Momentum (Investing)
Momentum is the empirical tendency of assets that have performed well (or poorly) over the recent past to continue performing in the same direction over the near term.
Momentum investing is built on a persistent market anomaly: securities with strong trailing returns tend to keep outperforming, and laggards tend to keep underperforming, over horizons of roughly three to twelve months. A classic cross-sectional momentum strategy ranks a universe of stocks by their past returns (often skipping the most recent month to avoid short-term reversal) and goes long the top performers while shorting the bottom. Time-series, or absolute, momentum instead judges each asset against its own past, holding it when its recent return is positive and exiting or shorting when negative. Momentum is one of the most documented factors in finance and contrasts directly with mean reversion, which bets on the opposite behavior over different horizons.
Momentum is well supported across asset classes and decades of data, but it carries distinctive risks. The most serious is the momentum crash: sharp reversals, often at market turning points after a sustained decline, can wipe out months of gains quickly because the strategy is by construction long the recent winners and short the recent losers exactly when leadership flips. Momentum also has meaningful turnover and transaction costs, and its returns are correlated with broad volatility regimes. Defining momentum precisely (lookback length, rebalancing frequency, and the skip period) materially changes the result, which makes disciplined, out-of-sample testing important.
For investors, momentum is valuable both as a standalone return source and as a diversifier, because it has historically been lightly or negatively correlated with value-style strategies. It is frequently combined with other factors so that the portfolio is not exposed to a single anomaly's drawdowns. Understanding momentum also helps interpret why trend-following can persist even in markets often described as efficient.
On hedgewing.ai, recent-return and trend signals are part of the 45 engineered features feeding the model ensemble, and the sequence-aware architectures, particularly the LSTM, GRU, and TCN, are well suited to detecting persistence in price paths. The Transformer's attention mechanism can weight which past periods matter most for continuation. Rather than committing to a fixed lookback rule, the models learn momentum's strength from data and attach calibrated confidence to each forecast, while walk-forward backtesting checks that any momentum edge survives in genuinely unseen periods, including the crash-prone reversals where naive momentum fails.
Related terms
Mean Reversion · Factor Analysis · Temporal Convolutional Network (TCN) · Calibrated Confidence
Back to the hedgewing.ai glossary · See AI stock forecasts