Decomposing Momentum Strategies Across Timeframes

1. Momentum Is Not One Thing

Momentum — the tendency for assets that have recently outperformed to continue outperforming — is among the most robust anomalies in empirical finance. Jegadeesh and Titman (1993) documented it in US equities; Asness, Moskowitz, and Pedersen (2013) extended the evidence across asset classes and geographies. It is one of the few anomalies that survived the replication crisis in finance largely intact.

Yet the academic literature's focus on "momentum" as a single phenomenon obscures a more nuanced reality. A 1-week momentum signal and a 12-month momentum signal are capturing different market dynamics. The former is likely driven by order flow persistence and short-term under-reaction to information. The latter is driven by gradual diffusion of fundamental information, investor herding, and the slow adjustment of institutional portfolios to new regimes.

For systematic traders, this distinction matters because it determines how to construct, combine, and hedge momentum exposures. If short-term and medium-term momentum were simply the same phenomenon at different speeds, there would be no diversification benefit from combining them. If they are distinct — driven by different mechanisms, with different risk profiles and different failure modes — then combining them is a genuine free lunch.

2. Methodology

We construct momentum signals at seven look-back/hold horizons: 1 day, 1 week, 1 month, 3 months, 6 months, 9 months, and 12 months. For each horizon, the signal is the total return over the look-back period, and positions are held for a period equal to the look-back period. We skip the most recent 1 day for all signals except the 1-day signal itself, following the standard practice to avoid contamination from bid-ask bounce.

The asset universe consists of 24 liquid futures markets: 8 equity indices (S&P 500, Nasdaq 100, EuroStoxx 50, FTSE 100, Nikkei 225, Hang Seng, ASX 200, DAX), 8 FX pairs (EUR/USD, GBP/USD, USD/JPY, AUD/USD, USD/CAD, USD/CHF, NZD/USD, EUR/GBP), and 8 commodities (WTI Crude, Brent, Gold, Silver, Copper, Natural Gas, Corn, Soybeans). The evaluation period is January 2005 to December 2024.

At each rebalancing date, We rank all 24 assets by their momentum signal, go long the top quintile and short the bottom quintile, and hold until the next rebalancing date. Position sizes are volatility-adjusted: each asset's notional exposure is inversely proportional to its trailing 60-day realised volatility, targeting a constant ex-ante contribution to portfolio risk.

3. Performance by Horizon

Table 1: Performance of momentum strategies by look-back horizon, January 2005 – December 2024. All returns are gross of transaction costs.

The results show a hump-shaped pattern: momentum profitability peaks at the 1–3 month horizon and declines at both shorter and longer horizons. This is consistent with the existing literature. More interesting is the return skewness column: short-term momentum (1 day, 1 week) shows modest negative skewness, consistent with the reversal risk inherent in very short-term trend-following. Medium-term momentum (1–3 months) is approximately symmetric. Longer horizons show increasingly negative skewness, reflecting the crash risk that is well-documented in long-horizon momentum strategies — the risk of sudden, violent reversals when a long-running trend ends abruptly.

4. Correlation Structure

The key question for portfolio construction is whether momentum signals at different horizons are correlated. If they are, combining them provides little diversification benefit. If they are not, a multi-horizon momentum portfolio can achieve a substantially better Sharpe ratio than any single-horizon strategy.

Table 2: Pairwise correlations of daily returns across momentum horizons.

The correlation matrix reveals two distinct clusters. Short-term signals (1 day, 1 week) are moderately correlated with each other (0.32) but nearly uncorrelated with medium-term signals (3–12 months). Within the medium-term cluster, correlations are higher (0.48–0.72), reflecting the substantial overlap in the information used by these signals. The 1-month signal sits at the boundary, showing meaningful correlation with both clusters.

This structure implies that the primary diversification benefit comes from combining short-term (1 week) and medium-term (3 month) momentum rather than from spreading across many medium-term horizons.

5. A Combined Strategy

We construct a combined momentum portfolio by allocating 40% to the 1-week signal and 60% to the 3-month signal. The allocation is not optimised; it reflects a qualitative judgment that the higher Sharpe 3-month signal should receive greater weight, tempered by the diversification benefit of the weekly signal.

Table 3: Performance of individual and combined momentum strategies. The combined portfolio achieves a materially higher Sharpe ratio and Calmar ratio than either component.

The combined portfolio achieves a Sharpe ratio of 0.82, compared to 0.63 for the best single-horizon strategy. More importantly, the maximum drawdown drops from 24.8% (3-month only) to 14.1%. This improvement comes entirely from diversification — the low correlation between short-term and medium-term momentum means that the worst periods for one signal often coincide with reasonable performance from the other.

6. Asset Class Differences

The momentum premium is not uniform across asset classes. Commodities show the strongest momentum at all horizons, consistent with the term structure and inventory-based theories of commodity momentum (Gorton, Hayashi, and Rouwenhorst, 2013). FX momentum is strongest at short horizons (1 week to 1 month), likely reflecting order flow persistence and gradual adjustment to macroeconomic news. Equity index momentum peaks at 3–6 months but carries the highest crash risk, reflecting the concentration of equity momentum reversals during market stress events.

For practitioners building multi-asset momentum portfolios, this suggests that the optimal horizon mix should vary by asset class. A commodity-focused portfolio might emphasise 3-month momentum, while an FX-focused portfolio would benefit more from shorter look-back periods.

7. Transaction Costs and Capacity

Short-term momentum generates substantially more turnover than medium-term momentum. The 1-week signal has an average annual turnover of approximately 2,600% (each position turns over roughly 50 times per year), compared to approximately 400% for the 3-month signal. At a transaction cost of 2 basis points per round trip in liquid futures, the 1-week strategy's cost drag is approximately 2.6% annually, versus 0.4% for the 3-month strategy.

After costs, the 1-week strategy's Sharpe ratio falls from 0.53 to approximately 0.24, which is marginally attractive. The 3-month strategy's post-cost Sharpe remains at 0.59. The combined portfolio's post-cost Sharpe is approximately 0.68 — still substantially above either component.

Capacity constraints are asymmetric by horizon. Short-term signals in liquid futures markets can accommodate tens of millions of dollars before market impact degrades performance. Medium-term signals, which trade less frequently and in the same liquid markets, can accommodate substantially more — likely several hundred million.

8. Conclusion

Momentum is not a single phenomenon but a family of related effects operating at different timescales and driven by different mechanisms. Short-term momentum (1–4 weeks) and medium-term momentum (1–6 months) are largely uncorrelated, creating a genuine diversification opportunity. A combined strategy that allocates across both horizons achieves a Sharpe ratio 30% higher than the best single-horizon strategy, with a 43% reduction in maximum drawdown. For systematic traders, treating momentum as a multi-scale phenomenon and constructing portfolios accordingly is one of the most accessible improvements available.