Multi-Pair Portfolios vs Single-Pair EAs

A portfolio of EAs across several pairs does not add edge, but it smooths the equity curve and cuts drawdown when the systems are truly different. Five to ten uncorrelated systems capture most of the benefit. A single-pair EA is simpler to understand and watch.

Every automated trader eventually faces the same fork: keep refining one system on one pair, or run several systems across several pairs and let diversification do some of the work. Both roads are legitimate. What matters is understanding what each one actually buys you - and what it costs.

What a portfolio really buys: smoother, not bigger

Diversification does not add edge. Each system's expectancy is whatever it is. What a portfolio changes is the path: when losing periods do not line up, one system's flat month overlaps another's good month, and the combined equity curve is smoother than any component.

Smoother matters more than it sounds. Drawdown is the number that ends trading careers - it drives the account size you need, the leverage you can safely use, and whether you abandon a sound system at its low. Halving your worst drawdown for the same return is a better deal than most optimisation will ever hand you.

The correlation trap

The whole benefit rests on one assumption: the systems are actually different. That fails more often than people expect, in quiet ways:

• Shared currency exposure. Long EURUSD, short USDCHF and long GBPUSD are three tickets on the same horse: the dollar. A dollar squeeze hits all three at once.
• Shared logic. The same trend-following idea deployed on six pairs is one bet on "trends persist", six times. When markets chop, all six chop together.
• Shared regime. Systems born from the same data window often share the same blind spot, and a regime change hits them as a group.

The honest check is empirical: line up the systems' daily returns over the same calendar window and look at how often they lose together - especially in the worst weeks. Correlation in the tails is what hurts; averages flatter the book.

Risk budgeting across legs

The classic error is sizing every leg as if it were the only system on the account. Ten legs each risking 1% per trade can easily stack into simultaneous exposure far beyond what any single-system trader would accept. The discipline that keeps a book safe:

• Set a total account risk budget first, and divide it across legs - a leg's size falls as the book grows.
• Treat correlated legs as one allocation, not several.
• Judge the portfolio on its combined worst drawdown on shared history, never on the sum of each leg's own best-case numbers.

The sizing maths itself is the same as for one system - covered in position sizing and, for the adventurous, volatility scaling - applied to the book instead of the leg.

When single-pair focus wins

Portfolios are not free. Every extra leg is another system to validate, monitor and eventually retire. Complexity is a real cost, and it compounds. Focusing on one strong system is the better road when:

• your capital is small enough that minimum lot sizes make multi-leg risk budgets impossible to hit;
• you cannot yet tell a robust system from a curve-fit one - learn that on one system first, not ten;
• the candidate systems fail the correlation check anyway, so the "portfolio" would be one bet in disguise.

The practical path

Start with one system you trust because it survived honest out-of-sample testing. Add a second only when it is genuinely different - different pair, different logic, different holding period - and measures that way on shared history. Let the book grow one proven, uncorrelated leg at a time.

FAQ

Is a portfolio of EAs better than one good EA?

A portfolio of genuinely different systems usually produces a smoother equity curve for the same total risk, because the systems' losing periods rarely line up. The benefit disappears if the systems are near-copies of each other.

How many systems make a sensible portfolio?

Diversification benefit climbs quickly over the first handful of genuinely uncorrelated systems and flattens after that. Five to ten distinct systems capture most of the effect; fifty near-duplicates capture none.

What is the biggest mistake in EA portfolios?

Counting correlated systems as diversification. Ten strategies that all buy USD strength are one position in disguise, and they will all draw down together.

Nebula builds and validates individual MetaTrader systems on the free tier; the portfolio engine on Pro and above assembles tested systems into a book and judges the combination on shared history rather than on each leg's own highlight reel.