# Multi-strategy portfolios: holding a strategy beside your funds

Canonical URL: https://marketheist.io/portfolio/learn/multi-strategy-portfolio
By Jeffrey Lin · Published 2026-11-12 · Updated 2026-11-12 · Data through Sep 2026
Category: Diversification

## Key takeaways

- A strategy held as a holding is one return stream: the result of its rule applied month by month. The parent portfolio gives it a weight and rebalances it against the other holdings like any fund.
- Diversifying across assets means owning things that react differently to the economy; diversifying across strategies means owning different rules, which can be built on the same assets. A trend rule on US stocks and US stocks held outright own the same market but are in it at different times.
- The correlation between a strategy sleeve and the buy-and-hold sleeves, and its share of the portfolio's risk, show how much it adds; a strategy that is highly correlated with what the portfolio already holds adds little.
- Strategy results carry a hindsight risk that funds do not: the rule was usually designed or published after studying the same history it is tested on, so its past record is likely to be better than its future one.
- MarketHeist composes each strategy sleeve from its own rules on monthly data first, then simulates the parent over the sleeves' returns, with the same metrics, risk shares and caveats as a plain portfolio.

## Definition

A multi-strategy portfolio holds one or more rules-based strategies, such as a trend-following or momentum rule, as holdings in their own right, at fixed weights beside ordinary buy-and-hold funds. It diversifies across ways of holding assets as well as across the assets themselves, since a strategy that trades the same market at different times can behave quite differently from simply owning it.

Also known as: multi-strategy portfolio, strategy diversification, combining strategies, strategy sleeve, trend following in a portfolio, diversifying across strategies, portfolio of strategies.

## How it works

A buy-and-hold fund is always invested. A strategy is a rule that decides what to hold, and when: a trend filter holds the stock market only while its price is above its moving average and moves to cash otherwise; a momentum rotation holds whichever of a few funds has risen most recently. Either can be run on its own, producing a monthly return stream, and that stream can be held as a sleeve in a larger portfolio at a fixed weight.

The diversification it brings is of a different kind. Two stock funds from different countries diversify a little because their markets differ; a trend rule on the very same stocks can diversify more in the months that matter, because it steps out of the market during long declines. Over calm, rising years the strategy and the stocks move almost together; in a prolonged bear market they part ways. Their correlation over the whole window averages those two regimes.

Measuring it uses the same tools as for funds. Each sleeve's share of the portfolio's risk shows whether a strategy is a small seasoning or the main ingredient, and the co-movement groups show whether the strategy counts as a separate bet or as more of what is already there. Combining strategies is like adding a second cook to a kitchen: if both follow the same recipe you get more of the same dish; if one cooks differently, the menu changes even when they share the ingredients.

The hard part is honesty about the record. Funds have prices that anyone could have bought; a strategy's history is a simulation of a rule, and rules that get published and named are usually the ones that looked good on the history available when they were written. Parameters such as a 10-month window or a 12-month lookback were chosen with hindsight, trading costs and taxes fall on every switch, and a rule that worked in past bear markets may meet a decline of a different shape.

## Formula

**A strategy sleeve's monthly return (trend-timing example)**

```
R_t^{strat} = s_{t - 1} r_t, s_{t - 1} = {\begin{pmatrix} 1 & P_{t - 1} > MA_{10} (P)_{t - 1} \\ 0 & otherwise \end{pmatrix}
```

where:

- Rₜˢᵗʳᵃᵗ: the strategy sleeve's return in month t
- rₜ: the underlying fund's total return in month t
- sₜ₋₁: the rule's decision at the end of month t − 1: 1 = invested, 0 = cash earning 0
- Pₜ₋₁, MA₁₀(P)ₜ₋₁: the fund's level and its 10-month simple moving average at the end of month t − 1

The decision uses only month-ends already passed, so a month's own return never decides whether that month is held.

**The parent over its sleeves**

```
R_t = \sum_s X_{s,t - 1} R_{s,t}
```

where:

- Rₜ: the whole portfolio's return in month t
- R_s,t: sleeve s's return in month t: a fund's total return or a strategy's simulated return
- X_s,t−1: sleeve s's actual share of the portfolio at the end of month t − 1, reset to its target on the parent's rebalance months

**What a sleeve adds to volatility**

```
RS_s = \frac{W_s (\Sigma W)_s}{W^⊤\Sigma W}, (\Sigma W)_s = \sum_j W_j \rho_{s j} \sigma_s \sigma_j
```

where:

- RSₛ: sleeve s's share of the portfolio's risk (shares sum to 1)
- Wₛ, Wⱼ: the parent's target weights for sleeves s and j
- Σ: covariance matrix of the sleeves' monthly returns
- ρₛⱼ: correlation between sleeves s and j (1 when j = s)
- σₛ, σⱼ: the sleeves' volatilities

A strategy sleeve with a low correlation to the others contributes less risk than its volatility alone would suggest.

## Worked example

The classic 60/40 portfolio rebuilt with half of its stock portion handed to a strategy: SPY (the S&P 500) 30%, BND (US bonds) 40%, and the remaining 30% in the trend-timing strategy, which holds the US stock market (VTI) while it is above its 10-month moving average and cash otherwise. The sleeves are reset each December. Compared with the plain 60/40 portfolio from Apr 1993 to Sep 2026, about 34 years.

The multi-strategy version grew 8.1% a year against 8.5% for the plain 60/40: a little less growth. In exchange, volatility fell from 9.2% to 7.5% and the deepest decline from -30.3% to -16.0%, and the Sharpe ratio was 1.08 against 0.93.

The strategy diversified without adding a new asset. The SPY sleeve and the strategy sleeve both own US stocks, yet the highest correlation between any two sleeves was 0.69, well short of the near-perfect correlation of two broad US stock funds. The strategy holds the very market SPY holds, but at different times: invested through most rising years, in cash through much of the long declines of 2000 to 2002 and 2008.

Risk shares show who carries what. SPY was 30% of the money and 54.7% of the risk; BND 40% of the money and only 8.4% of the risk; the strategy sleeve carried the remainder. Counted as co-movement groups, the three sleeves made 1.34 independent bets.

Now the caveat. The trend rule was published by Mebane Faber in 2007, after part of this window, and its record here rests mostly on two deep bear markets that unfolded slowly enough for a monthly rule to step aside. A decline of a different shape, such as a fast crash and rebound, can cost it instead, as the trend-filter guide shows. Read the result as what a strategy sleeve did in this history, not what it is owed.

Data through Sep 2026. Reproduce id: a6e109428a3de635 (this guide, its lab setup and the data month).

## How MarketHeist computes it

- **Data**: Monthly total returns (dividends reinvested). Funds younger than the period shown are extended backward with older funds, spliced on returns, not prices.
- **Window**: The worked example and the lab use the common window of the sleeves and the comparison, Apr 1993 to Sep 2026 (402 months), ending at the data month of this page.
- **Nested portfolio**: Each sleeve is the monthly return stream of a whole portfolio (as the MarketHeist MCP reads it, with that portfolio's own rebalancing rule and overlay) or of a fund. The parent holds the sleeves at their target weights and rebalances them at the stated cadence, as the portfolio engine composes a portfolio of portfolios. Risk shares come from the sleeves' covariance; bets count the sleeves' co-movement groups (average correlation at or above the grouping threshold) the way the risk guides do.
- **Strategy sleeves**: A strategy sleeve is the strategy's own monthly return stream, built from its rule on monthly data before the parent sees it. The trend-timing strategy is the whole-book trend filter on VTI with a 10-month moving average: decided at month-end, applied the next month, cash earning 0, no trading costs. (`applyPortfolioOverlay` in packages/portfolio/engine/overlay.ts)
- **Rotation strategies**: Momentum strategies such as dual momentum rank their funds on trailing returns that end at least one month before the month held, so a ranking never uses the month it trades. (`effectiveSkipMonths` in packages/portfolio/engine/nodes.ts)
- **Parent**: Simulated with the same function as any portfolio over the sleeves' monthly returns: each sleeve starts at its target share, drifts with its own returns and is reset at the end of each rebalance month. No costs or taxes. (`simulatePortfolio` in packages/portfolio/engine/portfolio.ts)
- **Risk shares and bets**: Risk shares on the parent's target weights with the population covariance of the sleeves' monthly returns; independent bets count co-movement groups of sleeves (clusters whose average correlation is at least 0.5). (`exposureOverlap` in packages/portfolio/engine/overlap.ts)
- **Window**: The common window of every sleeve and the comparison, ending at the latest data month. A strategy's stream starts after its own warm-up (the months its rule needs before the first decision), which can make it the sleeve that limits the window.
- **In the app and the MCP**: Strategy sleeves can be canonical strategies, rotations, a rule from the backtester applied to a fund or another portfolio, or a published backtest; nested portfolios of strategies are checked for cycles and a maximum depth. (`composeNodeSeries` in packages/portfolio/engine/nodes.ts)
- **Engine functions**: `simulatePortfolio` (packages/portfolio/engine/portfolio.ts), `riskContributions` (packages/portfolio/engine/risk.ts), `exposureOverlap` (packages/portfolio/engine/overlap.ts), `effectiveBetsFromGroups` (packages/portfolio/engine/overlap.ts)

## Multi-strategy portfolio (a strategy as a holding) compared with its neighbours

Diversifying across assets versus across strategies

| Approach | Adds | Shows up as | Main risk |
| --- | --- | --- | --- |
| More assets, bought and held | Different reactions to growth, inflation and rates | Lower correlation between holdings in most periods | Correlations rise in sell-offs |
| A strategy on an asset already held | Different timing in the same market | Correlation well below 1 over full cycles, near 1 in calm rises | Hindsight in the rule; whipsaws; costs |
| A strategy on new assets (rotation) | Both new assets and different timing | Low correlation, concentrated holdings at times | Turnover and concentration |
| Several strategies together | Rules that fail at different times | Risk spread across sleeves | Rules built on the same idea fail together |

## Limitations

- A strategy's record is a simulation of a rule, usually designed or published after studying much of the history it is tested on. Expect a strategy sleeve's past to flatter its future more than a fund's does.
- Strategy sleeves trade. The simulation charges no costs or taxes on the rule's switches or on the parent's resets, and the strategies here earn nothing on cash.
- Over a short window a strategy can look like a diversifier or a drag depending on whether a large decline happened in it; the full-window correlation hides which.
- Risk shares are measured on the sleeves' streams and a strategy's risk changes with its holdings, so a strategy sitting in cash for long spells has a risk share that understates its risk when invested.
- Long histories are extended backward with older funds or indexes, so early years approximate rather than reproduce them.

## Frequently asked questions

**What is a multi-strategy portfolio?**

A portfolio that holds rules-based strategies, such as trend following or momentum rotation, as holdings at fixed weights beside buy-and-hold funds, and rebalances among them like any other holdings. In MarketHeist each strategy is one sleeve with its own return stream.

**Is diversifying across strategies better than diversifying across assets?**

They are different, and they combine. Asset diversification relies on markets reacting differently; strategy diversification relies on rules being in the market at different times. A trend rule on stocks can diversify a stock holding in long declines while adding little in calm rising markets, and it carries risks of its own: whipsaws, costs and hindsight in its design.

**Can I hold a trend-following strategy as part of a 60/40 portfolio?**

In MarketHeist, yes: a portfolio can hold a strategy such as the 10-month trend-timing rule as one holding beside stock and bond funds, and the analysis reports its share of the risk and its correlation with the rest. The worked example on this page replaces half of a 60/40 portfolio's stocks with that rule.

**Why might a strategy's backtest overstate its future?**

Because the rule and its settings were usually chosen after seeing how they would have done. Even with no deliberate fitting, the strategies that get published and named are the survivors of many ideas tried. Tests on data the rule was not designed on, such as later years or other markets, are a better guide; see the train/test split guide.

**How many strategies does a portfolio need?**

There is no fixed number. What matters is whether the strategies fail at different times. Two momentum rules on similar funds behave much like one; the co-movement count of independent bets shows whether added sleeves are new bets or repeats.

## References

1. Faber, M. T. (2007) A Quantitative Approach to Tactical Asset Allocation <https://papers.ssrn.com/sol3/papers.cfm?abstract_id=962461>. Journal of Wealth Management 9(4). The 10-month moving-average timing model.
2. Antonacci, G. (2014) Dual Momentum Investing. McGraw-Hill.
3. Hurst, B., Ooi, Y. H. and Pedersen, L. H. (2017) A Century of Evidence on Trend-Following Investing. Journal of Portfolio Management 44(1).
4. Asness, C. S., Moskowitz, T. J. and Pedersen, L. H. (2013) Value and Momentum Everywhere. Journal of Finance 68(3). Diversification between strategies.
5. McLean, R. D. and Pontiff, J. (2016) Does Academic Research Destroy Stock Return Predictability?. Journal of Finance 71(1). Returns after publication.

## Try it

App: https://marketheist.io/portfolio/s/faber-trend-timing

MCP prompt (MarketHeist MCP server, https://marketheist.io/mcp):

```
Using MarketHeist.io, analyze a portfolio that is 30% SPY, 40% BND and 30% the Faber 10-month trend-timing strategy, rebalanced yearly, and compare it with a plain 60/40 portfolio over the same months: CAGR, max drawdown, Sharpe, each holding's share of the risk and the correlation between the strategy and SPY
```

Historical simulation, not advice.
