Portfolio Allocation

Quick answer: Portfolio allocation is the decision of how to divide capital and risk across multiple strategies or instruments, where correlation between the components, not their individual returns, is the dominant driver of the combined equity curve's drawdown and risk-adjusted performance.

In simple words

Portfolio allocation is deciding how to split your capital across several strategies or instruments rather than putting everything into one. The key insight is that how the pieces move together matters more than how good each one is alone. Two decent strategies that rarely lose at the same time can produce a smoother combined curve than one excellent strategy on its own.

Purpose

Portfolio allocation exists because combining imperfectly correlated return streams reduces aggregate risk for a given return, so the backtest of a well-allocated book has shallower drawdowns and a higher risk-adjusted ratio than any single component.

Professional explanation

Correlation is the dominant lever

When you combine return streams, the variance of the whole is not the average of the parts; it depends on the covariances between them. Two strategies with the same standalone volatility combine into a portfolio whose volatility is lower the less correlated they are, and only equal to the average when they are perfectly correlated. This is why allocation is fundamentally about correlation rather than about picking the single best strategy. A book of moderately good, weakly correlated strategies routinely beats a single strong one on drawdown and Sharpe, because their bad periods do not coincide.

Equal-weight, risk-weight and risk parity

The simplest allocation is equal capital weight, but equal capital does not mean equal risk, because a more volatile component contributes more risk per rupee. Risk-weighting, and its formal cousin risk parity, instead allocate so that each component contributes the same amount of risk to the portfolio, typically by scaling inversely to volatility and then accounting for correlations. Risk parity tends to produce more balanced, diversified exposure than capital weighting, though it can require leverage to reach a target return and is sensitive to the volatility and correlation estimates it relies on.

Correlations are unstable, especially in stress

The central danger in allocation is that correlations are not constant. In calm markets strategies may look nicely diversified, but in a crisis correlations across risk assets tend to rise toward one, so the diversification that looked solid in the backtest evaporates exactly when it is needed. A portfolio backtest that assumes a single full-sample correlation matrix will overstate the diversification benefit. Honest allocation studies examine correlations in stressed sub-periods separately and stress-test the book against a scenario where cross-correlations spike.

Estimation error and the case for simplicity

Optimised allocations that minimise variance for a target return are notoriously fragile, because they amplify estimation error in the expected returns and covariances, often producing extreme, concentrated weights that perform poorly out of sample. This is why simple, robust schemes such as equal risk contribution frequently outperform mean-variance optimisation in live trading despite looking inferior in-sample. Responsible backtesting treats the covariance matrix as an uncertain estimate, applies shrinkage or caps on weights, and validates the allocation out of sample rather than trusting the in-sample optimum.

Rebalancing, capacity and cost in the backtest

An allocation is not static: as components drift, the book must be rebalanced back to target weights, and rebalancing itself has costs and look-ahead traps. Rebalancing too often incurs brokerage, STT and slippage that erode the diversification benefit; too rarely lets the riskiest component dominate. The rebalance must use only information available at the rebalance date, and the frequency is a parameter that should survive sensitivity testing. Capacity also matters: an allocation that works for a small account may not scale, because larger positions move the market and increase slippage, which a naive backtest on historical prices will not capture.

Equal-weight vs Risk-parity allocation

Practical example

Limitations

• Correlations are unstable and tend to rise toward one in crises, erasing diversification when it matters most
• Mean-variance optimisation amplifies estimation error, producing fragile, concentrated weights out of sample
• A single full-sample correlation matrix overstates the diversification benefit a stressed book would see
• Rebalancing to target weights adds brokerage, STT and slippage that erode the combined edge
• Capacity limits mean an allocation that works small may not scale, which historical-price backtests miss

Why it matters in practice

• Combining weakly correlated streams lowers aggregate drawdown and raises Sharpe more than picking one strategy
• It shifts the focus of validation from single-strategy returns to the covariance structure of the whole book

Common mistakes

• Believing many instruments equals diversification when they share a common driver like broad market beta
• Trusting a full-sample correlation matrix instead of examining correlations in stressed sub-periods
• Running a mean-variance optimiser and accepting extreme in-sample weights without shrinkage or caps
• Ignoring rebalancing costs so the backtested diversification benefit is overstated
• Equating equal capital weight with equal risk, letting the most volatile component dominate
• Assuming the allocation scales to any account size without modelling capacity and market impact

Professional usage

Multi-strategy desks treat the covariance structure of their return streams as the primary object of study, not the individual strategies, and they allocate risk rather than capital, most often via equal risk contribution with shrinkage applied to the covariance estimate. They deliberately measure correlations in stressed regimes, cap single-component weights, and stress-test the book against a scenario where cross-correlations spike toward one. Rebalancing frequency, transaction costs and capacity are modelled explicitly, and simple robust allocations are preferred over fragile in-sample optima that fail to generalise.

Frequently asked questions

Voice search & related questions

Natural-language questions people ask about Portfolio Allocation.

Sources & references

Last reviewed 12 July 2026. Educational content only — not investment advice. Markets and rules change; verify current conventions with SEBI, NSE/BSE and your broker.