L1 · MULTI-ASSET PUBLIC SHADOW PORTFOLIO

From views
to weights.

A rules-based research portfolio that connects strategic allocation, a transparent tactical overlay, HKD translation, transaction costs, risk concentration and return attribution. The objective is a repeatable decision process, not a claim of market foresight.

Read the evidence Download latest weights

This is research, not investment advice

The series is a paper portfolio using public ETFs and foreign-exchange data. It contains no real money, personal account or broker execution.

01 INVESTMENT QUESTION

Does a simple multi-asset policy plus a trend overlay improve the decision process?

VERSION POLICY_V1

The portfolio is judged against a static policy portfolio, a 60/40 index proxy and an equal-weight multi-asset basket. The question is not whether it earned the highest return in history; it is whether the rules are explicit, repeatable and explainable after costs and currency effects.

Decision mandate

Invest across equity, rates, credit, gold, commodities and cash.
Hold a fixed strategic allocation as the primary benchmark.
Apply one transparent trend rule to equity sleeves only.
Rebalance monthly with no leverage and no shorting.
Preserve the policy version when evaluating the process.

The central conclusion

Over the full sample, the simple 60/40 proxy delivered a higher annualised return and Sharpe ratio than either L1 portfolio. The trend overlay slightly reduced volatility but did not create enough return to justify a claim of superior allocation skill.

02 METHOD

Separate the strategic portfolio from the tactical decision.

MONTH-END DECISIONS

L1 SAA

Policy weights

40% equity, 35% duration and credit, 10% gold, 5% commodities and 10% short-term bills.

L1 TAA

Trend overlay

If a regional equity sleeve has a negative 10-month return, halve that sleeve and hold the proceeds in bills.

Evaluation

Decision-time discipline

Targets use information through the prior month-end. Returns are in HKD, unhedged, with 10 bps cost per unit of turnover.

Monthly return_t = Σ w_j,t · r_j,t − turnover_t · 10 bps

Table 1 | Strategic allocation and public proxies
Sleeve	Ticker	Policy weight	Role
US large-cap equity	SPY	20%	Global growth
Developed ex-US equity	VEA	10%	Geographic diversification
Emerging-market equity	EEM	10%	Higher-growth and FX exposure
US 7–10 year Treasury	IEF	20%	Duration and recession hedge
US investment-grade credit	LQD	15%	Carry and credit exposure
Gold	GLD	10%	Real-rate and tail hedge
Broad commodities	DBC	5%	Inflation sensitivity
Short-term Treasury bills	BIL	10%	Liquidity and tactical buffer

03 RESULTS

The benchmark won. The tactical overlay bought a little stability.

FULL SAMPLE · AFTER COST

L1 TAA CAGR—

60/40 CAGR—

L1 TAA volatility—

60/40 Sharpe—

Cumulative growth of HKD one for L1 SAA, L1 TAA, 60/40, and equal weight. — Figure 1 | Cumulative net asset value after transaction costs. Source: public adjusted ETF closes, translated to HKD, monthly rebalancing.

Table 2 | Full-sample outcome by portfolio
Portfolio	CAGR	Volatility	Sharpe	Max drawdown	Avg. turnover
L1 SAA	—	—	—	—	—
L1 TAA	—	—	—	—	—
60/40	—	—	—	—	—
Equal weight	—	—	—	—	—

Full-sample annualised return and volatility for L1 SAA, L1 TAA, the 60/40 benchmark and equal weight. — Figure 2 | The 60/40 proxy sits highest on return, while L1 TAA sits furthest left on volatility. Source: public ETF data translated to HKD; monthly net returns after 10 bps transaction costs.

How to read the risk-return map

The chart makes the trade-off visible without ranking portfolios by a single number. The 60/40 proxy has the highest return in this sample but also higher volatility than L1 TAA. L1 TAA moves left on risk, not up on return: the overlay reduced variance without repairing the performance gap.

Drawdown paths for L1 SAA, L1 TAA, and the 60/40 benchmark. — Figure 3 | Drawdown path. Source: monthly portfolio returns after transaction costs, HKD translation and monthly rebalancing.

04 CURRENT SNAPSHOT

Policy and current allocation remain close because every equity trend is positive.

DATA THROUGH 29 OCT 2026

Policy and current portfolio weights by asset sleeve. — Figure 4 | Strategic policy weights versus the latest tactical allocation. Source: policy v1 and current public-data snapshot.

Stacked area chart showing tactical target weights for equity, fixed income and credit, real assets and cash through the full sample. — Figure 5 | Grouped tactical target weights through time. The dashed white line marks the 40% strategic equity allocation; trend-driven reductions transfer equity to cash. Source: month-end targets from policy v1.

The tactical overlay is episodic, not a monthly forecast

Most of the time the grouped allocation stays close to policy. When an equity trend turns negative, the visible cash band expands as the affected sleeve is halved. This is a bounded risk-control rule, not a claim that the model can time every turning point.

Table 3 | Current portfolio weights
Sleeve	Ticker	Current weight
US large-cap equity	SPY	—
Developed ex-US equity	VEA	—
Emerging-market equity	EEM	—
US 7–10 year Treasury	IEF	—
US investment-grade credit	LQD	—
Gold	GLD	—
Broad commodities	DBC	—
Short-term Treasury bills	BIL	—

05 RISK

Capital weights understate the equity concentration.

TRAILING 36 MONTHS

Risk contribution by asset for the latest portfolio. — Figure 6 | Contribution to portfolio variance using a trailing 36-month covariance matrix. Source: public data and current portfolio weights.

Equity is the main risk engine

The three equity sleeves are 40% of policy capital but account for roughly 61% of model-based portfolio variance. The conclusion is structural, not a market forecast.

Cash and commodities can hedge

Their risk contributions differ from their capital weights and can be negative when their covariance with the portfolio is favourable. That makes capital allocation and risk allocation separate decisions.

06 ATTRIBUTION

Where the return came from, before pretending it came from a clever model.

ANNUALISED APPROXIMATION

Annualised return contribution by asset sleeve. — Figure 7 | Average monthly contribution multiplied by twelve. Source: target weights and monthly asset returns; this is an additive approximation, not an exact decomposition of compounded wealth.

07 DECISION LOG

The current decision can be replayed from the written rule.

OPEN SNAPSHOT

October 2026 decision

US equity trend: —.
Developed ex-US equity trend: —.
Emerging-market equity trend: —.
All signals positive, so no equity sleeve is reduced.
Hold the policy allocation until the next month-end review.

What would change the decision?

A negative 10-month return in any equity sleeve would halve that sleeve and move the proceeds to short-term bills. A policy change requires a new version number; it cannot be applied retroactively.

08 LIMITS

What this case proves, and what it does not.

PAPER RESEARCH

It does show

A repeatable process from allocation policy to monthly weights.
Explicit handling of HKD translation, costs and signal timing.
The difference between capital weight and risk contribution.
An honest benchmark result, including underperformance.

It does not establish

Real-money execution or personal PnL.
A statistically significant alpha from the trend rule.
Forward-looking market views or an investment recommendation.
Production-grade market impact, taxes or closed-market execution.

Why a 60/40 is important here

A well-known benchmark outperformed the more diversified allocation over this sample. The correct response is to investigate period dependence, currency exposure and risk objectives, not to rewrite the backtest until the desired winner appears.

09 REPRODUCE

Data, weights and methodology are all inspectable.

PUBLIC DATA

Evidence pack

Download the latest weights, monthly history and investment memo.

Weights CSV Monthly history Data manifest Investment memo

Run locally

python -m pip install -e ".[dev]"
python scripts/build_case.py
python -m pytest -q
python -m http.server 8000

Data boundary

Adjusted ETF closes and USD/HKD are downloaded from Yahoo Finance via yfinance. Raw vendor data is not redistributed. Returns are unhedged HKD translation of USD-listed proxies.