Correlation Cap

Multi-pair EAs (Trendopedia AI across 8 majors, Market Trader AI Pro across 12 instruments) face a structural risk that single-pair EAs don't: pair correlations are not constant. EURUSD, GBPUSD, AUDUSD, NZDUSD, and CHFUSD often trade as independent instruments during normal market conditions but converge to near-perfect correlation during USD-strength episodes. A multi-pair EA that treats these as 5 independent risk allocations during the normal regime can find itself with 5× the intended exposure during the stress regime — exactly when the trader doesn't want that.

Correlation cap implementations:

• **Static per-currency-base caps**: max N pairs sharing the same base or quote currency open simultaneously (e.g. max 3 USD-quote longs at any time) • **Dynamic correlation matrix**: EA computes pairwise correlations over rolling 30-day windows; pairs above 0.7 correlation threshold count toward shared exposure for sizing purposes • **Direction-aware caps**: long USDJPY and short EURJPY are both short-JPY exposures and count as correlated for purposes of yen-related regime stress • **Aggregate volatility cap**: total portfolio VaR (value at risk) limit imposed across all open positions, scaling per-trade sizing down when aggregate VaR approaches the cap • **Beta-to-DXY cap**: limit total exposure that correlates with the dollar index, preventing concentrated dollar-direction bets across the portfolio

The simplest implementation ("max N positions per base currency") is the most common and easiest to audit. The more sophisticated implementations (dynamic correlation matrices, beta-to-DXY caps) provide better protection but require continuous computation and are harder to verify from the buyer's perspective.

Why correlation caps matter for EA selection:

• Multi-pair EAs without explicit correlation caps are NOT diversified during stress regimes — they're concentrated • A 5-pair EA with no cap, running at 1% risk per trade, can have 5% effective risk during USD-strength regimes • The realised drawdown during regime stress can be 2-4× what the published backtest suggests because the backtest period may not have included a coordinated regime episode

For buyer evaluation: confirm the multi-pair EA documents its correlation-management approach. Vendor docs that explicitly describe the correlation cap (mechanism, threshold, cap size) signal serious risk engineering. Vendors that don't address correlation at all are either (a) running un-capped multi-pair exposure that creates regime stress vulnerability or (b) hand-waving on a topic they don't understand. Either way, the absence of correlation cap documentation is a buyer red flag for multi-pair systems.