SystemTrader
ETFsIntermediateTrend Following

Multi-Asset Trend Following

A diversified trend-following strategy that allocates across multiple asset class ETFs using momentum signals and volatility-adjusted weighting. Captures trends across equities, bonds, commodities, and real estate while controlling portfolio risk.

Universe
Multi-Asset ETFs
Formation
6-12 Months
MA Filter
100-200 Days
Rebalance
Monthly

Overview

The multi-asset trend following strategy exploits the diversification power of ETFs combined with systematic momentum signals. By taking positions across different asset classes—equities, bonds, commodities, real estate—the strategy captures trends while benefiting from low cross-asset correlations.

Unlike sector rotation which picks winners from a single asset class, this strategy weights all ETFs with positive momentum, using sophisticated weighting schemes that account for volatility. The key insight: simple momentum weights overweight volatile assets, so volatility-adjustment improves risk-adjusted returns.

The strategy computes cumulative returns over 6-12 months, filters for positive momentum (optionally using a moving average filter), and then assigns weights proportional to momentum divided by volatility. This approach, studied by Faber (2007) and refined by Du Plessis & Hallerbach (2015), creates a well-diversified trend portfolio that adapts to market conditions.

Key Insight

Multi-Asset Diversification
Spreads risk across asset classes
Volatility-Adjusted
Prevents overweighting risky assets
Trend Capture
Only holds assets in uptrends

Weighting Formula

wi = γ · Rcum / σ²

Higher momentum + lower volatility = more weight

Legend

Positive momentum
✗Negative momentum (excluded)

Key Insight

US Bonds (AGG) gets the highest weight despite modest momentum—its low 5% volatility results in excellent risk-adjusted allocation.

Research

Multi-asset trend following combines insights from tactical asset allocation research with momentum and volatility-weighting literature. The strategy has strong academic foundations spanning several decades of empirical finance.

The Mathematics

In Plain English

The math behind this strategy is straightforward. Here's what you're actually doing:

  1. 1
    Calculate cumulative returns for each ETF over the formation period (6-12 months)
  2. 2
    Filter for positive momentum: only keep ETFs with R_i^cum > 0
  3. 3
    Optional: Apply moving average filter: keep only ETFs where Price > MA(100-200 days)
  4. 4
    Calculate historical volatility σ_i for each remaining ETF
  5. 5
    Assign weights using momentum and volatility: w_i = R_i^cum / σ_i²
  6. 6
    Normalize weights so they sum to 1 (γ ensures Σw_i = 1)
  7. 7
    Optional: Impose weight caps (e.g., w_i ≤ 25%) to prevent over-concentration

That's it. The formulas below just express this process precisely.

Technical Formulas

1
Cumulative Return (Momentum Signal)

Formula
R_i^{cum} = \frac{P_i(t)}{P_i(t-T)} - 1

Where P_i(t) is the price of ETF i at time t, and T is the formation period (typically 6-12 months). Only ETFs with positive R_i^cum are included.

2
Moving Average Filter

Formula
\text{Include ETF } i \text{ if } P_i > MA_i(T')

Optional filter where T' is typically 100-200 days. Only include ETFs trading above their long-term moving average to avoid catching falling knives.

3
Momentum-Proportional Weights

Formula
w_i = \gamma_1 \cdot R_i^{cum}

Simplest weighting: proportional to cumulative returns. Issue: overweights volatile ETFs since high-volatility assets tend to have larger cumulative returns.

4
Volatility-Adjusted Weights

Formula
w_i = \gamma_2 \cdot \frac{R_i^{cum}}{\sigma_i}

Dividing by volatility σ_i helps control overweighting of volatile assets. Improves Sharpe ratio by reducing exposure to high-volatility assets.

5
Sharpe-Optimal Weights

Formula
w_i = \gamma_3 \cdot \frac{R_i^{cum}}{\sigma_i^2}

Optimizes portfolio Sharpe ratio under diagonal covariance (ignoring correlations). Dividing by σ_i² provides stronger volatility adjustment. The γ constants ensure Σw_i = 1.

Normalization Constant γNote

The γ values (γ₁, γ₂, γ₃) are computed to ensure weights sum to 1: γ = 1 / Σ(numerator). For example, γ₃ = 1 / Σ(R_i^cum / σ_i²) where the sum is over all ETFs passing the filters.

Why Volatility-Adjust?Note

Simple momentum weights (Eq. 371) overweight volatile ETFs because expected returns are proportional to R_i^cum. Volatility adjustment (Eq. 372-373) mitigates this by penalizing high-volatility assets, improving risk-adjusted returns.

Weight CapsNote

Imposing bounds w_i ≤ w_i^max (e.g., 25%) further mitigates overweighting and ensures diversification. This is especially useful when one asset has exceptionally strong momentum.

Strategy Rules

Universe Selection

  1. Include diverse asset class ETFs: US stocks, international stocks, bonds, commodities, real estate
  2. Example universe: SPY, EFA, EEM, AGG, TLT, GLD, DBC, VNQ, TIP
  3. Ensure all ETFs have sufficient history (>5 years) and liquidity
  4. Avoid leveraged, inverse, or sector-specific ETFs
  5. Consider currency-hedged versions for international exposure

Momentum & Filter Rules

  1. 1Calculate 6-12 month cumulative returns for each ETF
  2. 2Keep only ETFs with positive cumulative returns (R_i^cum > 0)
  3. 3Optional: Apply 200-day MA filter (Price > 200-day MA)
  4. 4Calculate trailing volatility (e.g., 60-day rolling standard deviation)
  5. 5Assign weights to all remaining ETFs (no rank-based selection)

Weighting Rules

  1. Use volatility-adjusted weights: w_i = R_i^cum / σ_i or w_i = R_i^cum / σ_i²
  2. Normalize weights so Σw_i = 1
  3. Impose maximum weight cap (e.g., 25% per ETF)
  4. If all ETFs fail filters, hold cash or short-term bonds
  5. Rebalance monthly on a fixed schedule

Risk Management

  1. 1Volatility-adjustment inherently reduces exposure to risky assets
  2. 2Weight caps prevent over-concentration in any single asset
  3. 3Moving average filter provides crash protection
  4. 4If no ETFs pass filters, move fully to cash/bonds
  5. 5Monitor correlation regime changes across asset classes

Implementation Guide

Implementing multi-asset trend following requires calculating momentum and volatility for a diversified ETF universe, then applying the weighting formula. The key is consistent monthly execution.

1

Build Your ETF Universe

Select 8-15 ETFs covering major asset classes. Aim for diversification across equities (US, international, emerging), fixed income (government, corporate), and alternatives (commodities, real estate).

Tips
  • Core US: SPY (S&P 500), QQQ (Nasdaq)
  • International: EFA (Developed), EEM (Emerging)
  • Bonds: AGG (Aggregate), TLT (Long-term Treasury), TIP (TIPS)
  • Alternatives: GLD (Gold), DBC (Commodities), VNQ (Real Estate)
  • Consider: HYG (High Yield), LQD (Corporate Bonds)
2

Calculate Momentum Signals

For each ETF, calculate the cumulative return over your chosen formation period (6-12 months). Use adjusted close prices to account for dividends and splits.

Tips
  • Formula: R_i^cum = (Price_today / Price_12mo_ago) - 1
  • Skip the most recent month if using 12-1 momentum
  • Calculate as decimal (e.g., 0.15 for 15% return)
  • Filter out ETFs with negative momentum
3

Apply Optional Filters

For additional crash protection, apply a moving average filter. Only include ETFs trading above their long-term moving average.

Tips
  • Calculate 200-day simple moving average for each ETF
  • Filter: Include ETF if Current Price > 200-day MA
  • This removes assets in downtrends from the portfolio
  • If all ETFs fail the filter, hold cash or short-term bonds
4

Calculate Volatility

For each ETF passing the filters, calculate historical volatility. This will be used to adjust the momentum weights.

Tips
  • Use 60-day rolling standard deviation of daily returns
  • Annualize by multiplying by √252
  • Alternative: Use 20-day or 90-day windows
  • Store both momentum (R_i^cum) and volatility (σ_i) for each ETF
5

Compute Weights

Apply the volatility-adjusted weighting formula. Divide momentum by volatility (or volatility squared), then normalize to sum to 1.

Tips
  • Calculate raw weight: R_i^cum / σ_i² for each ETF
  • Sum all raw weights: Total = Σ(R_i^cum / σ_i²)
  • Normalize: w_i = (R_i^cum / σ_i²) / Total
  • Apply weight caps: if w_i > 0.25, set to 0.25 and redistribute

After capping, renormalize weights to ensure they sum to 1.

6

Execute and Rebalance

Execute trades to match target weights. Rebalance monthly by repeating the entire process with updated data.

Tips
  • Rebalance on first trading day of each month
  • Use limit orders to minimize market impact
  • Track turnover - expect moderate changes monthly
  • Document weights and rationale for performance review

Broker Requirements

This strategy can be implemented at any broker offering the selected ETFs. Commission-free ETF trading is widely available. No special account type required for the long-only version.

Helpful Tools & Resources

Data & Analysis
Yahoo Finance, Portfolio Visualizer, Python/Pandas
Backtesting
Portfolio Visualizer, Quantconnect, Zipline
Execution
Interactive Brokers, Schwab, Fidelity

Strategy Variations

Explore different ways to implement this strategy, each with its own trade-offs and benefits.

Equal-Weight Momentum

Instead of volatility-adjusted weights, equal-weight all ETFs that pass the momentum and MA filters. Simpler to implement but may overweight volatile assets.

Good baseline to compare against volatility-adjusted versions.

Risk Parity Overlay

Apply risk parity weighting (inverse volatility) first, then tilt toward positive momentum assets. Combines risk parity diversification with momentum signals.

Provides smoother returns with trend-following tilt.

Dual Momentum Filter

Add absolute momentum: only invest when the broad market (SPY) is above its 200-day MA. Otherwise hold bonds or cash regardless of individual ETF signals.

Significantly reduces drawdowns in bear markets.

Correlation-Adjusted Weights

Instead of assuming diagonal covariance, use the full covariance matrix to compute optimal weights. Accounts for correlations between asset classes.

More complex but theoretically optimal.

Country/Sector Extension

Expand the universe to include country ETFs or sector ETFs alongside asset class ETFs. Increases the opportunity set for trend capture.

Higher turnover but more diversified trends.

Consider combining multiple variations or testing them against your specific investment goals and risk tolerance.

Risks & Limitations

High(2)
Medium(2)
Low(2)
Trend ReversalsHigh

Trend-following strategies suffer during sharp market reversals when momentum flips quickly. The strategy may hold long positions as markets turn down, experiencing drawdowns before the filters trigger.

Impact:
Correlation Spikes in CrisisHigh

During market crises, correlations across asset classes spike toward 1. The diversification benefit assumed by the strategy breaks down precisely when needed most.

Impact:
Volatility Estimation ErrorMedium

Volatility-adjusted weights depend on accurate volatility estimates. Historical volatility may not predict future volatility, especially during regime changes.

Impact:
Filter WhipsawMedium

In choppy, range-bound markets, ETFs may frequently cross above and below their moving averages, causing frequent entry/exit signals that generate losses and transaction costs.

Impact:
Cash DragLow

When all ETFs fail the momentum/MA filters, the strategy moves to cash. Holding cash during false signals means missing market gains when trends resume.

Impact:
Model ComplexityLow

The volatility-adjusted weighting adds complexity. Errors in calculation, especially the normalization step, can lead to unintended portfolio exposures.

Impact:
Understanding these risks is essential for proper position sizing and portfolio construction. Consider combining with other strategies to mitigate individual risk factors.

References

  • Faber, M.T. (2007). A Quantitative Approach to Tactical Asset Allocation. Journal of Wealth Management, Spring 2007 [Link]
  • Du Plessis, J. & Hallerbach, W.G. (2015). Volatility Weighting Applied to Momentum Strategies. SSRN Working Paper [Link]
  • Bekkers, N., Doeswijk, R.Q. & Lam, T.W. (2009). Strategic Asset Allocation: Determining the Optimal Portfolio with Ten Asset Classes. Journal of Wealth Management, 12(3), 61-77 [Link]
  • Doeswijk, R.Q., Lam, T.W. & Swinkels, L. (2014). The Global Multi-Asset Market Portfolio 1959-2012. Financial Analysts Journal, 70(2), 26-41 [Link]

Multi-asset trend following involves risk of loss. Diversification does not guarantee profits or protect against losses. Past performance of tactical strategies does not guarantee future results. This is educational content, not investment advice.

Ready to explore more ETF strategies?

Discover additional tactical ETF strategies including sector rotation, factor timing, and mean-reversion approaches.

Browse ETF Strategies