Momentum vs. Mean Reversion: A Quantitative View

1. What is Momentum?

At its core, momentum is the empirical observation that assets that have performed well in the recent past tend to continue performing well, while assets that have performed poorly tend to continue performing poorly. It's often summarized by the adage, "winners keep winning, and losers keep losing."

From a quantitative trading perspective, this translates into a simple rule: buy past winners and (in a long/short portfolio) sell past losers.

To see this concept defined and illustrated, please watch the following segment from the video "Quantitative Momentum."

As you saw, the strategy involves ranking assets based on their performance over a specific lookback period (e.g., the past 12 months) and investing in the top performers. The academic work by Jegadeesh and Titman demonstrated that this simple strategy generated statistically significant excess returns, challenging the classical Efficient Market Hypothesis.

The persistence of this effect is striking. Let's look at the historical data.

The graph powerfully illustrates the momentum effect: a portfolio of high-momentum stocks substantially outperformed the market, while a portfolio of low-momentum stocks dramatically underperformed even risk-free assets. This provides the empirical motivation for building strategies around this phenomenon.

2. The Quantitative View: Momentum vs. Mean Reversion

Now, let's formalize these concepts. From a statistical standpoint, momentum and its counterpart, mean reversion, can be described using the concept of autocorrelation in asset returns.

• Momentum implies positive autocorrelation: high past returns are correlated with high future returns.
• Mean Reversion implies negative autocorrelation: high past returns are correlated with low future returns (and vice-versa), suggesting a "reversion to the mean."

The following reading provides a detailed breakdown of these definitions and their theoretical foundations.

To put it simply, momentum strategies bet on the continuation of a trend, while mean-reversion strategies bet on its reversal.

The video below offers a concise, practical contrast between the two approaches.

3. Time Horizons and Practical Nuances

An important insight from your reading is that momentum and mean reversion are not mutually exclusive; they often dominate at different time horizons.

• Short-Term (days to weeks): Mean reversion is often observed. This is sometimes called the "short-term reversal" effect. It can be caused by market-microstructure effects (e.g., liquidity providers pushing prices back to equilibrium) or overreactions to noise.
• Intermediate-Term (3-12 months): Momentum is the dominant effect. This is the timeframe most classic momentum strategies target.
• Long-Term (3-5 years): Mean reversion tends to reappear. Stocks that have been winners for several years may become overvalued and subsequently underperform.

The "Momentum and Mean Reversion Primer" you read earlier has an excellent section that details this. I recommend you review the section "Manifestation Across Different Timeframes" (pages 3-8) when you have time, as it provides a comprehensive overview of this crucial concept.

A Key Refinement: The "12-2" Rule

This time-horizon dependency leads to a key refinement in quantitative momentum strategies. Researchers found that excluding the most recent month's performance from the lookback period often improves results. This is because the short-term reversal effect can contaminate the intermediate-term momentum signal.

This leads to the common "12-2" or "12-1" convention: calculate momentum based on the returns from 12 months ago to 2 (or 1) months ago.

The following clip explains the rationale for this adjustment.

4. Implications for Strategy Design

The fundamental differences between momentum and mean reversion lead to trading strategies with starkly different performance characteristics.

• Momentum Strategies typically have a win rate below 50%. They make money by capturing rare but very large wins when a strong trend persists, which more than offset the frequent small losses from failed trends. This results in a positively skewed return distribution.
• Mean Reversion Strategies typically have a high win rate. They profit by making many small, consistent gains as prices oscillate around a mean. However, they are vulnerable to rare but very large losses when a price breaks out of its range and begins a new, strong trend. This results in a negatively skewed return distribution.

This clip and the following chart illustrate this critical difference.

Understanding this distinction is vital for risk management and for aligning a strategy with your own psychological tolerance for frequent small losses versus rare large ones.

Conclusion

In this lesson, we have established a foundational understanding of momentum and contrasted it with mean reversion.

Key Takeaways:

• Momentum is the tendency for past winners to continue winning. Quantitatively, it is defined by positive autocorrelation in returns and is often measured over an intermediate-term lookback period (e.g., 3-12 months).
• Mean Reversion is the tendency for prices to revert to an average value after an extreme move. It is defined by negative autocorrelation in returns.
• These phenomena are driven by different behavioral biases: momentum is largely attributed to investor underreaction to news, while mean reversion is linked to investor overreaction.
• The dominance of momentum or mean reversion depends on the time horizon. Short-term returns often revert, intermediate-term returns trend, and long-term returns revert again.
• Strategies built on these principles have very different risk profiles: momentum strategies aim for large, infrequent wins (positive skew), while mean-reversion strategies aim for small, frequent wins (negative skew).

This conceptual framework is the starting point for everything that follows. In our next lesson, we will get more practical and address the learning outcome: "Calculate logarithmic and arithmetic returns for crypto assets and explain the implications of their statistical properties." We will move from the 'what' to the 'how' by focusing on returns—the fundamental building block for calculating momentum.