What the market actually prices in

Every sharp market move seems to trace back to something nobody saw coming — a sudden war, a policy shock, some discontinuity with no precedent to price against. That’s the part of market behavior I keep coming back to: the future isn’t unknowable because we’re bad at analysis, it’s unknowable because the thing that actually moves prices is, by definition, the thing nobody could have modeled in advance.

Which is why I’m skeptical of any strategy built on watching how the market behaved last time and assuming it will behave that way again. If a pattern is visible enough to spot on a chart, or explainable in a five-minute video, it’s visible enough that someone else has already traded on it, and once enough people do, the pattern gets priced away before it’s useful ¹ ². History doesn’t repeat cleanly enough to trade on casually; it mostly just rhymes enough to make a convincing story after the fact.

But “you can’t trade on a hunch” is different from “there is no edge.” Real, persistent inefficiencies do exist — patterns and premia that survive being known about ³ ⁴, either because they only pay off if you can hold through long stretches of being wrong, or because exploiting them profitably takes more infrastructure than any single chart or backtest can capture. Finding and harvesting those is a fundamentally different exercise than pattern-matching a chart. It takes rigorous, systematic, data-driven research: testing an idea across decades of data and dozens of markets, controlling for how a signal decays once too many people find it, and sizing every bet so that being wrong on any one of them doesn’t matter much.

That’s the actual case for active management done well — not someone with a strong opinion about where the market goes next, but a process built to separate real, durable signal from noise at scale, and to keep doing that as the noise itself adapts. It’s also exactly why I’d trust that kind of process over a hot tip from a video or a confidently-written blog post — including, a little uncomfortably, this one. The distance between “I noticed a pattern” and “I found something real” is a lot of rigorous work, and most of us, myself included, aren’t equipped to do that work on the side.

There’s decent academic evidence behind that skepticism about “visible” patterns, and it’s more specific than just a hunch. McLean and Pontiff tested 97 variables that academic papers had shown to predict U.S. stock returns and checked how well each kept predicting. Portfolio returns were 26% lower out-of-sample and 58% lower after publication, which they read as investors learning about an anomaly from the research itself and trading it away ¹. Chordia, Subrahmanyam and Tong found that rising liquidity and trading activity made a handful of anomalies disappear ². And many patterns never were real. Hou, Xue and Zhang replicated 452 published anomalies and found that 65% failed a basic significance test, and 82% failed a stricter one ⁵.

What’s harder to arbitrage away is a premium that only pays off if you can stomach being wrong for a long stretch. Momentum is the standard example: well documented across more than eighty years of U.S. data and in dozens of other countries ³, and yet it hasn’t disappeared the way you’d expect a fully-known free lunch to. Part of the explanation may be that momentum strategies occasionally suffer sharp, persistent crashes, and they come in panic states, after market declines and in high volatility, just as the market rebounds ⁶. A separate NBER paper covering 140 years found the same two facts together: abnormal returns, and large crashes that were partly predictable. It argues that money managers competing for return-chasing capital have an incentive to stay invested even when the crash risk is known to be high ⁴. The premium survives, in other words, partly because collecting it means sitting through a crash you can see coming.

August 5, 2024 was a fresh reminder of the other half of the argument — the unmodelable part. Japan’s Nikkei 225 fell about 12.4% in a single session, its worst day since the 1987 Black Monday crash, as a rate hike from the Bank of Japan collided with a weak U.S. jobs report and unwound a huge, crowded yen carry trade almost overnight ⁷. Nobody had that exact combination on a five-minute chart the week before. That’s the asymmetry this post is really about: the patterns that are visible enough to spot get competed away, often faster than a decade ago; the moves that actually reprice the world tend to arrive from a direction the chart never showed you.

References

R. David McLean & Jeffrey Pontiff, “Does Academic Research Destroy Stock Return Predictability?” (The Journal of Finance 71(1), 2016). Out-of-sample returns on 97 published return-predicting variables are 26% lower, and 58% lower after publication.
Tarun Chordia, Avanidhar Subrahmanyam & Qing Tong, “Have Capital Market Anomalies Attenuated in the Recent Era of High Liquidity and Trading Activity?” (Journal of Accounting and Economics 58(1), 2014).
Tobias J. Moskowitz, “Explanations for the Momentum Premium” (AQR white paper, 2010). Summarizes risk-based and behavioral explanations for momentum and notes its presence in 83 years of U.S. data, 40 other countries and more than a dozen asset classes.
Benjamin Chabot, Eric Ghysels & Ravi Jagannathan, “Momentum Trading, Return Chasing, and Predictable Crashes” (NBER Working Paper 20660, 2014).
Kewei Hou, Chen Xue & Lu Zhang, “Replicating Anomalies” (Review of Financial Studies 33(5), 2020).
Kent Daniel & Tobias J. Moskowitz, “Momentum Crashes” (Journal of Financial Economics 122(2), 2016).
Reuters, “Japan’s Nikkei sees biggest rout since 1987 Black Monday” (August 5, 2024).