Momentum Is Real, But You're Probably Measuring It Wrong
Among all the factors that academic finance has documented over the past five decades, momentum is arguably the most uncomfortable. It shouldn't work. In an efficient market, past price performance should carry no information about future returns. And yet it does — persistently, across geographies, across asset classes, across time. The question isn't whether momentum is real. It's whether the version most retail investors use bears any resemblance to what the research actually describes.
The Gap Between the Factor and the Folklore
When most people talk about momentum investing, they mean buying things that have gone up recently. That intuition isn't entirely wrong, but it collapses a lot of nuance into a phrase. The academic momentum factor — the one documented by Jegadeesh and Titman in 1993 and replicated hundreds of times since — uses a specific formation window: roughly months two through twelve of prior returns, explicitly excluding the most recent month. That exclusion exists because the last 30 days of price movement tend to mean-revert, not continue. Leading with that short-term signal doesn't give you momentum; it gives you the opposite. It's a subtle distinction, but the return difference between measuring momentum correctly versus naively can be substantial, especially in volatile markets.
There's also the question of what's being measured. Raw price return over a window is one approach. Risk-adjusted momentum — normalizing by the volatility of those returns — is another, and generally a stronger predictor. A stock that gained 18%25 over ten months with low volatility is a very different signal than one that gained 18%25 while swinging 4%25 in either direction most weeks. The second case reflects noise as much as trend; the first reflects sustained directional pressure. Conflating them is one of the more common errors in DIY quantitative screens.
Momentum isn't about what went up. It's about what went up in a way that suggests the move isn't finished.
Where It Breaks Down — and Why That Matters
Momentum also has a known failure mode: it crashes. During sharp market reversals — think March 2020, or the 2009 recovery — high-momentum portfolios can unwind violently as crowded positions reverse simultaneously. This isn't a flaw in the factor so much as a feature of how it operates. Momentum works because behavioral biases cause markets to underreact to new information over medium-term horizons, leading to persistent trends. But when sentiment shifts hard, those trends don't just stop — they snap. A pure momentum strategy that ignores regime context is a strategy that will periodically deliver severe drawdowns at the worst possible moments.
This is precisely where multi-factor models earn their keep. Combining momentum with fundamental quality signals or mean-reversion indicators doesn't dilute the momentum signal — it conditions it. You're not abandoning the factor; you're filtering for the environments where it's most reliable. Platforms like AIStockAnalytics are built around exactly this kind of synthesis, scoring equities across technical, fundamental, and hybrid dimensions simultaneously rather than relying on any single factor in isolation. That's not a novel idea in institutional quant shops — it's been standard practice there for years. Making that same analytical architecture available to private investors is a meaningful shift, not a marginal one. Momentum, measured correctly and embedded in the right context, is a durable source of signal. Measured carelessly, it's a reliable way to buy the top.