EA Portfolio Drawdown: What Happens When Your EAs Lose at the Same Time

Every EA seller shows you the backtest. A smooth equity curve climbing steadily upward, with a modest 12-15% maximum drawdown. The numbers look great — so you buy five EAs, put them all on your $10,000 account, and expect smooth sailing.

Three months later, your account is down 45% and you're questioning everything. What went wrong?

Drawdown stacking. It's the single biggest risk in running a portfolio of Expert Advisors, and almost nobody talks about it.

What Is Drawdown Stacking?

EA drawdowns stack additively when multiple EAs trade the same account simultaneously, typically producing 1.5-2x the drawdown of any individual EA at moderate correlation.

The concept is simple, but its implications are devastating:

Traders intuitively think portfolio drawdown equals the worst individual drawdown. In reality, it almost always exceeds it — sometimes dramatically.

Why Individual EA Backtests Lie About Portfolio Risk

MetaTrader Strategy Tester runs each EA in complete isolation, so its results miss the three biggest sources of portfolio drawdown when EAs share an account.

This creates three critical distortions:

1. No Shared Balance Effect

Every EA on a real multi-EA account computes its lot size from one shared balance, so when EA-A loses $1,000 every other EA's next position is smaller. If EA-B uses percentage-based lot sizing, its next trade uses a smaller lot. This cascading effect doesn't exist in individual backtests.

2. No Margin Interaction

Open positions across multiple EAs consume margin from the same pool, and total usage can approach your account limit at exactly the worst moment. This can prevent EAs from opening new trades at critical moments — exactly when they might need to take recovery trades.

3. No Temporal Overlap Visibility

Individual backtests hide whether two EAs draw down in the same week, the single most important data point for portfolio risk. EA-A's max drawdown might happen during the same week as EA-C's max drawdown. You can't see this unless you simulate them together.

The Math: Correlated vs. Uncorrelated Drawdowns

Correlation between EA return streams determines almost the entire range of possible portfolio drawdowns, from mild diversification benefit to fully additive losses.

Best Case: Completely Uncorrelated EAs (ρ = 0)

Truly uncorrelated EA returns scale portfolio volatility with the square root of the EA count, halving expected risk relative to a single EA at four EAs. Drawdowns still stack, but the probability of all EAs drawing down simultaneously is low.

Notice that even with perfectly uncorrelated EAs, the portfolio drawdown doesn't improve as dramatically as you might hope. It's better than one EA alone, but far from zero.

Worst Case: Perfectly Correlated EAs (ρ = 1)

Perfectly correlated EAs produce additive drawdowns, so five EAs at 15% individual drawdown can collectively reach 75% portfolio drawdown.

Realistic Case: Partial Correlation (ρ = 0.3 to 0.6)

Most retail EA portfolios cluster between 0.3 and 0.6 average correlation due to shared USD exposure, similar strategy types, and common market regime sensitivity. Even EAs on different pairs often have correlations of 0.2-0.5 due to shared USD exposure, similar strategy types, or common market regime sensitivity.

This is the critical insight: with average correlation of 0.4, five EAs provide the diversification benefit of roughly two independent EAs. Your portfolio is far less diversified than you think.

Real Scenario: 5 EAs, Each With 15% Max Drawdown

Five EAs each showing 15% individual drawdown can collectively produce 38-52% portfolio drawdown depending on average correlation, as the worked example below shows.

Even at moderate correlation (0.35), the portfolio drawdown is more than double the worst individual EA. At higher correlation — which is common during market stress — you're looking at a drawdown that could require a 100%+ gain just to recover.

Why the Worst-Case Is Worse Than You Model

EA correlation jumps from 0.2-0.3 in calm markets to 0.7-0.9 during volatility spikes, a phenomenon called correlation convergence that hits hardest at peak drawdown. This phenomenon, called correlation breakdown (or more accurately, correlation convergence), means:

• During calm markets, your EAs might show 0.2-0.3 correlation
• During a volatility spike, correlation jumps to 0.7-0.9
• The jump happens exactly when drawdowns are largest
• Your "diversified" portfolio suddenly acts like a single concentrated bet

How to Measure Portfolio Drawdown Properly

Realistic portfolio risk requires four measurements that individual EA backtests cannot provide: correlation matrix, shared-balance simulation, Monte Carlo paths, and allocation optimization.

Step 1: Build a Correlation Matrix

A pairwise correlation matrix exposes which EA pairs are effectively the same strategy, where one highly correlated pair can dominate the entire portfolio risk profile. You need to see every EA-to-EA correlation pair, not just the average. One highly correlated pair can dominate your portfolio risk.

Step 2: Simulate on a Shared Balance

Shared-balance simulation runs all EAs forward together so each EA's lot size responds to other EAs' wins and losses, capturing what trade-list merging cannot. This captures the lot-sizing feedback effect and margin interactions. FXOptimize does exactly this: upload your MT4/MT5 backtest HTML files and see the combined equity curve with realistic shared-balance simulation.

Step 3: Run Monte Carlo Simulations

Monte Carlo simulation randomizes trade order across thousands of paths to expose the realistic 95th-percentile worst case rather than the single historical sample. This reveals:

• 95th percentile drawdown: The drawdown you'd see in 5% of scenarios — use this for risk sizing
• Expected drawdown range: What's typical vs. what's extreme
• Recovery time distribution: How long drawdowns last, not just how deep they go
• Risk of ruin: The probability of hitting a drawdown level where the account can't recover

Step 4: Optimize Allocation Weights

Allocation weights should reflect each EA's risk-adjusted return and correlation to the rest of the portfolio, not equal-weight defaults. If EA-A and EA-B are highly correlated, reducing their combined weight improves the portfolio. If EA-C has the best Calmar ratio and low correlation with others, it should get more weight.

Portfolio optimization finds the allocation weights that maximize risk-adjusted returns while keeping drawdown within your tolerance.

Practical Rules for Managing Portfolio Drawdown

Five operational rules cover the most common multi-EA blowups: tolerance buffer, correlation cap, stress test multiplier, lot sizing, and monitoring cadence.

Rule 1: Target 50% of Your Tolerance

Simulated portfolio drawdown should target half your maximum acceptable loss to leave buffer for correlation spikes and out-of-sample surprises. This gives you buffer for correlation spikes and out-of-sample surprises.

Rule 2: Limit Correlated Pairs

No two EAs with correlation above 0.5 should have a combined allocation above 30% of the portfolio. Either reduce one or replace it with something less correlated.

Rule 3: Stress Test with 2x Drawdown

Whatever your simulation shows as max drawdown, assume the real number is 1.5-2x higher. Backtest periods don't contain every possible market regime. The future will produce new worst cases.

Rule 4: Use Portfolio-Level Lot Sizing

Don't let each EA calculate lots independently on the full balance. Allocate a fraction of the balance to each EA, or use a portfolio-level risk manager that limits total exposure across all EAs.

Rule 5: Monitor Monthly

Correlations shift as markets change. Re-run your portfolio analysis monthly to catch correlation drift before it causes a blowup.

Key Takeaways

• Drawdown stacking means your portfolio drawdown is always larger than your worst individual EA drawdown
• Correlation determines severity — at 0.4 average correlation, 5 EAs behave like ~2 independent EAs
• Individual backtests miss shared-balance effects, margin interaction, and temporal overlap
• Monte Carlo simulation gives you the realistic worst-case, not just the historical worst-case
• Portfolio simulation tools like FXOptimize model what actually happens when EAs share an account
• Size for 2x the expected drawdown to survive correlation spikes and out-of-sample surprises

The traders who survive long-term with multi-EA portfolios are the ones who measure risk at the portfolio level, not the EA level. Don't be the one who discovers drawdown stacking with real money.