Traditional economics approaches market prediction through a causal chain: identify the fundamental drivers, model their relationships, and project forward. Interest rates, corporate earnings, macroeconomic indicators, and historical price patterns serve as the inputs, and the model produces a forecast of the most likely future state. This framework has genuine analytical value, and in conditions of relative stability, it produces useful approximations. Its failure mode, however, is not gradual deterioration. It is sudden, catastrophic breakdown at precisely the moments when accurate prediction matters most.
The 2008 financial crisis demonstrated this failure with exceptional clarity. The fundamental models in widespread use at the time analysed housing prices, credit spreads, and corporate balance sheets through established causal relationships. What they could not capture was the intricate web of interdependencies between mortgage-backed securities, the balance sheets of global financial institutions, and the feedback dynamics that converted an initial rise in mortgage defaults into a cascading system-wide collapse. The interconnectedness meant that a disturbance in one part of the system propagated through the entire market with a speed and completeness that no linear model predicted. Investor behaviour, driven by panic and the sudden repricing of risk, amplified the structural failure in ways that models focused on quantitative fundamentals were not designed to accommodate. The crisis was not an anomaly that slipped through an otherwise reliable framework. It was the predictable consequence of applying linear models to a system whose defining property is non-linearity.
Entropic Thinking as an Alternative Framework
The alternative is to approach financial markets not as systems moving predictably from one fundamental state to another, but as entropic systems: dynamic, probabilistic, and capable of transitioning between states in ways that cannot be fully predicted from prior conditions alone.
Entropic thinking does not attempt to identify the single most likely future state and draw a straight line toward it. It maps the landscape of possible states, estimates the probabilities of transitions between them, and designs strategies that remain viable across the full range of outcomes rather than optimising for the single expected one. The analogy is the difference between a weather forecast that predicts tomorrow’s temperature as a single number and a forecast that models the probability distribution of possible conditions, including the tail scenarios that the point estimate ignores entirely.
Applied to financial markets, this shift in perspective reframes what constitutes a sound investment strategy. A linear approach asks: what will happen? An entropic approach asks: what is the probability distribution of what might happen, and how should a strategy be positioned to survive the full range of that distribution, including the fat-tail scenarios that the distribution’s centre does not represent?
Systematic trend-following strategies are structurally aligned with entropic thinking. They do not attempt to predict specific market events or identify the fundamental cause of a price move. They analyse the directional momentum of price series and adjust positions based on the probability that observed trends will persist. This probabilistic orientation allows the strategy to remain responsive to the market’s actual behaviour rather than committed to a prior prediction about what that behaviour should be.
The Card Shuffle Analogy
The concept of entropy in financial markets is well illustrated by the behaviour of a deck of cards moving through successive shuffles.
Begin with a deck in perfect order, arranged by suits and numbers from ace to king. This represents a state of maximum order and minimum entropy: the position of every card is known, and the system is fully predictable. The first shuffle introduces some randomness. Most cards remain close to their original positions, but a few move significantly, appearing as outliers in the overall distribution. The deck retains a recognisable semblance of its original order. At this stage, the system remains largely predictable: the positions of most cards can be estimated with reasonable confidence, even if the exact location of the displaced few cannot. This initial disruption is analogous to a market experiencing minor fluctuations from small, isolated events. The overall structure is intact, and conventional models remain informative.
As shuffling continues, each subsequent pass further disrupts the original order. After several shuffles, no discernible pattern connects the current arrangement to the starting state. The deck appears disordered to any observer relying on knowledge of the original configuration. This progressive transition from order to apparent disorder illustrates the increasing entropy of the system: each shuffle represents a new state that follows from the preceding one but cannot be precisely predicted from the original. The system is not random in the sense of being memoryless. Each configuration is the direct consequence of the preceding shuffle. But the accumulation of small perturbations has made the system’s state effectively unpredictable from the initial conditions alone.
The financial market parallel is direct. Markets in stable regimes exhibit behaviour that conventional models describe reasonably well. Prices fluctuate, but the fluctuations cluster around a central tendency in a way that statistical models calibrated to recent history can approximate. As the regime evolves and the number of interacting variables and feedback loops accumulates, the system’s state becomes progressively less predictable from its initial conditions. The market equivalent of multiple shuffles is not time alone but the accumulation of interacting influences: geopolitical shifts, regulatory changes, technological disruptions, changes in market participant composition, and the feedback dynamics between all of these. Each adds entropy to the system, progressively decoupling the current state from the prior ordered baseline.
The crucial insight from the card shuffle analogy is that entropy increases predictive difficulty asymmetrically across the return distribution. In the early stages of shuffling, when the system retains most of its initial order, the state of most cards can be predicted reliably. But a small number of cards will have moved significantly from their starting positions, and their exact locations cannot be predicted. These are the outliers of the distribution. As shuffling continues, the proportion of cards in predictable positions decreases and the proportion in unpredictable positions increases. The tails of the distribution thicken. In financial terms, the market produces more fat-tail events as regime transitions accumulate, and fewer of its movements remain within the range that Gaussian models describe reliably.
Gaussian and Fat-Tailed Distributions in an Entropic System
Understanding the dynamics of an entropic market requires distinguishing precisely between the two distributional regimes that coexist within it.
Gaussian distributions describe the central behaviour of the market: the routine, everyday price movements driven by normal trading activity, minor news, and typical shifts in market sentiment. These movements cluster around the mean, and extreme deviations are rare. In the early stages of the card shuffle, before entropy has accumulated to the point of displacing many cards significantly, the distribution of positions is approximately Gaussian. Models calibrated to this regime produce reasonable predictions for the majority of outcomes, and the small losses that occur when predictions are wrong remain manageable.
Fat-tailed distributions describe the behaviour of the market’s tails: the rare, significant events that produce substantial displacements from prior equilibria. Economic shocks, major policy changes, geopolitical dislocations, and structural regime shifts produce outcomes that lie in the tail regions of the distribution. Continuing the card shuffle analogy, after several shuffles, a meaningful proportion of cards have been displaced significantly from their original positions. These displacements appear in the tail regions of the distribution. Gaussian models, calibrated to the central behaviour, systematically underestimate the probability of these tail outcomes. They treat extreme displacements as near-impossible when the entropic dynamics of the system make them structurally more frequent than the Gaussian assumption predicts.
The practical consequence for investment strategy is that a model that works well in the central regime of the distribution will fail at precisely the moments when the tail events arrive, which are also the moments when the failure is most costly. A strategy designed only for the Gaussian regime will be unable to accommodate the fat-tail events that an entropic system produces with structurally higher frequency than Gaussian assumptions predict. Conversely, a strategy designed to survive and benefit from fat-tail events will incur the cost of that preparation during the periods of Gaussian central-regime behaviour, in the form of frequent small losses accumulated while waiting for the tail events to arrive.
The appropriate response is not to choose between these two regimes but to design strategies that function across the full distribution. This means risk management that prepares explicitly for fat-tail events through diversification, hedging, and position sizing calibrated to tail scenarios rather than central tendency. It means trend-following approaches that adapt to both Gaussian and fat-tailed market behaviour, participating in the small predictable movements while maintaining the structural architecture required to capture the large unexpected shifts. And it means incorporating both distributional regimes into the models used for risk assessment, rather than defaulting to Gaussian assumptions that provide false comfort about the rarity of extreme outcomes.
The Illusion of Complete Disorder
The perception that an entropic system has become completely unpredictable is itself a cognitive error that entropic thinking corrects.
When a deck of cards has been shuffled many times and its original order is no longer recognisable, it is tempting to conclude that the system is now in a state of complete randomness. But this conclusion confuses the loss of the original reference point with the loss of all structure. Each configuration of the deck, however disordered it appears relative to the starting state, is a specific and unique arrangement that follows directly from the preceding shuffle. The system has not become random. It has become unpredictable from the perspective of an observer anchored to the original ordered state.
The same confusion arises in financial markets when analysts conclude that a market in apparent chaos has lost all structure. The market has not become random. It has transitioned to a configuration whose structure cannot be recovered by extrapolating from prior conditions. New patterns and new trends are continuously forming within the apparent disorder. The fat-tail events that appear as chaos from the perspective of a Gaussian model are, from the perspective of entropic thinking, transitions between market regimes: movements from one basin of attraction to another in the complex adaptive landscape of market states.
Markov processes and hidden Markov models represent an attempt to formalise the prediction of state transitions in entropic systems. They estimate the probability of moving from one market state to another based on the current state and a transition matrix calibrated to historical behaviour. These tools provide genuine predictive value in the short term, when the number of shuffles since the last regime establishment is small and the transition probabilities remain stable. Their limitation is structural: they assume that the transition matrix itself is stable, that the market cannot introduce transition types that have not been observed before. In an entropic system that generates fat-tail events and regime shifts by nature, this assumption is precisely the one that fails at the moments of greatest consequence.
The appropriate response to this limitation is not to abandon predictive models but to calibrate strategy design to the full range of possible transitions, including those that historical models have not encountered. This is the essence of the loose pants principle applied to entropic market thinking: strategies should be flexible enough to accommodate regime transitions that have not yet occurred, not precisely optimised for the transitions that have.
Adapting Strategy to the Entropic Market
Recognising that financial markets are entropic systems does not produce helplessness. It produces a more accurate framework for what strategies can and cannot be expected to achieve, and it points toward the design principles that are appropriate for a complex adaptive system rather than a linear one.
The card shuffle analogy is instructive here too. A strategy applied to a shuffled deck that attempts to predict the exact position of every card will fail comprehensively as entropy accumulates. A strategy that instead identifies the cards most likely to have moved in a consistent directional pattern, the cards whose displacement from the prior equilibrium has been sustained across multiple shuffles, and commits to following that directional displacement until it reverses, will extract genuine edge from the entropic system without requiring the full prediction that the system’s complexity makes impossible.
This is precisely the logic of systematic trend following applied to financial markets. The strategy does not predict specific market events. It does not attempt to identify the fundamental cause of a price move or to project a causal chain from current conditions to a specific future state. It identifies directional momentum in price series, a property that the entropic market produces when regime transitions create sustained, serially correlated movements in asset prices, and commits to following that momentum until the trailing stop signals that the trend has reversed.
In an entropic system where the sequence of shuffles produces occasional large, sustained displacements alongside frequent small random fluctuations, a strategy designed to participate in the large displacements while limiting exposure to the small random ones is structurally well-suited to the distributional properties of the system. The frequent small losses accumulated during the Gaussian central-regime periods are the cost of maintaining the architecture that captures the fat-tail events when they arrive. The positive skew of the resulting return distribution, frequent small losses offset by infrequent large gains, is the direct geometric consequence of applying an asymmetric exit structure to an entropic market that produces fat-tail events with structurally higher frequency than Gaussian models predict.
Investors who embrace this framework do not eliminate uncertainty. They align their strategy with the actual distributional properties of the system they are operating in, rather than with the simplified Gaussian approximation that linear models provide. The result is not certainty about any specific outcome but resilience across the full range of outcomes that an entropic market produces, including the fat-tail regime transitions that define the long-run compounding advantage of the Outlier Hunter’s approach.