Momentum investing is the philosophy that securities which have performed relatively well (or poorly) over an intermediate horizon — typically the past 3 to 12 months — tend to continue performing in the same direction over the next several months. It buys recent winners and, in its long-short academic form, sells recent losers. Its core tension is that this directly contradicts the textbook claim that past prices contain no information about future returns (weak-form market efficiency), yet the effect is one of the most robust and widely replicated patterns in all of empirical finance. The catch: momentum's edge is bought with the risk of rare, violent "crashes," high turnover, and the discomfort of buying what already looks expensive.

How it's measured / formed

Momentum is defined on past returns over a lookback (formation) window, then a position is held over a holding window. The classic academic specification ranks all stocks by their cumulative return and goes long the top decile, short the bottom decile.

• Cross-sectional (relative) momentum — rank securities against each other; winners are simply the best relative performers. This is the Jegadeesh–Titman construction.
• Time-series (absolute) momentum / trend-following — judge each asset against its own past return: long if its trailing return is positive, short/flat if negative (Moskowitz, Ooi & Pedersen, 2012).
• 52-week-high momentum — proximity of price to its 52-week high; George & Hwang (2004) found this nearness predicts future returns and partly subsumes raw past-return momentum.

The standard academic factor is UMD (Up-minus-Down), also called WML (Winners-minus-Losers) or MOM — the formation period is conventionally the past 12 months skipping the most recent month (the "2–12" window), because the most recent month exhibits short-term reversal rather than continuation. This is the fourth factor in the Carhart (1997) four-factor model, added to Fama-French's market, size, and value factors.

How it's used in practice

• Quant equity factor sleeves. Institutions run UMD-style long-only or long-short books, usually with sector and beta neutralization, monthly or quarterly rebalancing, and liquidity screens.
• Momentum ETFs / smart-beta. Index products (e.g. MSCI USA Momentum-tracking funds) deliver a long-only, capacity-aware version to retail and advisors.
• Cross-asset trend-following / managed futures (CTAs). Time-series momentum is the engine of the trend-following industry, applied to equity indices, bonds, currencies, and commodities.
• Relative-strength rotation. Tactical allocators rotate into the strongest asset classes, sectors, or countries by trailing relative strength.
• Retail "leadership" stock-picking. O'Neil's CANSLIM and Minervini's SEPA are practitioner cousins — they explicitly buy high relative-strength leaders breaking to new highs.

A central design choice is the skip-month convention and managing turnover: shorter holding windows raise gross returns but also raise the trading bill.

Standing & evidence

Momentum is among the best-documented anomalies, but the honest picture is layered:

• Original finding. Jegadeesh & Titman (1993, Journal of Finance) showed buying winners / selling losers over 3–12-month windows earned roughly ~1% per month (their strongest cell, the 12-month formation / 3-month holding portfolio, averaged ~1.31%/month gross over 1965–1989). This is gross, in-sample, US equities.
• Pervasiveness. Asness, Moskowitz & Pedersen ("Value and Momentum Everywhere," 2013) found a consistent momentum premium across eight markets and asset classes with common factor structure. Geczy & Samonov extended US tests back to the 1800s and to multiple assets, finding momentum profits remained positive and significant out-of-sample — strong evidence it is not pure data-mining.
• The crash risk. Daniel & Moskowitz ("Momentum Crashes," 2016) document that momentum is prone to infrequent but severe drawdowns. They report that 14 of the 15 worst momentum months occurred when the prior two-year market return was negative and the contemporaneous market return was positive — i.e. in panicked, rebounding, high-volatility markets (notably 2009). Momentum returns are negatively skewed with fat tails; the premium is partly compensation for, or a symptom of, this tail.
• Why it exists (contested). The leading explanations are behavioral — investor underreaction to news (information diffuses slowly) and delayed overreaction from return-chasing (Hong & Stein; Daniel-Hirshleifer-Subrahmanyam). Risk-based stories struggle to explain its magnitude and crash structure. Fama & French (2008, "Dissecting Anomalies") — efficient-markets proponents — labeled momentum the "premier anomaly," notable because it is the effect their own framework most struggles to dismiss.
• Net-of-cost feasibility. Momentum is turnover-heavy. Korajczyk & Sadka (2004) estimated that after price-impact costs roughly $5 billion (in 1999 terms) could be deployed in well-designed momentum strategies before profits vanish; in small/illiquid names trading costs can swallow the gross edge entirely. Implementation, not existence, is the binding constraint.

Strengths & limitations

Works best: in trending, calm-to-moderate-volatility regimes; in liquid, large-cap universes where costs are manageable; over the 3–12-month horizon (the "sweet spot"); and as a complement to value, with which it is strongly negatively correlated — the two diversify each other well.

Fails: at sharp regime turns and market rebounds off panic lows, where crashes concentrate; at the very short horizon (1 month shows reversal, hence the skip-month) and the very long horizon (3–5 years shows reversal — De Bondt & Thaler); and once costs are charged in illiquid names.

The single most common misuse is conflating the disciplined, diversified, intermediate-horizon momentum factor with naïve performance chasing — piling into whatever is hottest right now with no skip-month, no risk control, and no exit rule. That is the short-term reversal zone and behaves oppositely. A related error is borrowing the academic factor's credibility for a single high-RSI chart: the cross-sectional momentum factor (Jegadeesh-Titman) is robust as a diversified portfolio premium; a single-name RSI or "it's going up" read is not the same thing and carries no comparable evidence.

System relevance

Within the Delvantic corpus, this node is the investment-philosophy anchor for relative-strength and trend concepts that recur in the Technical Analysis and Swing-Trading branches — cross-link to the TA definitions of relative strength, the 52-week high, and RSI (noting RSI is a single-name oscillator, not the cross-sectional momentum factor), and to leadership setups (CANSLIM / SEPA). For the Augustus trade-setup agent, the load-bearing caveats are: (1) momentum is a portfolio-level premium — single-name "it's trending" reads inherit none of the academic edge; (2) skip the most recent month / week of price action to avoid the reversal zone; and (3) momentum's worst losses cluster in high-volatility rebounds off declines, so any momentum-tilted signal should be gated by the regime engine — its edge degrades precisely when the market is panicking and rebounding.

Sources

• Jegadeesh & Titman (1993), "Returns to Buying Winners and Selling Losers," Journal of Finance 48(1):65–91 — onlinelibrary.wiley.com/doi/abs/10.1111/j.1540-6261.1993.tb04702.x
• Daniel & Moskowitz (2016), "Momentum Crashes," Journal of Financial Economics — nber.org/system/files/working_papers/w20439/w20439.pdf
• Asness, Moskowitz & Pedersen (2013), "Value and Momentum Everywhere," Journal of Finance — onlinelibrary.wiley.com/doi/abs/10.1111/jofi.12021
• Moskowitz, Ooi & Pedersen (2012), "Time Series Momentum" — w4.stern.nyu.edu/facdir/lpederse/papers/TimeSeriesMomentum.pdf
• George & Hwang (2004), "The 52-Week High and Momentum Investing" — referenced via alphaarchitect.com / bauer.uh.edu
• Korajczyk & Sadka (2004), "Are Momentum Profits Robust to Trading Costs?", Journal of Finance — kellogg.northwestern.edu/faculty/korajczy/htm/Korajczyk%20Sadka.jf2004.pdf
• Carhart (1997), four-factor model / WML — en.wikipedia.org/wiki/Carhart_four-factor_model
• Geczy & Samonov, "Two Centuries of Price Return Momentum" — ssrn.com/abstract=2292544
• AlphaArchitect, "Risk-Based Explanations for the Momentum Premium" — alphaarchitect.com/risk-based-explanations-momentum-premium/

Dispute flags: the underreaction-vs-overreaction (behavioral) vs risk-based explanation for momentum is genuinely unresolved; the cited Jegadeesh-Titman ~1%/month figure is gross, in-sample US equity data and overstates net-of-cost, post-publication returns.