one of the two structural phenomena that constitute the programme’s actual operation. Living through one is not the abnormal state. It is part of what running an Outlier Hunting programme is.
The Definition and the Geometry
A drawdown is the decline in portfolio value from a peak to a subsequent trough, measured as a percentage of the peak. It begins the moment the portfolio sets a new high and ends only when that high is recovered and exceeded.
By this definition, every programme that ever generates a return also generates drawdowns. There is no compounding path that does not pass through periods of loss. The question is not whether drawdowns will occur. It is what they reveal about the system, and what the correct response is when they do.
The geometry of recovery, first encountered in Foundation 2, applies here with full force. A 25% drawdown requires a 33% gain to recover. A 40% drawdown requires a 67% gain. A 50% drawdown requires a 100% gain. Each additional percentage point of depth adds a disproportionately larger recovery requirement, because the base from which the recovery must be made shrinks with each loss. The cost of a drawdown is not the loss itself. It is the loss plus the compounding that did not occur during the recovery period. Both costs accumulate silently while the programme is below its high-water mark.
This is why the Cut Back Rule, introduced in Foundation 2, is not optional. It is the mechanism that prevents a normal drawdown from deepening into a structural one. As depth increases, position sizes are reduced formulaically. The programme moves itself further from the ruin boundary before the next adverse sequence can push it closer. When recovery begins, sizing scales back up. The programme returns to full exposure as capital is restored, not in anticipation of the recovery but in response to it.
Two Drawdowns That Are Not the Same Thing
Before going further, a distinction from Foundation 2 must be brought back into the centre of this discussion, because the rest of the essay depends on it. The word “drawdown” is being used in trading writing to describe two different events that have very different meanings for an Outlier Hunting programme.
A drawdown in total equity, which includes unrealised profit and loss on every open position, occurs whenever a winning trade retraces from its high-water mark on its way to the trailing stop. Every captured outlier in the programme’s history produces such a drawdown at the end of its life. The position runs into profit, the trailing stop follows the price up, the trend exhausts, the price retraces, the stop is eventually struck, and the position closes well below its peak unrealised value. The total equity figure declines through that retracement. This is not a drawdown in the relevant sense. It is the structural mid-life of an outlier capture, doing exactly what an outlier capture is supposed to do. The Cut Back Rule does not engage. The programme holds.
A drawdown in closed balance equity is different in kind. It occurs when the realised compounding base, the figure that represents the programme’s actual progress through time, is being eroded by an accumulating sequence of closed losing trades. Positions are opening, hitting their stops, and closing into the realised account at small losses. Across many markets, across weeks or months, those small losses accumulate. The closed equity figure declines. This is the drawdown the Cut Back Rule responds to, the drawdown the rest of this essay is principally about, and the drawdown that produces the operational pressure to override the validated process that Foundation 6 named as the structural threat to systematic execution.
The distinction matters because the two events feel similar to a trader watching the equity curve and are managed by entirely different rules. A retracement of unrealised profit on a winning trade in cocoa or fixed income is not a programme drawdown. It is the geometry of capturing the trend. A sustained accumulation of small losses in the realised account is a programme drawdown, and it triggers the formulaic response described in Foundation 2.
For the rest of this essay, “drawdown” refers to the second event: the decline in closed balance equity that arrives when the programme is operating through a market environment the system is not built to capture, and that the Cut Back Rule was designed to manage. Where unrealised retracements are relevant, they are named explicitly.
Warehoused Risk and the Release Mechanism
A drawdown in closed equity does not arise from nothing. It is the visible accumulation of warehoused risk being released into the realised account.
Every position a programme opens carries risk that has not yet been expressed as a closed outcome. The programme may have dozens of open positions at any given moment, each carrying an unrealised result that sits between the entry price and the current stop. That aggregate of unrealised risk is warehoused within the open position register. It is not present in the closed equity figure because nothing has yet been closed. It is real risk, sitting in the portfolio, awaiting the conditions that will resolve it in one direction or the other.
When a regime shifts, when the trending conditions that justified the positions give way to a ranging environment that works against them, the warehoused risk begins to release. Positions hit their stops in sequence. The losses that were unrealised become realised. They flow from the open position register into the closed account, one position at a time, as each stop is struck. Across many markets and weeks or months, those realised losses accumulate. The closed equity curve, which had been advancing as winning trades closed above the previous high-water mark, turns and descends as losing trades dominate the closed flow. The drawdown is not the arrival of unexpected risk. It is the scheduled conversion of warehoused open-position risk into realised losses in closed equity.
Understanding this changes the interpretive frame. The trader who sees a drawdown as an attack on the portfolio from outside is likely to respond defensively, closing positions, reducing exposure beyond what the Cut Back Rule specifies, or abandoning the process entirely. The trader who understands it as the scheduled release of warehoused risk through the closure mechanism is more likely to let the programme respond mechanically, as it was designed to, and to hold through the release until the trending regime returns.
“The drawdown you are in always feels different from the ones in your backtest. That feeling is the human response to non-ergodic risk. It is real. It is manageable. But only if you understand what it is.”
What a Drawdown Does Not Prove
A drawdown in a non-predictive structural system is not evidence that the system has stopped working. This is the interpretation that destroys more systematic programmes than any market condition. The most natural response to watching capital decline over weeks or months is to conclude that the approach is broken, that the market has changed, that the historical edge no longer applies. For a programme of the kind this series describes, that conclusion is almost always premature and is most likely to be reached at exactly the moment the programme is closest to its recovery.
The distinction matters because not all systematic processes are alike. A predictive system attempts to forecast where prices will be at some future moment. A drawdown in such a system carries direct diagnostic weight, because the forecast was wrong and the view that justified the position has been refuted. A non-predictive structural system does not forecast. Its rules read structural features of how markets actually work as complex adaptive systems: the formation of breakouts from compressed states, the development of trending regimes, the alternation between feedback states described in the Fractals research. The Outlier Hunter’s programme is of this second kind. It does not predict that crude oil will rise. It reads the structure of crude oil’s price action and responds to specific terrain features when those features develop. When a position in crude oil moves adversely and hits its stop, no prediction has been refuted, because no prediction was made. The drawdown is an accumulation of these closed positions across markets where the terrain did not develop into the trending regimes the system requires for profit. It is not a failure of forecasting. It is the cost of waiting in a non-predictive frame.
For the Outlier Hunting framework, drawdowns carry no diagnostic weight about edge integrity. The edge is structural, arising from how markets work rather than from the system’s ability to predict. Drawdowns in this framework are the mechanical cost of operating a non-predictive process through a market environment where the structure the system reads is not currently expressing itself. They do not reveal that the system has stopped working. They reveal that the conditions for its work are not currently present. There is a separate signal that does carry diagnostic weight, and that signal is the subject of a later section. It is not the drawdown.
The systematic record of interventions during drawdowns is not encouraging. Programmes abandoned or modified during drawdowns that would have recovered lose the recovery. Programmes modified during drawdowns to reduce the very rules that produced the historical edge lose the edge too. The pattern is consistent enough to be treated as a structural property of how humans interact with systematic processes under stress: the intervention that feels responsible is, statistically, the one most likely to convert a temporary drawdown into a permanent impairment. This is the override impulse named in Foundation 6, expressed through the specific operational pressure that drawdowns produce.
A drawdown is also not bounded by historical precedent. The trader who has studied the system’s historical maximum drawdown sometimes treats that number as a ceiling: if the drawdown reaches that level without having reached it before, something genuinely unusual must be happening. In a fractal market, the historical maximum is not the ceiling of future risk. It is only the largest adverse structure that the sample has so far been long enough to expose. The next section addresses why.
A drawdown is also not uniform in its causes. The shape, speed, and composition of a drawdown carry information that a single percentage number does not. A sharp, correlated drawdown that arrives rapidly across many markets simultaneously suggests a systemic event: a regime shift, a liquidity shock, a crisis in which correlations have converged and every position is being repriced simultaneously. This kind of drawdown tends to be followed by the sustained directional moves that trend following captures most powerfully. A slow, grinding drawdown that accumulates over many months across many small losing positions suggests a ranging environment in which trends are absent and the programme is paying the normal cost of waiting for the next trending regime to develop. Each type calls for patience, but the character of the patience required is different.
The Worst Drawdown Is Always Ahead, and So Is the Largest Outlier
The most important and most uncomfortable truth about drawdowns in a fractal market is this: regardless of how long a programme has been running, regardless of how extensive its historical record, the worst drawdown it will ever experience has not yet occurred. The best one has not occurred either. Both claims are true given enough time, and the structural reason is the same.
The reason is easier to see through an image than through statistics.
Imagine a tree that is itself a fractal system. Small twigs branching from larger branches, larger branches from limbs, limbs from a single dominant trunk. If you sample only a small section of this tree, you are likely to find leaves and twigs and perhaps a few small branches. You are unlikely to sample the trunk. The small section misrepresents the tree by under-representing the largest structural features that define what the tree actually is. As you increase the size of the sample, you progressively expose more of the underlying structure. You encounter larger branches, then limbs, eventually the trunk. The tree was always like this. The small sample showed only the periphery. The larger sample reveals the architecture.
Markets behave the same way. A short historical sample shows the small structural features: minor moves, contained ranges, ordinary trends. As the sample is extended, it begins to expose the larger structural features. The Volcker shock of the early 1980s. The collapse of 1987. The dotcom decline. The 2008 cascade. The 2020 inflation regime. The cocoa trend of 2024. These are not statistical anomalies bolted onto an otherwise smooth distribution. They are the trunks and major limbs of the market’s actual fractal structure, present at every scale, exposed by the sample only when the sample is long enough to encounter them.
The Fractals of Finance research provides empirical support for what the structural argument predicts.
Drawing on forty-one years of daily data across sixty-eight global futures contracts, the research finds that five-sigma events appear roughly five thousand seven hundred times more often than the bell curve predicts. The Hurst exponent averages 0.866 across the universe, far above the 0.5 baseline that would describe a random walk. The autocorrelation of absolute returns averages 0.353 at lag one and remains positive and significant for more than a year. These are the statistical signatures of a fractal system whose tail is not a remote possibility but a structural feature that progressively reveals itself as the sample grows. The events that define long historical records are the trunks. They were always there. They were exposed by the sample being long enough to encounter them.
The implication for drawdowns is uncomfortable. The programme that has run for twenty years and survived a maximum drawdown of thirty percent has not proved that thirty percent is the limit of what it will face. It has proved that twenty years of operation did not happen to expose any structural feature larger than thirty percent of its capital. These are not the same statement. The next twenty years of operation will sample the same fractal market more deeply, and that deeper sample will eventually contain larger structural features than the first twenty years did. A drawdown larger than anything in the first twenty years is not improbable. It is structurally implied by the geometry of the system, given sufficient time, and it becomes more likely with each year of additional sampling, not less.
This is why the position sizing and Cut Back Rule discipline established in Foundation 2 must be maintained unconditionally, not relaxed because the programme has a long history of surviving its worst historical conditions. The long history is evidence of survival, not evidence that the tail has been exhausted. The tail is the trunk. The trunk is always larger than what the small sample has so far shown.
The same structural fact applies on the favourable side, and this is the half of the argument that justifies the entire approach. The fractal structure that guarantees worsening adverse events also guarantees larger favourable ones. The same deeper sample that will eventually contain a drawdown larger than anything in the historical record will eventually contain an outlier larger than anything in the historical record. The Outlier Hunting framework is not built around the worst drawdown ahead. It is built around the largest outlier ahead, with an architecture that contains the first while remaining structurally present for the second. Small left tail, controlled by mechanical stops sized from closed balance equity. Uncapped right tail, captured by trailing stops that hold winners as far as the trend will carry them. The asymmetry of the architecture is what makes the asymmetry of the structure compoundable rather than fatal.
The reader who internalises only the worse-drawdown half of this argument has half the case. The reader who internalises both halves has the full case. The largest favourable structural feature in your trading career has not yet arrived, and neither has the largest adverse one. The discipline of the Outlier Hunting programme is to remain present, solvent, and correctly positioned for both, in the knowledge that both are structurally guaranteed by the same property of the system, and that the architecture of the programme is what determines whether the symmetric exposure compounds or destroys the capital base.
“In a fractal market, the worst drawdown is always ahead of you, not behind you. The historical maximum is the floor of your experience, not the ceiling of your risk.”
What Should Genuinely Concern an Outlier Hunter
The argument above says that drawdowns in a non-predictive structural system carry no diagnostic weight about edge integrity. This raises a question. Is there any signal that should genuinely concern an Outlier Hunter? The answer is yes, and there are two, operating at different timescales.
Before naming them, it is worth being clear about what is not a diagnostic.
The signal is not depth. A deep drawdown in closed equity is what the architecture is designed to handle through the Cut Back Rule. Depth alone tells you the programme is operating through an adverse period of the kind that has occurred before and will occur again.
The signal is not duration. A long drawdown is psychologically corrosive but not diagnostic. The structural conditions for trends arrive on no fixed schedule. Eighteen months without a new high is consistent with a market environment that is producing few trends. It does not mean the system has lost the ability to capture trends when they arrive.
The signal is not the experience of being in the drawdown. As Foundation 6 established, the override impulse is most intense at the moments when the discipline of non-interference is most necessary. The discomfort of holding through a drawdown is not evidence that something is wrong with the system. It is evidence that the system is being run through the conditions it was designed to be run through.
The two genuine diagnostics ask a different question. They ask whether the system is aligned with the structural features it was designed to capture. They operate at different timescales and through different mechanisms, but they are asking the same underlying question.
The Map-to-Market Test
The first diagnostic operates at the regime level and can be performed continuously while the programme is running. It is the simplest test there is: plot the equity curve against the market price data over the same period, and observe whether the two move together in the way the system’s design intends.
The Outlier Hunter’s system is designed to capture trends. Trends are visible in the market price data. The equity curve should reflect the presence or absence of those trends. If the market price data shows sustained directional moves and the equity curve is rising in step-ups that correspond to those moves, the system is doing what it was designed to do. If the market price data shows sustained directional moves and the equity curve is flat or in drawdown, the system is failing to capture what it should be capturing. If the market price data shows no sustained directional moves and the equity curve is rising anyway, the system is finding edge in places it was not designed to find it, which is a different kind of mismatch but equally diagnostic.
Both kinds of mismatch matter. The first is the more obvious failure mode: the system has stopped being aligned with the conditions it was designed for, perhaps because of an execution problem, a parameter drift, or a change in the structural features the rules respond to. The second is a more subtle failure mode but equally significant: the system is producing returns from sources other than the trends it was meant to capture, which suggests the system has acquired a structural bias unrelated to the fat-tail mechanism. A system finding edge in noise is not capturing the edge it was built to capture, even if the edge feels favourable in the short term.
The map-to-market test is qualitative in implementation but rigorous in its logic. The trader who knows what the markets did during a given period can perform the test by visual inspection. Plot the equity curve. Plot the major markets in the universe over the same period. Look at whether the rises in the equity curve correspond to the periods when those markets produced sustained directional moves, and whether the flat or declining periods in the equity curve correspond to the periods when those markets did not. The correspondence does not have to be perfect, because individual markets do not all produce trends at the same time, and the equity curve is the aggregate of the universe rather than any single market. But the broad correspondence should be present. When it is not, the question worth investigating is why.
This test has an important property the deeper diagnostic does not have: it can detect a mismatch within a single regime, often within months rather than years. The trader does not need to wait for the right tail of the lifetime distribution to develop. The map-to-market test asks a more immediate question, and the answer is available continuously as the programme runs.
The Size Distribution of Winning Trades
The second diagnostic operates at the lifetime distribution level. Foundation 4 named this as the diagnostic that requires the longest timeframe to evaluate but provides the deepest signal about whether the structural mechanism that produces the programme’s edge remains intact.
If the major step-ups have stopped arriving, or if the gains realised from winning positions have shrunk to the size of the average loss, the structural property the system was designed to capture, the fat-tailed distribution of returns with its rare but structurally large favourable events, may be operating differently in the current environment. Not because the system has predicted wrong, but because the structural feature itself may have shifted in a way that the system’s design did not anticipate.
In practice, this signal is rare and slow-developing. It does not arrive in the middle of a normal drawdown. It develops over many years of accumulating evidence that the right tail of the return distribution has changed in character. The Outlier Hunter who watches for it correctly is not anxious during normal drawdowns and is appropriately careful during the unusual periods when the right tail’s behaviour genuinely deserves examination.
The Two Diagnostics Together
The two tests are complementary rather than redundant. They detect different kinds of failure at different timescales.
The map-to-market test is faster, more direct, and detectable within single regimes. It catches the system that has drifted out of alignment with the conditions it was designed for. The size-distribution-of-winners test is slower, more inferential, and requires years of accumulated data. It catches the deeper question of whether the structural mechanism that produces the right tail of returns remains operative in the markets the programme is operating in.
A programme that passes both tests is a programme operating as designed in conditions that continue to support its design. A programme that fails the map-to-market test but passes the size-distribution test has an alignment problem worth investigating in the near term. A programme that fails the size-distribution test has a deeper question to engage that no near-term remedy will resolve. The two diagnostics together provide a more complete picture than either alone.
The discipline is to know what to watch and what to ignore, and to recognise that depth and duration of drawdown are not on the watch list. The watch list is short. Map the equity curve to the market data. Watch the right tail of the return distribution across many years. Everything else, including the discomfort of holding through any specific drawdown, is the structural cost of operating in a fat-tailed market and is not the diagnostic.
The Cocoa Precedent
The Cocoa Conundrum, documented in detail on the site, is a concrete reference point for the relationship between drawdowns and the structural events that resolve them.
Cocoa’s extraordinary trend in 2024 produced returns that defined the year for many systematic programmes. The years preceding the breakout looked like every other accumulation period the framework has experienced. Small losses in cocoa specifically, ranging behaviour, warehoused risk releasing into the closed account through stops being struck. The Outlier Hunter holding cocoa positions through that period was running a non-predictive structural system through a market environment that was not yet producing what the system was built to capture.
When the conditions developed, the system was present and positioned. The breakout was caught. The trailing stop followed the trend up. The position ran for months, accumulated unrealised profit, retraced from its high-water mark as the trend exhausted, and closed when the stop was struck. The closed equity took a discrete step up. The drawdown that had preceded the position was resolved not by the system changing but by the system continuing.
This is what drawdown management protects. Not the avoidance of drawdowns but the preservation of the programme through them, so that the architecture is present when the trunk events that resolve them arrive. Cocoa is the favourable case. The adverse equivalents are the deep drawdowns that resolve through the same architecture continuing to operate. The system does not know which is coming. It is built to be present for both.
Reading the Drawdown That Is Currently Happening
When a programme is in a drawdown, three dimensions of it deserve attention. They do not change the correct behavioural response, which is to let the system run and apply the Cut Back Rule mechanically. But they inform how the drawdown is understood, which affects the quality of the patience the trader can bring to holding through it.
Depth tells you how far the programme has fallen from its high-water mark in closed balance equity. A drawdown at ten percent of peak is within the normal operating range of most systematic trend following programmes. A drawdown at thirty percent is deeper, consistent with the historically bad periods for the approach, and likely to be producing significant psychological pressure. Neither number, by itself, tells you whether the edge is intact. The edge is structural, as the previous sections established, and does not depend on any specific historical drawdown range.
Duration tells you how long the programme has been below its high-water mark in closed equity without setting a new high. Duration is often the harder dimension to live with. A deep but brief drawdown resolves before the psychological pressure has time to compound. A shallow but extended drawdown, one that grinds along for twelve or eighteen months without a new high, tests patience in a different and often more corrosive way. The programme does not look catastrophically broken. It just looks like it has stopped working. The length of that ambiguity is what drives most interventions, and it is what the override impulse from Foundation 6 acts on most powerfully.
Shape tells you something about the cause. A drawdown that arrived rapidly across many correlated positions on specific days suggests an event-driven cause: a regime shift, a correlation event, a liquidity shock. A drawdown that accumulated gradually across many small losses over many markets suggests an environmental cause: the market is ranging and the programme is paying the normal cost of waiting for the next trending regime.
None of these three dimensions, individually or together, engages the diagnostic questions that actually matter. Those questions, the map-to-market alignment and the size distribution of winners, were addressed in the previous section. The depth-duration-shape framework is for reading the drawdown’s character, not for evaluating the edge. The edge is evaluated through the two diagnostics established earlier, not through the depth, duration, or shape of the drawdown the programme is currently inside.
What Comes Next
Foundation 7 has addressed the most operationally difficult experience in systematic trend following: living inside a drawdown with the correct understanding of what it is and what it is not. The correct response is mechanical, not psychological. But the mechanical response can only be maintained if the understanding is genuine. The drawdown is the structural cost of operating a non-predictive system through a market environment where the structure the system reads is not currently expressing itself. It is one of two phenomena that constitute the equity curve, the other being the outliers that justify the entire approach. The architecture of the Outlier Hunting framework contains the first while remaining structurally present for the second, and the discipline of holding through the drawdown is the operational condition for the architecture to work as designed.
Foundation 8 addresses the tool most commonly used to develop the understanding of drawdowns before the live drawdown arrives: the backtest. Specifically, why the backtest is necessary but not sufficient, where it fails to represent what the live programme will encounter, and how to use it correctly rather than as false reassurance that the worst is already known. The trunk events we have just discussed are the events backtests systematically under-represent, and Foundation 8 will examine why.
READ DEEPER
→ Why Your Drawdowns and Opportunities Are Always Ahead of You
→ Who Holds the Risk: Why Markets Survive and Traders Do Not
→ The Cocoa Conundrum: A Classic Trend Follower’s Perspective on Risk
→ Better Brakes, Faster Gains: Unlocking the True Power of Convexity in Outlier Hunting
→ The Fractals of Finance: The Random Walk Is Dead
→ The Fractals of Finance: The Impossible Keeps Happening
Previous: Foundation 6: Why Systematic Beats Discretionary: The Evidence | Next: Foundation 8: The Difference Between Backtesting and Reality
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