Risk Variation in Trend-Following Systems: Why Performance Differences Occur Despite Identical Signals

motivation

Why do trend-following managers with similar approaches achieve such different results? The answer lies not only in signal generation, but also in risk management design .

This article systematically analyzes how different risk allocation rules affect performance – even with identical trend signals.

The year 2019

2019 was a perfect textbook year for this question.

While fixed income delivered spectacular trends, other asset classes (AC) showed significantly weaker or no trends:

• Fixed Income: Strong, consistent downward trends in yields

Equities: Volatile and trendless with large drawdowns

Commodities: Mixed signals without a clear direction

• FX: Predominantly moving sideways

The result: The Barclays BTOP50 Index showed a return range of -5% to +12% between the best and worst-performing managers. This discrepancy cannot be explained by differing signals – all managers had access to the same markets and trends.

The five critical design decisions

A study by AlphaSimplex identifies five fundamental design questions in trend following:

1. Signal generation

Absolute vs. Cross-Sectional Momentum

• Absolute: Time series of each market independently (classical approach)

• Cross-sectional: Relative strength between markets

Most diversified CTA managers use absolute momentum signals.

2. Cross-Sectional Risk Allocation

EAR vs. TVR (The decisive choice)

Equal Asset Class Risk (EAR):

Fixed risk budgets per asset class (e.g. 25% for equities, FI, commodities, FX)

Time-Varying Risk (TVR):

Dynamic allocation based on trend strength (signal conviction)

2019 Impact: TVR systems were able to benefit massively from the fixed income trend, while EAR systems were forced to shift capital into weaker asset classes.

3. Portfolio Level Risk Targeting

CRT vs. SRT

Constant Risk Targeting (CRT):

Fixed portfolio volatility (e.g., 10% per annum), independent of signal strength

Signal-Based Risk Targeting (SRT):

Higher risk with strong trend signals, lower risk with weak ones.

Philosophy: CRT stabilizes the risk profile. SRT increases "conviction"—taking more risk during strong trends.

4. Rebalancing

Continuous vs. Periodic

• Continuous: Daily or during significant market movements

• Periodic: Fixed intervals (weekly, monthly)

More frequent rebalancing means higher transaction costs, but better trend participation.

5. Timing Approach

End-of-day vs. Intraday

A technical issue of signal implementation, less relevant for strategic performance differences.

The four risk rescaling combinations

In its study, AlphaSimplex simulates four core strategies for 2019 with the following result:

strategy	Cross-sectional	Portfolio Level	2019 Performance
EAR-CRT	Fixed 25% AC Risk	Constant 10% Volatility	~+6%
EAR-SRT	Fixed 25% AC Risk	Signal-Based	~+4%
TVR-CRT	Time-Varying	Constant 10% Volatility	~+12%
TVR-SRT	Time-Varying	Signal-Based	~+8%

The results in detail:

1. TVR-CRT: The clear winner of 2019 (+12%)

• Maximum participation in the fixed income trend

• A constant risk budget (10% volatility) is optimally focused on trending markets.

• No forced allocation to weak markets

2. EAR-SRT: The loser (+4%)

Fixed 25% risk budgets per asset class force diversification into trendless markets.

Signal-based risk targeting further reduces overall exposure.

• Double restriction: neither flexible AC allocation nor constant risk-taking

3. EAR-CRT & TVR-SRT: Midfield

• EAR-CRT: Fixed AC allocation, but constant risk budget stabilizes performance (~+6%)

• TVR-SRT: Flexible AC allocation, but reduced exposure due to SRT (~+8%)

The practical implications

What does this mean for your trading system?

1. TVR is not always better

2019 was a "TVR-friendly" year with a clear trend concentration in one asset class. In years with broad diversification across all asset classes, EAR can deliver more stable results.

2. CRT vs. SRT is a philosophical question.

• CRT: Preference for a consistent risk profile, regardless of the market environment

SRT: Preference for "adaptive conviction" – more risk in strong trend phases

3. Asset class caps as a middle ground

AlphaSimplex recommends TVR with asset class caps (e.g., maximum 40% portfolio risk per asset class):

• Allows flexibility for strong trends

• Prevents extreme concentration

• Balances diversification and opportunity

An example implementation would be as follows:

Target Portfolio Volatility: 10%

Max AC Risk: 40% × 10% = 4% per asset class

If fixed income trends generated an 8% risk contribution:

• → Cap at 4%

• → Distribute the remaining budget (6%) to other ACs

Historical comparison: 2008 vs. 2019

The paper compares 2019 with 2008 – a year in which both approaches worked well:

2008:

Strong, synchronous trends across all asset classes

· Equities crash, fixed income rallies, commodities collapse, USD rises

EAR and TVR deliver similar performance (~+20%)

2019:

• Trend concentration in fixed income

Other asset classes are trendless or contrary to current trends

TVR is massively superior (+12% vs. +6% for EAR-CRT)

The lesson: In crises with broad, synchronous trends, risk allocation is less relevant. In "single-trend years," the design determines success or failure.

Practical implementation:

TVR with Asset Class Caps

Here's how to implement a TVR approach using caps:

Step 1: Calculate Raw Positions

w_raw = Signal / Volatility (per market)

Cap e.g. at ±2.0 (Max position per market)

Step 2: Scale to target volatility

Portfolio Vola (raw) = √(Σ(w_raw × σ)²)

Scale Factor = Target Vola / Portfolio Vola (raw)

w_base = w_raw × Scale Factor

Step 3: Asset Class Risk Contribution

Risk Contribution (per market) = w_base × σ

AC Risk (per class) = √(Σ Risk Contrib²) for all markets in AC

Step 4: Apply AC caps

Cap Level = AC Cap (e.g. 40%) × Target Vola

AC Scale Factor = MIN(1; Cap Level / AC Risk)

w_TVR (final) = w_base × AC Scale Factor

Step 5: Final Portfolio Vola

Portfolio Vola (TVR) = √(Σ(w_TVR × σ)²)

Should be ≤ Target Volatility

Conclusion: Design determines performance

Trend following is not a homogeneous approach .

The performance differences between managers are less attributable to "alpha generation" and "magic indicators" than to conscious risk management decisions .

1. Flexible Asset Allocation (TVR) enables opportunistic participation in concentrated trends – such as in Fixed Income in 2019.

2. Constant Risk Targeting (CRT) stabilizes the risk profile and prevents unwanted exposure reduction during strong trend phases.

3. Asset class caps offer the optimal middle ground between flexibility and diversification.

For systematic traders, this means: Know your risk rescaling design and understand in which market environments it works. There is no "best" system – only the system that fits your risk philosophy and time horizon.

One note:

The systems we actually trade have extremely simple signal generation. Usually only with 1-2 parameters. If that...

Risk is the only thing we can manage during a trade.

You can download the specific implementation steps for a sample system via the following link:

TVR konkret.pdf

Download PDF • 360KB

resources

· Original Paper: Kaminski, K. & Yang, A. (2020). "Risk Variation in Trend-Following Systems". AlphaSimplex Group.

About the author

Tobias Lüke is a systematic trader with a focus on trend following and risk management.

On cja.capital he analyzes quantitative trading strategies and shares practical implementations.

Disclaimer: This article is for educational purposes only and does not constitute investment advice. Trend-following strategies can cause significant losses. Conduct your own analysis and consult professional advisors.