How the same price series can trend, mean-revert, and appear random depending on when you look
The Coastline
How long is a coastline?
The answer depends on the length of your ruler. Measure with a hundred-kilometre ruler and you get one number. Measure with a one-kilometre ruler and the coastline grows longer as your measurement captures more inlets and bays. Measure with a one-metre ruler and it grows longer still, tracing every rock and crevice. There is no true length. There is only the length at a given scale of measurement.
This is not a quirk of geography. It is a property of complex, irregular structures. The coastline is fractal: its character changes with the scale at which you observe it.
Markets share this property. The same price series looks different depending on the timescale of observation. What appears as noise at one scale reveals structure at another. What appears as trend at one horizon appears as mean-reversion at another. The market does not have a single character. It has a character at each scale, and these characters can contradict each other.
Understanding this is essential. Your architecture operates at a specific timescale, often one imposed by constraints rather than chosen freely. If you do not know what the market looks like at that scale, you do not know the market you are trading.
The Same Series, Different Stories
Consider a single price series observed at different frequencies.
At the tick level, prices jump erratically. The bid-ask bounce creates apparent reversals that are pure microstructure noise. Patterns that appear significant are often artefacts of how orders arrive and execute. The signal-to-noise ratio is low. Structure is hard to find.
At the daily level, a different picture emerges. Short-term momentum becomes visible: in aggregate, prices that rose yesterday are slightly more likely to rise today. Volatility clusters. Trends begin to form and persist for days or weeks. The noise of microstructure fades and behavioural patterns emerge.
At the monthly level, the picture shifts again. Momentum that was visible at daily horizons begins to exhaust as the capital that drove it becomes fully deployed and valuation anchors begin to bind. Mean-reversion appears: assets that outperformed over the past three to five years tend to underperform over the next three to five. The forces that drove prices away from equilibrium at shorter horizons now pull them back.
At the multi-year level, the picture shifts once more. Secular trends driven by economic fundamentals, demographics, and technological change become visible. These trends are invisible at shorter horizons, obscured by the fluctuations that dominate daily experience.
Each horizon tells a different story about the same underlying series. The stories are not wrong. They are true at their respective scales.
Why Scale Changes Structure
The forces that move prices operate at different frequencies.
At short horizons, price movements are dominated by order flow, liquidity provision, and the mechanics of execution. These forces are largely unstructured relative to fundamental value. They create noise that obscures signal.
At intermediate horizons, behavioural forces become more visible. Momentum arises from slow information diffusion, herding, and feedback between price and positioning. Mean-reversion arises from arbitrage, value investing, and the eventual correction of mispricings. These forces create exploitable structure, but they operate over weeks and months, not hours.
At long horizons, fundamental forces dominate. Economic growth, inflation, interest rates, and earnings drive returns. These forces are more predictable in direction but slower in expression. They are visible only if you are patient enough to wait for them.
The market is not a single system with a single character. It is a superposition of systems operating at different timescales, where shorter-horizon dynamics can mask or delay the expression of longer-horizon forces. What you see depends on where you look.
Volatility Across Horizons
Volatility itself is scale-dependent.
If volatility scaled linearly with time, you could calculate long-horizon volatility by simply multiplying short-horizon volatility by the square root of time. This is what simple models assume. It is not what markets do.
In reality, volatility exhibits complex scaling behaviour. At short horizons, volatility is often higher than linear scaling would predict: intraday moves are larger relative to daily moves than simple models suggest. At longer horizons, mean-reversion in volatility compresses realised dispersion: annual volatility is often lower than you would expect from scaling daily volatility.
This matters for architecture. A system calibrated to daily volatility will underestimate short-term tail risk and may overestimate long-term dispersion. The same position can appear appropriately sized at one horizon and dangerously sized at another.
The volatility surface, discussed earlier in this series, encodes some of this scale-dependence. The term structure of implied volatility reflects how the market prices volatility at different horizons. But implied volatility is an expectation, not a guarantee. Realised scaling can diverge from what the surface suggests.
Correlations That Come and Go
Correlations are also scale-dependent.
Two assets may appear uncorrelated at daily frequency but highly correlated at monthly frequency, or vice versa. The diversification that exists at one timescale may vanish at another.
This is particularly important during stress. At short horizons, correlations tend to spike during market dislocations: everything moves together as liquidity withdraws and feedback mechanisms synchronise. At longer horizons, the diversification may return as idiosyncratic factors reassert themselves.
A portfolio that appears well-diversified at the horizon of years may offer little protection at the horizon of days. The correlation structure you measure depends on the window you use to measure it. There is no single true correlation, just as there is no single true coastline length.
Matching Architecture to Horizon
Every trading system, every risk model, every investment strategy operates at an implicit or explicit timescale.
A high-frequency strategy operates at milliseconds to seconds. It exploits microstructure patterns that exist only at that scale. It is irrelevant to an investor operating at monthly or annual horizons.
A momentum strategy typically operates at weeks to months. It captures behavioural patterns that emerge at that scale. It will underperform if evaluated at daily horizons, where noise dominates, or at multi-year horizons, where mean-reversion takes over.
A value strategy operates at years to decades. It exploits slow mean-reversion in valuations. It requires patience that shorter-horizon participants cannot or will not exercise.
Mismatch between strategy and horizon is a common source of failure. A momentum strategy evaluated on daily returns will appear to underperform. A value strategy evaluated on monthly returns will appear to lag. The strategy is not failing. The evaluation horizon is wrong. Much of what appears as disagreement about markets is actually disagreement about timescale.
Architecture must be designed for a specific horizon. Risk limits, position sizing, rebalancing frequency, and performance evaluation all need to match the timescale at which the strategy operates. A system that mixes horizons without awareness will produce confusion and poor decisions.
The Horizon You Inhabit
You do not choose your horizon freely. Constraints choose it for you.
A day trader is constrained by the need to close positions before market close. An institutional investor is constrained by quarterly reporting and annual performance reviews. A pension fund is constrained by liability duration that stretches decades into the future.
These constraints are not merely practical. They determine what the market looks like to you. The day trader sees a market of noise and microstructure. The pension fund sees a market of secular trends and mean-reverting valuations. They are looking at the same prices, but they are not seeing the same market.
Understanding your horizon means understanding what structure exists at your scale and what structure is invisible to you. It means accepting that patterns visible to others operating at different scales may be irrelevant to your situation, and patterns visible to you may be invisible to them.
The Ruler and the Coast
The coastline has no true length. It has a length at each scale of measurement, and these lengths are all valid.
Markets have no single character. They have a character at each timescale, and these characters are all real. Trend and mean-reversion, signal and noise, structure and apparent randomness coexist depending on where you look.
Your architecture is a ruler. It measures the market at a specific scale. What it reveals at that scale is real. What it obscures at other scales is also real, but it is not yours to capture.
The question is not what the market is doing.
The question is what the market is doing at your horizon.
This is the tenth article in a series exploring the deep structure of markets. Next: “Position Sizing: The Geometry of Survival”
Want the theoretical foundation for why trend following works?
The Fractals of Finance: Determinism, Adaptation and the Geometry of Markets bridges complexity science with practical trading implementation. With a foreword by Jerry Parker, original Turtle Trader.
Available now on Amazon in paperback, hardcover, and Kindle.
