Running multiple EAs: why five "safe" bots can breach one prop-firm drawdown cap

Tóm tắt — Five XAUUSD builds in our own tests each hold relative drawdown under 10%, from 3.84% to 6.77%. Run together, their worst-case combined drawdown is 28.86% — nearly three times the cap each one passes alone. Correlation lives in the instrument, not the strategy name. Two comparison tables, the honest measurement range, and how to size a multi-EA portfolio.

Running multiple EAs: why five "safe" bots can breach one prop-firm drawdown cap

TL;DR — Running more EAs only reduces risk if they trade different instruments. If they share one symbol, you have not diversified — you have leveraged a single bet.

• Five of our profitable XAUUSD builds each hold relative drawdown under 10%, from 3.84% to 6.77%.
• Run together, their worst-case combined drawdown is 28.86% — nearly three times a typical prop-firm cap that each one passes alone.
• Prop firms measure account equity, not per-EA performance. Five compliant EAs can still produce one non-compliant account.
• The fix is to budget the account's drawdown cap first, then divide it across EAs — not to size each EA as if it traded alone.

Does running more EAs reduce risk?

Only if the EAs lose money at different times. Diversification is not a count of strategies — it is a lack of correlation between them. Two EAs with different names, indicators and timeframes that both trade XAUUSD will still take their worst losses in the same week, because the thing that hurts them is the same price series.

That matters most on funded accounts, where one account-level breach ends the challenge no matter how sensibly each EA was configured on its own.

Why do five EAs that each pass a 10% cap fail together?

Because drawdowns can add up, and each EA's tested drawdown assumed it was the only strategy on the account. Here are the five profitable XAUUSD builds from our own published tests that individually sit under a 10% relative-drawdown limit:

Every row passes a 10% cap. The total does not. Even the two calmest builds — gold-ny-momentum and News Straddle — sum to 9.55%, which sits on the cap with nothing left over.

Disclosure, because it matters: those are full-window figures from 2021 to 2026. Under our current near-window listing gate, only one of these five builds (swing-breakout-3m) is on sale today. Two record zero trades in the most recent year — a known execution issue we are still fixing — one is negative on a four-trade sample, and one has only eleven recent trades. We are using them here as a worked example of correlation arithmetic, not as a recommended basket. Treat this table as math, not a shopping list, and check current status on /store.

What is the honest range for a combined drawdown?

Somewhere between 6.77% and 28.86%, and we have not measured it. That is the important admission. The true figure depends on whether the five equity curves dip at the same moment:

• If the drawdowns were perfectly offsetting, the portfolio drawdown would be the largest single one: 6.77%.
• If they were perfectly synchronised, it would be the arithmetic sum: 28.86%.

Because all five trade the same instrument, the realistic answer sits far closer to the sum. We publish per-EA backtests on /proof but no combined-portfolio equity curve — so nobody, including us, should quote a precise portfolio drawdown for this basket. Assume the pessimistic end until a combined test exists.

How correlated are EAs really? (look at the symbol, not the strategy name)

Correlation lives in the instrument. Grouping every build that was profitable over its tested window by symbol shows how lopsided a "buy the top performers" basket becomes:

Ranking EAs by profit factor and buying the top four hands you four gold EAs. Ranking them by instrument first is what actually spreads risk. You can filter by symbol on /store and line builds up side by side on /compare.

How should I size a multi-EA portfolio?

Start from the account limit and divide down. If the cap is 10% and you plan to run four EAs, each gets a 2.5% drawdown budget — not the 6% it was tested at alone. In practice that means cutting each EA's risk-per-trade percentage in proportion, using the percent-risk logic in EA position sizing. Two rules follow:

1. One EA per instrument before adding a second EA anywhere.
2. Add EAs by subtraction — when a new EA joins, the existing ones must shrink, because the cap did not grow.

What do prop firms actually measure?

Account equity, on a single curve, in real time. No funded-account rule set evaluates EAs individually. Whether a breach came from one EA or five at once is irrelevant — daily-loss and maximum-drawdown rules read the aggregate. Deep combined drawdowns are also disproportionately hard to recover from: see drawdown recovery math.

FAQ

How many EAs should I run on one account? Start with one per instrument and no more than three or four total on a funded account. Beyond that, the per-EA drawdown budget becomes so small that each EA trades at a size where its edge barely covers costs.

Do different timeframes make two EAs on the same pair uncorrelated? No. Timeframe changes trade frequency, not the underlying price series. Our five gold builds span M15 to H4 and still share every gold shock.

Can I just pick the EA with the lowest drawdown five times over? That is exactly the trap this article documents. Five builds at 3.84% to 6.77% drawdown sum to 28.86%. A portfolio of individually compliant EAs is not a compliant portfolio — the account has one drawdown limit, and correlated EAs spend it simultaneously.

Risk caveat

All figures come from our own historical backtests over the windows published on /proof and are not forecasts. Historical performance does not represent future results. Backtested drawdown is routinely exceeded live, so plan for at least 1.5 times the tested figure and read the risk disclosure before allocating capital. Nothing here is financial advice or a promise of profit. Traders sourcing EAs for client accounts should review the terms on /partners.