The Future Is Not Hidden, It Is Unwritten
What a forecast does when its subjects can act on it
Previous: Episode 1, What Bohm Saw
Next: Episode 3, What Survives Is What Could Not Be Eliminated
Imagine a bar that is pleasant when fewer than sixty people turn up and crowded when more do. You want to go on Thursday, but so does everyone else. A prediction that the bar will be quiet brings people through the door. A prediction of a crowd keeps them at home. W. Brian Arthur’s El Farol problem makes an awkward fact plain: when people use an expectation to decide what to do, the expectation can help change what it predicts. [1]
The rule sounds easy until a quiet forecast is posted and everyone who read it turns up. A second forecast warns of a crowd, and people stay away. The room changes because the guests use the prediction.
The forecast walks into its own story
Suppose a researcher finds that a breakout of a particular kind has often continued. A small trader applies the rule in a liquid market. The trade might have no discernible effect on the wider price. Now imagine the rule is taught, automated and used by funds managing much larger sums. Some participants anticipate the entries; others provide liquidity at less generous prices. The breakout may arrive earlier, become more violent or offer less return after costs. None of those possibilities follows automatically. The point is that the rule’s adoption becomes part of the setting in which its result is measured.
This is feedback in a precise sense. An observation produces an expectation. The expectation produces orders. The orders affect price and available liquidity. Those become the next observation. A model fitted to the first part of that loop may fail when its use alters a later part.
The size and reach of the expectation matter. My one order is not a large fund’s liquidation. A private model is not a public recommendation followed by thousands of accounts. Feedback should be investigated at the scale on which it could work, rather than invoked as a universal explanation for every failed backtest. Yet when the mechanism is present, the forecast has ceased to be a neutral description. The forecast is a participant dressed in the language of observation.
Arthur’s bar gives this problem a memorable small stage. It does not say that every prediction will defeat itself. Some forecasts are robust, some participants never see them, and some prices reflect forces far larger than the forecaster. It says that the conditions of a forecast include its audience and their capacity to act. In adaptive markets, I must make that part of the question.
In finance, the rules are multiplied by capital and institutions. A trader may enter because of an economic view, another because an investor redeemed, another because a stop was reached. The same price move can be a signal for one participant and a constraint on another. Some actions cancel; some reinforce. The finished price is a visible settlement among them. To understand why it might change sharply, I ask which rules could become aligned and what would make participants act together.
Hidden or unwritten?
If I believe the future is hidden, my main problem is access. I look for a better model, a bigger dataset or an expert whose signal can uncover what is already there. If I understand that tomorrow’s market has not yet been formed by tomorrow’s participants, my task changes. I still study history and estimate probabilities. I stop searching for a completed future and prepare to participate in one that is taking shape.
Nassim Nicholas Taleb has pressed the difference between uncertainty concealed by an inadequate model and outcomes that our retrospective stories make look more predictable than they were. [2] I take a practical lesson from that distinction. We should be suspicious of the polished explanation that arrives only after a rare event and describes the path as though it had been available in advance. We should also be suspicious of the claim that an adaptive future is merely a puzzle that a more powerful computer will solve.
I am making no claim here about whether the universe is deterministic. “Unwritten” describes my situation inside a market: tomorrow’s prices have not yet been formed by tomorrow’s orders and responses. That is enough to change where I put my attention.
In the language of Episode 1, today’s explicate prices and positions are part of the conditions from which tomorrow unfolds. The implicate structure includes the constraints and relations under which participants can act, not a ledger containing tomorrow’s completed closing price. Some of those relations are stable enough to study. Others change as the people who make the market learn. The lesson is not that knowledge is impossible; it is that the kind of knowledge available to an embedded trader is conditional and has to remain open to revision.
This is where hindsight plays a trick. After a sustained trend I can identify the report, policy shift or positioning that seems to have set it in motion. I might have known those facts at the start. What I could not know was the full sequence of later decisions that would carry the move farther, interrupt it or reverse it. Retrospective explanation makes one realised path easy to narrate. It often conceals how many plausible continuations existed when I had to act. The archive of other signals that failed is my defence against that trick.
The distinction does not forbid prediction. Conditional forecasts can be useful. A manufacturer may hedge a known exposure; an investor may assess a range of possible returns; I may estimate whether the historical evidence for a trend rule is strong enough to justify risking capital. The mistake is to treat a conditional estimate as knowledge of a completed event, or to forget that others’ responses may change the condition.
A method of participation
Here is the trading process I actually use. I seek sustained price moves across a diversified set of futures markets. Each eligible market begins with a small, equal allowance of initial risk. I use average true range, or ATR, to scale the position to recent price movement. ATR is a useful measure of that movement, not a measure of every risk in a contract or an assurance that realised losses will be equal. Entries and trailing exits are predefined. I use no profit targets, because I cannot identify the exceptional move in advance or know how far it will run. Most attempts fail or produce little. A few large outliers must pay for many of them.
An exit is an instruction, not an insurance policy. A gap, interruption or poor execution can produce a loss larger than the amount I intended at entry. Several markets can turn together. My Cut Back Rule reduces exposure when its conditions are triggered by closed balance equity. That rule is part of the strategy tested and adopted before the difficult week arrives, rather than an improvised response to fear. I keep implementation records because a beautifully researched rule is of little use if the live orders do something else.
This arrangement does not tell me what tomorrow will bring. It tells me how I intend to behave when the next signal arrives, when a trade survives, when it fails, and when the account follows an adverse path. A choice made before the outcome can later be judged against the record. I prefer that to a narrative rewritten around whichever trade just won or lost.
There is a real tension in remaining open to evidence. Change the rules every time they hurt and the researched method never gets traded. Refuse to revisit them and discipline becomes loyalty to a deteriorating arrangement. I separate live execution from scheduled research and review. A proposed change must have a stated reason, a test and a date. It does not acquire legitimacy merely because I felt relief after making it.
One reason this process feels different from forecasting is that it leaves room for a move larger than my initial expectation. A breakout is an invitation to take a limited position, not an assertion that I know where price will end. The exit follows the market’s subsequent path. My forecast, if the word applies, is about a distribution of possible trades under the rule. The individual result remains open.
Consider two valid entries in different markets. The first reverses almost at once. The second climbs, stalls, gives back part of its gain and then resumes for months. At entry, both presented the same kind of decision. The rule lets the unfolding market separate them over time. If I had demanded a persuasive account of why the second would win, I might have rejected both or selected the first for a story that later proved false. The design chooses repeated limited attempts and a way to stay with the occasional move that survives. That is how I participate in a future I cannot read in advance.
The Cut Back Rule adds another temporal element. It responds to realised account outcomes under conditions set before the drawdown. A reduction in exposure may preserve the capacity to continue if losses cluster, although it may leave the account smaller when a recovery starts. That trade-off is part of the method’s tested path. It is different from cutting size because the last few losses suddenly feel exceptional. The first can be evaluated with the strategy; the second changes the strategy at the very point when its design is being tested.
Who is writing?
The person in the episode’s image stands above moving water. Their view may help them judge where a wave will break, but the wave does not read their forecast. My market has people on both sides of the trade. They bring information, needs, financing limits, fear, patience and rival rules. When they respond, the observable price changes. We are neither omnipotent authors nor passive spectators. We are participants whose actions enter a collective result.
This distinction affects what I ask a model to do. I want it to organise evidence, expose assumptions, estimate the consequences of a rule under varied historical conditions and identify where implementation could fail. I do not ask it for an oracle’s position outside the participants it describes. When its use might change a market, I include adoption and capacity among the hypotheses to examine. When my own footprint is negligible, I still attend to what other large or coordinated users might do. These are questions about participation, not a rejection of calculation.
This matters even when my own influence is tiny. The question “Did my order move the market?” is narrower than “Am I part of the system I am trying to understand?” The latter is true of every trader who must commit capital before the next configuration appears. A model can inform the commitment. It cannot move me to a vantage outside the market.
The eventual price path will show what happened after participants acted. A forecast is part of that action when its audience can trade on it. I use models to estimate risk and test rules, then take the next signal as one participant among others. The future is not hidden. It is being written by us, now, in the actions we take.
Arthur’s bar closes for the night and leaves behind a count of who attended. Episode 3 asks what such a record tells us about everything that failed to persist.
Previous: Episode 1, What Bohm Saw
Next: Episode 3, What Survives Is What Could Not Be Eliminated
Sources and notes
[1] W. Brian Arthur, Inductive Reasoning and Bounded Rationality: The El Farol Problem, American Economic Review 84, no. 2 (1994): 406–411.
[2] Nassim Nicholas Taleb, The Black Swan: The Impact of the Highly Improbable (New York: Random House, 2007), especially the discussion of retrospective predictability.
Richard Brennan writes on systematic trading, complex adaptive markets, and the philosophical foundations of trend following at atstradingsolutions.com. His books include The Fractals of Finance, Complex Adaptive Markets, Carved by Impossibility and The Aussie Turtles Trend Following Guide.
Want to explore why structure exists at all?
Carved by Impossibility: What Remains When Everything Else Is Eliminated
The book explores the architecture of constraint, emergence, and reality itself, and what it means for how we understand markets, life, and the universe.
Available now on Amazon in paperback, hardcover, and Kindle.
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.