Price does not move because information arrives. It moves because traders act. That distinction is not semantic. It is the fault line between two fundamentally different models of how markets work, and which side you stand on determines almost everything about how you approach systematic trading.
The dominant framework in traditional economics is the Efficient Market Hypothesis. Under EMH, markets are consolidation devices: rational participants each form private estimates of an asset’s fundamental value, and market price reflects the aggregate of those estimates at any given moment. When new information enters the system, rational participants absorb it immediately and revise their estimates accordingly. Price updates instantly. Because the information itself is unpredictable, price moves are also unpredictable, each interval in a time series independent of the last. Serial correlation does not exist. Speculation is therefore a fruitless exercise, because any arbitrage opportunity requires the lag that EMH insists cannot exist.
This is an elegant model. It is also, increasingly, an empirically unsupportable one.
The Evidence Against the Information Model
Jean-Philippe Bouchaud, whose work on market microstructure has significantly advanced our understanding of how price actually behaves, set out to test a simple but powerful question: how many large price moves can be directly attributed to news? Using one-minute equity data and investigating directional moves of plus or minus four standard deviations, his research found that more than 90% of these events could not be assigned to any identifiable news event from Bloomberg, Reuters, or comparable sources. Only around 10% of major price moves had a traceable exogenous origin. The remaining 90% arose from within the market itself, from endogenous factors: the internal dynamics of collective trader behaviour.
This does not mean that news is irrelevant. A declaration of war or a central bank surprise will clearly move prices. But in a model where collective trader impact is the primary driver, even exogenous events work through the same mechanism: they alter trader behaviour, and it is that altered behaviour, the order flow, that physically moves price. Information may be the trigger, but the gun is always the same.
What makes Bouchaud’s findings more damaging still for EMH is not just the origin of price moves, but their shape. Under EMH, price should update immediately and then stabilise. The data shows something different. Both exogenous and endogenous price jumps are non-local in impact: they spread out in time, and the volatility surrounding them takes considerable time to decay. For endogenous events in particular, meaningful uncertainty persists for up to 300 minutes following the peak of a price jump. These are not discrete, independent events. They are extended disturbances with clustering properties, precisely the kind of serial dependence that EMH insists cannot exist.
The Excess Volatility Problem
Robert Shiller identified a related anomaly in what he called the Excess Volatility Puzzle: short-term price fluctuations are simply too large to be explained by changes in fundamental value alone. Something beyond information is driving volatility. The candidate Bouchaud and others propose is the feedback dynamics of collective trader behaviour, positive and negative loops that amplify or dampen signals in a lagged, non-local fashion, creating the kind of serially correlated price behaviour that a complex adaptive system (CAS) would naturally produce.
Introducing noise traders, participants with no informational edge, into the model is instructive here. Under EMH, uninformed traders should have no durable effect on price, because only genuine information can move prices. But if noise traders do contribute to long-run price volatility, and the evidence suggests they do, then the order-flow itself, whether informed or random, is the causal agent. The market is not a rational aggregator. It is a path-dependent, non-ergodic system in which collective behaviour shapes price independently of fundamental value.
From Microstructure to Macrostructure
The microstructural evidence aligns with findings at a larger scale. Research by Gabaix and Koijen, whose Inelastic Markets Hypothesis builds on this body of work, finds that a trader buying or selling one dollar of an individual stock increases or decreases that stock’s market capitalisation by approximately one dollar in the long run. The mechanism is the activity itself, not the information behind it. For index-level purchases, the multiplier rises to approximately five times, a finding that strains any model premised on rational price equilibration.
The decisive evidence, however, comes from the Square Root Law. Studies of metaorders, the incremental execution of large trades split across many smaller orders over minutes or days, reveal that price impact does not scale linearly with volume. It scales with the square root of volume. The second half of a metaorder consistently impacts price less than the first half. This is not consistent with a market that immediately and rationally absorbs information. It is consistent with a market that has a memory, a liquidity memory window during which the influence of past trades persists before eventually decaying.
This memory effect is formalised in Latent Liquidity Theory, which proposes that each participant holds a reservation price that updates as a function of time, news, prior price moves, and noise. Most of this liquidity is never publicly posted. It is latent, invisible to the order book, and it shifts gradually rather than instantly. Price moves are therefore discrete and jump-like, not smooth and continuous. They hover around an estimate for a period before stepping to a new level. The assumption of immediate price re-equilibration, the cornerstone of EMH, does not survive contact with this evidence.
What This Means for the Outlier Hunter
The implication for systematic trend following is direct. If collective trader impact, not information, is the dominant driver of price, then markets are not efficient in the EMH sense. Serial correlation exists. Lags exist. Order flow has a memory. The kind of extended, directional price moves that Outlier Hunters are designed to capture are not anomalies to be arbitraged away: they are a structural consequence of how markets actually function.
Trend following does not work despite the way markets behave. It works because of it.
This article draws on research covered in depth by Richard Brennan and Niels Kaastrup-Larsen at Top Traders Unplugged. For a more detailed treatment of the underlying research, including Jean-Philippe Bouchaud’s microstructural findings, read the full article at Top Traders Unplugged.