Standard Momentum Strategy: Mathematical Formulation & Implementation

Academic Background

In 1993, Narasimhan Jegadeesh and Sheridan Titman published their landmark paper "Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency", establishing that stocks with top past 3-to-12-month returns continued to outperform bottom-performing stocks over subsequent 3-to-12-month periods. In 1997, Mark Carhart formalized momentum as the fourth factor in institutional asset pricing models alongside market beta, size, and value.

Behavioral Hypothesis

Momentum persists due to documented cognitive and institutional behaviors:

1. Underreaction to New Information: Investors and analysts adjust earnings estimates incrementally, causing prices to drift toward new fundamental levels gradually over several months.
2. Disposition Effect: Investors sell winners prematurely while holding losers, slowing down the price adjustment process.
3. Institutional Herding: Mutual funds and asset managers systematically allocate capital toward leading stocks during quarterly performance window-dressing.

Strengths & Limitations

• Strengths: Strong multi-decade empirical track record across global equity markets; excellent participation in multi-month bull market cycles.
• Limitations: Vulnerable to sharp, non-linear momentum crashes during sudden market inflection points; higher turnover costs relative to passive indexing.

The Standard Momentum strategy applies rigorous return quantification across the equity universe:

1. Cumulative Return Calculation

For asset $i$ at time $t$ over lookback window $N$ (default $N = 252$ trading days, approximately 12 calendar months):

$$R_i(t, N) = \frac{P_{i,t}}{P_{i, t-N}} - 1.0 = \prod_{k=0}^{N-1} (1 + r_{i, t-k}) - 1.0$$

Where $P_{i,t}$ is the dividend-adjusted closing price.

2. Cross-Sectional Quantile Ranking

Let $\mathcal{U}_t = \{S_1, S_2, \dots, S_M\}$ represent the universe of $M$ liquid equities passing quality gates at time $t$. Each stock's cumulative return $R_i(t, N)$ is ranked:

$$\text{Rank}_i(t) = \text{ordinal rank of } R_i(t, N) \text{ within } \mathcal{U}_t \in [1, M]$$

The cross-sectional quantile thresholds for portfolio construction are defined as:

$$q_{\text{long}}(t) = \text{Quantile}(\mathcal{U}_t, \text{long\_percentile}), \quad q_{\text{short}}(t) = \text{Quantile}(\mathcal{U}_t, \text{short\_percentile})$$

The normalized cross-sectional percentile score $q_i(t) \in [0.0, 1.0]$ is computed as:

$$q_i(t) = \frac{\text{Rank}_i(t) - 1}{M - 1}$$

3. Volatility-Adjusted Momentum Score

To prevent low-quality, high-beta micro-caps from dominating rankings, returns are scaled by annualized historical realized volatility $\sigma_{i, \text{ann}}(t)$:

$$\sigma_{i, \text{ann}}(t) = \sqrt{252} \cdot \sqrt{\frac{1}{N_{\text{vol}}-1} \sum_{k=0}^{N_{\text{vol}}-1} (r_{i, t-k} - \bar{r}_i)^2}$$

$$\text{Score}_{\text{vol}, i}(t) = \frac{R_i(t, N)}{\sigma_{i, \text{ann}}(t)}$$

The strategy assigns signals based on cross-sectional percentile boundaries:

graph TD
    A[Compute 252-day Cumulative Returns across Universe] --> B[Rank Equities into Percentiles: q in 0 to 1]
    B --> C{Percentile Threshold Check}
    C -->|q >= 0.80 Top Quintile| D[LONG Signal: +1.0]
    C -->|q <= 0.20 Bottom Quintile| E[SHORT Signal: -1.0]
    C -->|0.20 < q < 0.80| F[NEUTRAL Signal: 0.0]
    D --> G[Apply Intraday 15m EMA Trend Confirmation]
    E --> G
    G --> H[Final Model Candidate Output]

Execution Protocol

• Long Directive (+1.0): Assigned when $q_i(t) \ge \text{long\_percentile}$ (default $0.80$, representing the top quintile of performers).
• Short Directive (-1.0): Assigned when $q_i(t) \le \text{short\_percentile}$ (default $0.20$, representing the bottom quintile of performers).
• Neutral (0.0): Assigned to intermediate percentiles where directional edge is statistically negligible.
• Rebalancing Cadence: To control transaction costs and portfolio churn, rankings are re-evaluated on periodic rebalancing schedules (monthly or bi-weekly).

• Momentum Crash Risk: When markets rebound violently following deep bear markets, historical losers frequently rally the hardest while past winners lag, causing sharp drawdowns for pure long/short momentum books.
• Tax & Turnover Management: High turnover can generate short-term capital gains for taxable accounts. Managing rebalance frequencies helps balance momentum responsiveness against execution drag.
• Sector Concentration: Pure cross-sectional sorting can lead to high sector concentration (e.g. 60%+ in technology during tech bull runs). Sector-neutralization constraints can be introduced to enforce balanced diversification.