Factor investing

Recent interesting research from Cakici and Zaremba, highlights an often-overlooked aspect of machine learning for equity return prediction: the choice of prediction target. Rather than focusing on increasingly sophisticated model architectures or feature engineering, the authors show that how returns are represented during training has a much larger impact on predictive performance. In particular, models trained to predict stock ranks instead of raw return levels generate substantially stronger portfolio performance—roughly doubling both returns and Sharpe ratios in large-cap universes.

Investors often face a simple but important problem. They receive a fund equity curve, a strategy track record, or a portfolio performance series, but they do not know what is actually inside. The manager may provide only a broad description, while the realized return stream may in practice be driven by a mix of momentum, tactical allocation, defensive overlays, cross-asset rotation, or other systematic effects.

In this test, the unknown equity curve was treated as a blind case. The “correct” answer was not used during the analysis. The task was therefore not to confirm a known decomposition, but to test whether an API-based workflow can identify which known systematic strategies best explain the behavior of a black-box curve.

An autonomous research agent replicated nine published US-equity anomalies on clean, survivorship-free data. The question is not only what it found (out-of-sample decay is the rule, and on a faithful build none survive — the lone apparent survivor turned out to be a construction error the discipline caught) but whether you can trust an agent to find it, and the checks that decide the answer.

What if a meaningful part of a usual trading strategy’s performance has nothing to do with your signal—but simply when you rebalance? A recent paper written by Carlo Zarattini & Alberto Pagani highlights a largely underestimated risk in systematic investing: rebalance timing luck (RTL). For practitioners running rotation or factor strategies, this is not noise—it’s a structural source of dispersion. Using a concentrated U.S. equity momentum strategy, the authors show that identical portfolios differing only by rebalance day can diverge by as much as ~350 bps in annual returns, compounding into dramatically different terminal wealth outcomes.

Carbon that has not yet been emitted should not be used to predict stock returns. While this sounds obvious, prior research papers have done exactly that. This critical observation forms the basis for the Robeco Institutional Asset Management research team’s re-examination of the relationship between climate risk and asset pricing. Investors and academics alike have sought to understand how environmental factors influence stock returns, often assuming that higher emitters command a risk premium. However, the timing of data availability is crucial in quantitative strategy formation, and misalignments here can lead to spurious conclusions about the pricing of carbon emissions.

Quantitative strategy research is rarely about discovering a single “perfect” trading rule. In practice, robust portfolios emerge from a structured research process that filters ideas, evaluates evidence, and combines complementary strategies.

To make the process concrete, we use value-based equity strategies as our working example. However, the goal of the article is not to identify the ultimate value strategy, but to illustrate how a systematic research workflow can be used to build a diversified portfolio of strategies around any investment hypothesis.