What the Quant Approach Cannot See
The measured path is the residue of a market still being made
Previous: Episode 6, The Practitioner’s Discipline
Next: Episode 8, Process Is the Only Refuge
In the week of 6 August 2007, several quantitative equity funds lost heavily while the broad market offered no equally simple explanation for the damage. Portfolios described as market neutral appeared to hold different stocks, in different firms, under separate risk reports. The losses pointed toward a connection that the separate risk reports did not show.
Amir Khandani and Andrew Lo later assembled evidence consistent with overlapping positions, forced reductions and liquidity disappearing as managers tried to trade. [1] Their reconstruction does not establish one complete chain for every fund. It exposes the possibility the separate portfolio reports had missed: many apparently independent positions could become one crowded trade when capital had to leave.
I am a quantitative trader. I use measured prices, code and written rules. A model fitted to a past path can omit something about the participants who will make the next one, especially when several users of similar models must act at once. The 2007 losses put that omission in the open.
What the variables inherit
Suppose two futures markets have had low measured correlation. Combining them may improve a portfolio’s historical distribution. The arithmetic is sound. Yet the number records earlier decisions made under earlier financing, positioning and economic conditions. If both markets become common holdings in similar portfolios, a shared need to reduce exposure can make them move together. The correlation is then recalculated, but the trader needed to understand the possible link before its full history appeared in the estimate.
The dominant quant orientation can analyse what the variables are doing with great precision while missing what their relationships are doing. A variable has a history; its name can conceal that fact. “Liquidity” might refer to quoted depth when orders are small; it may behave differently when many owners seek an exit. “Momentum” may describe a return pattern whose costs change as more capital uses it. “Diversification” may be visible across ordinary months and absent when funding pressure is shared. A more sophisticated model can investigate those contingencies. A more precise estimate of the old relationship alone cannot settle the new one.
In Bohm’s terms, these measurements belong to the explicate order. They are indispensable traces. The implicate, as I use the term here, is the constraint structure within which those traces were produced. If the structure changes, the same variable may carry a different practical meaning. The quant orientation fails when it treats the trace as an enduring property of a market considered apart from its participants.
Tracks bend through the grass in the hero image. A map can record their route. Once travellers read it, they may choose a path because it is on the map. Similarly, a market model can affect the opportunity it describes when its users become numerous or large.
I see this mistake in my own research when a parameter is given a name and acquires an air of permanence. A correlation matrix is an estimate from returns sampled over a chosen interval. It can be a good summary of the recent portfolio. It is tempting to treat the matrix as a map of enduring economic separations. If the same investors own several markets or the same funding source supports their positions, the unobserved connection may matter most when those investors must act together. A stress exercise asks how such a connection could emerge in the realised returns before its estimate has caught up.
A common exit beneath separate portfolios
An exposure that looked diversified within one manager’s model could overlap with positions held elsewhere. One fund’s selling changed prices faced by others. Those prices could trigger more selling. The relationships among portfolios, financing and market depth mattered at least as much as the estimated relationship among isolated return series. The shared constraint was in the relationship among managers, holdings and liquidity. Positions that looked separate under ordinary trading became linked when their owners tried to reduce them together. The arithmetic had described an earlier configuration accurately; it had not made that configuration permanent.
The timing changes the interpretation. A manager selling into falling prices may lock in losses and end the week with less exposure. If prices rebound after the urgent selling has passed, the fund does not automatically recover what the mark-to-market chart suggests it might have earned by holding. The sequence of margin pressure, execution and later price movement matters to the account that actually traded. Khandani and Lo’s evidence is indirect, so I cannot assign one manager’s loss to one forced order. The mechanism is still a concrete hypothesis to examine in positions and transaction records, rather than an incantation that correlations always rise in trouble.
Robert Lucas raised a related concern for economic policy evaluation: estimated historical relationships may change when people respond to a new policy. [2] Andrew Lo’s adaptive markets framework places competition and adaptation explicitly within financial analysis. [3] Albert Kyle’s market microstructure work treats information, trading and liquidity as interacting parts of price formation. [4] Some of the best quantitative thinkers have spent their careers studying the weakness of static estimates. My critique extends that inquiry, rather than arguing that quantitative research cannot see interaction.
The lesson is sharper than “correlations go to one in a crisis”, an aphorism too sweeping to be reliable. The question is which shared constraints would cause positions that appeared separate to become one trade, how that would affect execution, and what observable changes might alert us. A stress scenario and a capacity study can help. They still depend on an explicit account of who might respond and why.
A language for movement
Bohm wondered what happens when language gives movement the appearance of a settled thing. His proposed rheomode was an experiment in thinking with verbs and relationships nearer the centre. [5] It is no trading model. It reminds me that “liquidity” is the conditional willingness of others to transact, and “correlation” is a measure of earlier co-movement, not a substance stored in two markets. We need those nouns, but we need to know what activity gave them meaning.
The elder’s flowering tree from Episode 6 provides a challenge to that habit. Tree, insect, water and fish are separable objects for description. The location of the fish becomes intelligible through their association in a particular season and place. The elder could use one visible sign because a much longer relational record gave it meaning. A price series can also be a sign whose meaning depends on who acted, how and under which constraints. The analogy directs my attention; it does not grant me the elder’s precision in a different domain.
Trend following is one way I orient action toward what price is doing rather than toward a forecast of stable equilibrium. Its entry and exit rules respond as a move develops. Yet it is a quantitative model and faces the same test. Crowded entries can change execution. Several positions can reverse together. Whipsaw can continue longer than my research suggested. A price-responsive rule can be poorly designed, overfitted or traded at a size the market cannot support. The orientation helps me ask better questions. It confers no immunity.
The test I apply to myself is uncomfortable. Has the profitability of my entries relied on trading at a scale I could not actually carry? Would several systems using similar breakouts run into the same liquidity at once? Did I select the lookback that happened to flatter the historical path? If those questions reveal weakness, saying that my system is “adaptive” adds nothing. It must face the same capacity, implementation and out-of-sample tests I demand of the orientation I criticise.
The map I need
I shift from treating the measured path as the terrain to studying the activity that made it. Without measurement I cannot know whether a relationship held, how often it failed or whether my rules did what I claim. I need the map. I also need to remember who made the tracks, what conditions permitted their passage and what the map’s use may change. A backtest is a reconstruction from an observed path, not an experiment in which every possible participant reaction was held fixed.
I can make a model more honest by stating its generation story alongside its estimate. What created the return pattern? Who can act on it? What would adoption do to price and costs at the proposed scale? What observations would show that the relation has changed? These are research questions with imperfect answers. They are stronger than assuming that an estimated coefficient is a permanent law.
The map is indispensable to me. I also investigate the people whose decisions made its lines and the constraints under which they may travel together next time. In August 2007, positions held at different firms met the same urgency to sell. Separate models described separate portfolios; the exit was shared.
The tracks through the grass may guide the next traveller. Episode 8 asks what a trader can commit to when no map can guarantee the next route.
Previous: Episode 6, The Practitioner’s Discipline
Next: Episode 8, Process Is the Only Refuge
Sources and notes
[1] Amir E. Khandani and Andrew W. Lo, What Happened to the Quants in August 2007?, MIT working paper (September 2007). Their account of the mechanisms is partly inferential.
[2] Robert E. Lucas Jr., Econometric Policy Evaluation: A Critique, Carnegie-Rochester Conference Series on Public Policy 1 (1976): 19–46.
[3] Andrew W. Lo, The Adaptive Markets Hypothesis: Market Efficiency from an Evolutionary Perspective, Journal of Portfolio Management 30, no. 5 (2004): 15–29.
[4] Albert S. Kyle, Continuous Auctions and Insider Trading, Econometrica 53, no. 6 (1985): 1315–1335.
[5] David Bohm, Wholeness and the Implicate Order (London: Routledge, 1980), chapter 2, “The Rheomode: An Experiment with Language and Thought”.
Richard Brennan writes on systematic trading, complex adaptive markets, and the philosophical foundations of trend following at atstradingsolutions.com. His published books include The Fractals of Finance, Complex Adaptive Markets, Carved by Impossibility, and The Aussie Turtles Trend Following Guide, co-authored with Adam Havryliv. His forthcoming book, The Trend Following Manifesto, co-authored with Niels Kaastrup-Larsen, is coming soon.
You do not need to predict the future. You need a process that keeps you present for the rare moves that matter.
The Trend Following Manifesto: Building Robust Portfolios for an Uncertain World
Coming soon to Amazon in paperback and hardcover. Join the book launch list to receive the confirmed launch date, details of a special first-week discount, and personal emails about the story behind the book.
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.