There is a phenomenon in academic research known as HARKing — Hypothesizing After Results are Known. The investment world has its own version of this problem. It is pervasive, it is costly, and it is almost never discussed by name.
There is a phenomenon in academic research known as HARKing — Hypothesizing After Results are Known. It describes the practice of looking at data, finding a pattern, and then constructing a hypothesis to explain it, while presenting that hypothesis as if it had been formulated before the data was examined. The result is research that appears rigorous but is, at its foundation, a sophisticated form of storytelling dressed in the language of science.
The investment world has its own version of this problem. It is pervasive, it is costly, and it is almost never discussed by name.
What HARKing Looks Like in a Portfolio
The most common form of investment HARKing is the retrospective thesis revision. An investor builds a position in a company for one set of reasons — steady cash flows, a durable competitive moat, and an undervalued balance sheet. The stock then moves sharply for a completely unrelated reason: a supply chain disruption in a secondary business unit, a regulatory change in an adjacent market, or a short squeeze driven by retail momentum. The investor, rather than acknowledging that the gain was unrelated to their original thesis, quietly rewrites the story. By the time they discuss the position with a partner or write it up in a quarterly letter, the narrative has been reconstructed to suggest that the unexpected catalyst was anticipated all along.
This is not dishonesty in the conventional sense. It is a cognitive reflex, and it operates largely below the level of conscious awareness. The human mind is extraordinarily good at constructing coherent narratives from random events. It is far more comfortable with a story that has a logical arc than with the admission that a significant portion of investment returns are the product of chance. The problem is that this reflex, left unchecked, systematically corrupts the feedback loop that investment discipline depends on. If an investor cannot accurately distinguish between the gains that resulted from sound process and the gains that resulted from luck, they cannot improve. They will continue to attribute skill to outcomes that were driven by randomness, and they will be genuinely surprised when those outcomes fail to repeat.
The Quantitative Version: Overfitting
In systematic and quantitative investing, HARKing takes a more technical form known as overfitting. A quantitative analyst backtests a strategy against historical price data, adjusting parameters until the model produces attractive returns. The model is then presented with a rationale — a story about why these particular indicators, in this particular combination, capture a real and persistent market inefficiency.
The difficulty is that with enough parameters and enough historical data, it is almost always possible to find a combination that produces impressive backtested returns. The more parameters a model has relative to the number of observations in the dataset, the more likely it is that the apparent edge is simply a description of random noise in the historical data rather than a genuine structural feature of markets. When the model is deployed in live trading, the noise does not repeat, and the strategy fails.
This is the quantitative equivalent of the narrative fallacy: the analyst has constructed a compelling explanation for a pattern that does not actually exist in the forward-looking data. The backtested returns are real. The edge they appear to demonstrate is not.
The Fund Manager's Dilemma
The institutional version of HARKing is perhaps the most consequential, because it operates at the level of investor communication rather than internal decision-making. When a fund manager presents quarterly or annual results to limited partners, there is an enormous structural incentive to present the portfolio's performance as the product of deliberate, foresighted strategy — regardless of how much of that performance was actually driven by factors outside the manager's control or original thesis. A fund that underperformed its benchmark in its primary strategy but happened to benefit from a tail-risk hedge that was put on for unrelated reasons will often present the hedge as a deliberate defensive positioning decision. A fund that outperformed because a single position was acquired in a company that was subsequently taken private at a premium will often present the acquisition as the product of deep fundamental research into the company's strategic value to potential acquirers — even if the takeout was genuinely unpredictable. The result is that limited partners receive a systematically distorted picture of their manager's actual skill. Over time, this makes it nearly impossible to evaluate whether a fund's historical returns are the product of a repeatable process or a series of fortunate accidents.
The Alpha Crisis
The parallel to the replication crisis in academic science is instructive. In science, HARKing contributes to a situation in which a large proportion of published findings cannot be replicated by independent researchers. The findings are real in the sense that the data actually produced the reported results. But the results are artifacts of the specific dataset and the specific analytical choices made after the data was examined, not genuine discoveries about the structure of the world.
In investing, the equivalent is what might be called an alpha crisis: a large proportion of strategies that appear to generate excess returns in backtesting or in historical track records fail to produce those returns when deployed prospectively. The strategies are real in the sense that they did produce the historical returns. But those returns were artifacts of the specific historical period and the specific parameter choices made after the data was examined, not genuine structural edges.
The scale of this problem in asset management is difficult to overstate. A significant body of academic research suggests that the majority of actively managed funds fail to outperform their benchmarks over long periods after fees — and that a substantial portion of those that do outperform in any given period fail to sustain that outperformance in subsequent periods. HARKing is not the only explanation for this pattern, but it is a significant contributing factor.
How Discipline Prevents It
The antidote to HARKing in investing is the same as the antidote in science: pre-registered. In academic research, pre-registration means committing to a specific hypothesis, methodology, and analysis plan before collecting or examining the data. In investing, the equivalent is a set of practices that force explicit articulation of the investment thesis before a position is taken and that create an unalterable record of that thesis against which subsequent outcomes can be evaluated.
Investment policy statements serve this function at the portfolio level. An IPS that specifies explicit entry criteria, position sizing rules, exit conditions, and the specific factors that would constitute confirmation or disconfirmation of the investment thesis creates a framework that makes retrospective thesis revision visible and therefore harder to sustain unconsciously.
Pre-trade documentation serves the same function at the individual position level. Writing down, before execution, the precise hypothesis being tested, the expected timeframe for the thesis to play out, the specific risks that could invalidate it, and the conditions under which the position would be exited creates a paper trail that the investor must confront when reviewing performance. It is considerably harder to rewrite a thesis that is written down in a document with a timestamp.
Out-of-sample testing addresses the quantitative version of the problem. Holding back a portion of historical data that the model is never allowed to examine during development — and testing the final model only against that held-back data — provides a genuine test of whether the apparent edge is structural or artifactual. If the model performs well on the in-sample data but fails on the out-of-sample data, the edge is almost certainly the product of overfitting rather than a genuine market inefficiency.
The Stewardship Dimension
For the Walmer Family Fund, the discipline of pre-registration is not merely a technical matter of the investment process. It is a dimension of stewardship. The obligation of the steward is not to produce the appearance of skill but to exercise genuine, accountable judgment on behalf of those whose capital has been entrusted to their care. A steward who mistakes luck for skill, who rewrites their thesis after the fact, and who presents accidental gains as the product of foresight is not fulfilling that obligation — regardless of whether the returns happen to be attractive.
The biblical injunction to count the cost before undertaking a project — to think carefully and honestly about what is required and what is at stake before committing — is a form of pre-registration. It is the discipline of articulating, in advance, what you believe to be true and what you are prepared to stake on that belief. It is the opposite of HARKing.
Long-duration investing, by its nature, creates more opportunities for HARKing than short-term trading, because the gap between the original thesis and the eventual outcome is larger and the intervening events are more numerous. The discipline required to maintain an honest account of what was anticipated and what was not, over periods of years and decades, is considerable. It is also, for that reason, one of the most important disciplines in the long-duration investor's practice.