Investboard …
KnowledgeStrategy & portfolio
Thirty holdings across six asset groups, with a ten-year comparison and a clear view of technology, semiconductor and AI exposure. 30 holdings, an S&P 500 comparison and technology exposure.
Weather the Storm is a hypothetical 30-holding allocation with 54% equities and 46% in bonds, gold, commodity futures, managed futures and Treasury bills. The 2016–2025 comparison uses a retrospectively selected allocation. The report measures technology, semiconductor and AI exposure, including overlapping holdings inside funds.
Each position targets 3% or 4%, with a 5% rule applied after modeled daily-close rebalancing.
The historical stock selection used the same decade shown in the chart, creating substantial hindsight bias.
Looking through the funds reveals overlapping companies and shared economic exposures.
The full report documents every holding, its intended role, risks and the simulation assumptions.
Weather the Storm is a hypothetical 30-holding allocation with 54% equities and 46% in bonds, gold, commodity futures, managed futures and Treasury bills. The 2016–2025 comparison uses a retrospectively selected allocation. The report measures technology, semiconductor and AI exposure, including overlapping holdings inside funds.
Each position targets 3% or 4%, with a 5% rule applied after modeled daily-close rebalancing.
The historical stock selection used the same decade shown in the chart, creating substantial hindsight bias.
Looking through the funds reveals overlapping companies and shared economic exposures.
The full report documents every holding, its intended role, risks and the simulation assumptions.
The Investboard: Weather the Storm Portfolio is a 30-holding research allocation with 54% in equities. It explores how different sources of return can share a portfolio, with explicit target weights, rebalancing rules and a view of the companies held beneath the fund labels.
The design keeps an equity majority and allocates the remaining capital across bonds, gold, commodity futures, managed futures and Treasury bills. Each holding begins at 3% or 4%. Those small positions create room for different exposures, but the number of tickers alone tells us little about the risks underneath.
Research note: the allocation was selected on September 30, 2026 using the 2016–2025 period shown below. The comparison is a retrospective backtest.
Our companion essay, Is this time different?, examines a recurring problem in technology booms: the value a technology creates and the return earned by the capital financing it can diverge. AI can become indispensable while competition, overbuilding, depreciation or an excessive purchase price leave investors disappointed.
The “storm” refers to that uncertainty, including a possible AI-driven market correction. The portfolio retains AI-linked companies within a broader allocation. Below, we quantify its technology, semiconductor and AI exposure, including ownership through funds.
Ray Dalio’s diversification argument is useful because it asks what makes investments behave differently. Bridgewater’s All Weather framework organizes exposures around surprises in growth and inflation. Geographically broad equity funds can still share exposure to the same earnings cycle, interest rates and technology valuations.
Our allocation applies that question to a portfolio with a 54% equity majority. It assigns target capital weights across asset classes; Bridgewater’s institutional approach instead seeks to balance risk contributions.
The research measures total-return streams, including reinvested distributions. These combine price changes and cash income. Gold contributes through price changes, while equities, bonds and Treasury bills have their own return sources.
| Allocation | Target | Intended role |
|---|---|---|
| Equities | 54% | Growth, with several industries and regions |
| Bonds | 20% | Nominal Treasuries and inflation-linked bonds across maturities |
| Gold | 8% | A different response to monetary and confidence shocks |
| Commodities | 10% | Broad commodities, agriculture and industrial metals |
| Managed futures | 4% | A changing systematic mix of long and short futures |
| Treasury bills | 4% | Short-duration liquidity |
Each holding targets either 3% or 4%. The model rebalances the whole portfolio at completed quarter-ends and whenever a daily closing position exceeds 5%. The limit therefore applies after rebalancing: market moves can take a position above it between trades. Through September 2026, the largest observed weight was 5.55% before correction and 4.98% afterward.
The first version, Portfolio A, established the allocation and diversification baseline. We then kept the fund allocation fixed and searched for a stock basket with higher historical returns under related risk and correlation limits. Finally, we replaced the 3% Chunghwa Telecom holding with 3% Verizon to reduce direct Taiwan exposure. That revised portfolio is the one shown here.
The replacement involved a small trade-off: the portfolio now has three non-US direct stocks rather than four, and its weekly volatility approximation slightly exceeds the original search ceiling. The final simulation still met the corresponding daily volatility, drawdown and correlation limits. The final allocation therefore reflects both the search objective and the preference to reduce Taiwan dependence.
The comparison begins with $10,000 at the December 31, 2015 close and ends on December 31, 2025. Both series use USD total-return proxies with distributions reinvested. The portfolio includes modeled trading costs; the benchmark is buy-and-hold SPY, a fund tracking the S&P 500. SPY provides a familiar equity reference for the multi-asset allocation.
Growth of $10,000 in USD
2016–2025 · Hypothetical backtest · Distributions reinvested
| 2016–2025 measure | Weather the Storm | S&P 500 proxy (SPY) |
|---|---|---|
| Annualized return | 17.72% | 14.71% |
| Total return | 411.40% | 294.65% |
| Ending value of $10,000 | $51,140 | $39,465 |
| Annualized daily volatility | 9.85% | 18.01% |
| Maximum daily drawdown | −14.55% | −33.72% |
The table below makes the annual results behind the chart available without relying on color or reading a plotted line.
| Calendar year | Weather the Storm | S&P 500 proxy (SPY) |
|---|---|---|
| 2016 | 22.35% | 12.00% |
| 2017 | 20.62% | 21.70% |
| 2018 | 6.07% | −4.57% |
| 2019 | 20.77% | 31.22% |
| 2020 | 22.56% | 18.34% |
| 2021 | 15.38% | 28.73% |
| 2022 | 3.16% | −18.18% |
| 2023 | 18.31% | 26.18% |
| 2024 | 28.24% | 24.88% |
| 2025 | 22.24% | 17.72% |
The selected basket recorded ten positive calendar years. Within those years, its largest daily peak-to-trough decline was 14.55%, compared with 33.72% for SPY. Calendar-year results and the experience of holding through a drawdown tell different parts of the story.
The original Portfolio A returned 11.69% annually in the same decade. From January 1 to September 29, 2026, the revised portfolio returned 9.73% versus SPY’s 12.94%. The partial year sits outside the ten-year chart and optimization objective; its interpretation is addressed in the methodology below.
The search evaluated 2,398 full daily portfolio simulations from a current universe of 1,269 eligible US-listed stocks and ADRs. Selecting and evaluating on the same decade introduces hindsight and selection bias; using surviving listings adds survivorship bias. The result is not an independent performance test. The 2026 partial year also reflects current-universe selection and earlier research using 2026 observations. The window excludes the 2000–2002 and 2008 crises.
Prices come from FMP’s dividend-adjusted end-of-day series. The model deducts 0.10% for each dollar bought or sold, excludes initial entry costs and assumes fractional positions and closing-price execution. Fund expenses are reflected in market prices. Investor taxes, withholding and currency hedges are excluded.
The target-weighted mean absolute pairwise correlation of weekly holding returns was approximately 0.19. Its simulated daily correlation with SPY was approximately 0.82. The weekly statistic averages distinct holding pairs; the daily statistic measures the combined portfolio against SPY. They answer different questions. The portfolio brought together varied return drivers while remaining meaningfully tied to equity-market direction.
Some histories also span different businesses. TKO’s pre-September-2023 series represents predecessor WWE; Texas Pacific Land’s pre-2021 history includes its trust structure. The report records those changes when interpreting the price history.
The first ten positions in construction order account for 30% of target capital. The PDF provides the complete allocation and reasoning for all thirty.
| Holding | Target | Intended contribution and trade-off |
|---|---|---|
| SPY · SPDR S&P 500 ETF | 3% | US large-company participation; overlaps with QQQ and direct stocks. |
| QQQ · Invesco QQQ Trust | 3% | Growth and innovation exposure; adds technology concentration. |
| IWM · iShares Russell 2000 ETF | 3% | US small companies; sensitive to financing and the economic cycle. |
| VEA · Vanguard FTSE Developed Markets ETF | 3% | Developed markets outside the US; adds currency and overseas equity risk. |
| VWO · Vanguard FTSE Emerging Markets ETF | 3% | Emerging-market participation; retains Taiwan, China and policy exposure. |
| VZ · Verizon Communications | 3% | US telecom cash flows; competition, debt and capital spending remain risks. |
| CWST · Casella Waste Systems | 3% | Waste-service demand; valuation, acquisition and execution risks remain. |
| DHT · DHT Holdings | 3% | Oil-tanker freight economics; a highly cyclical shipping exposure. |
| DRD · DRDGOLD | 3% | Gold-related operating exposure; mining and South African risks differ from bullion. |
| FCN · FTI Consulting | 3% | Advisory and restructuring demand; exposed to utilization and personnel costs. |
Each holding in the report has an intended role and a specific account of the risks that can undermine it.
The individual ticker list understates technology exposure. NVIDIA appears directly and inside funds, while some AI-related businesses sit outside the formal technology sector. Looking through the five equity ETFs gives the following snapshot at target weights, using FMP constituents updated September 29–30, 2026.
| Exposure | Share of the whole portfolio | Definition |
|---|---|---|
| Technology sector | 7.82% | Direct NVIDIA plus technology holdings inside equity funds |
| Semiconductors | 5.44% | Identified chipmakers and semiconductor equipment companies |
| Identified AI basket | 6.59% | Selected chips, infrastructure, platforms and software companies |
| Broader AI basket | 10.17% | Identified basket plus Apple, Tesla and Texas Pacific Land |
All figures are percentages of the whole portfolio, and the categories overlap. Technology follows the sector classification. Semiconductors include identified chipmakers and equipment companies. AI is an analyst-defined basket that counts the full capital weight of selected companies; it is not a measure of AI revenue. The PDF lists its membership.
NVIDIA totals approximately 3.50% after the SPY and QQQ holdings are included, representing about 64% of the identified semiconductor weight. Technology is about 14.5% of the equity allocation. The full Samsung Electronics positions, classified separately by FMP, would increase the semiconductor estimate to approximately 5.53%.
Texas Pacific Land is a less obvious connection. Its approximately 3% position has a data-center development link through the Bolt agreement. Its existing economics remain tied largely to land, water and energy, so we include the whole holding only in the broader AI interpretation.
Taiwan-linked exposure remains through TSMC, approximately 0.44% via VWO, and NVIDIA’s reliance on Taiwan’s semiconductor and manufacturing ecosystem.
These ownership totals exclude FMF’s changing exposure through long and short futures.
The two physical-gold vehicles form one 8% economic exposure. DRDGOLD adds another 3% in gold-related equities with operating risk. Broad commodities overlap with the agriculture and metals sleeves. Looking through the instruments reveals those shared drivers.
The bond allocation diversifies maturities, but long Treasuries can fall sharply when yields rise. Inflation-linked bonds still carry real-rate risk. A shock to inflation and discount rates can hurt both stocks and bonds at once. Managed futures may help in persistent trends but can lose in reversals or directionless markets.
The portfolio is USD-based. For an investor whose spending is in euros, unhedged currency movements can dominate a supposedly defensive sleeve. Broker access, product structure, fund documentation and local tax treatment also matter. Implementation depends on the investor’s jurisdiction and access to these US-listed instruments.
Capital weights and risk contributions differ. A volatile 3% stock can contribute more risk than a 4% Treasury-bill position, which is why the allocation should be read alongside its company, currency and interest-rate exposures.
The report makes each design choice inspectable: target weights, overlapping companies, economic drivers, rebalancing rules and the conditions under which the intended role of a holding may fail. It includes the complete allocation and the reasoning for all thirty positions.
The next meaningful test begins after the research date. Freeze the allocation and rules, record revisions separately and evaluate subsequent results without rewriting the starting point. A written investment mandate should establish loss tolerance, liquidity needs and review conditions before market stress forces the decision.
The result is a transparent allocation whose assumptions can be inspected, whose exposures can be measured and whose rules can be evaluated over time.
All 30 holdings. Every role and risk.
Download the English PDF with target weights, the rationale and trade-offs for every position, annual results and the complete methodology. Available to registered users, including Free accounts.
Research edition: September 30, 2026. Retrospective portfolio research.
Research snapshot: September 30, 2026. Educational portfolio research; historical results reflect the selection process and assumptions described above.
At target weights, technology represents 7.82% of the whole portfolio, semiconductors 5.44%, the identified AI basket 6.59%, and the broader AI basket 10.17%. These categories overlap. AI measures whole-company ownership within defined baskets, using September 29–30, 2026 constituents.
The simulation uses FMP dividend-adjusted prices, reinvested distributions and modeled trading costs. Selection used the same 2016–2025 decade shown in the comparison, creating hindsight and selection bias. The methodology records the assumptions and limits of this retrospective backtest.
Registered users can download the English PDF after signing in and completing the account’s required consent step. It includes all 30 positions, target weights, rationale, risks and methodology. Access is included on Free and paid plans.