Multi Strategy EAs

[Isaac Uriel Arenas Caldera]

This topic is created in order to have a guide and discussion about multi strategy EAs.

I have seen this topic appeared in some topics but the discussion tends to die after some comments.

So the intention is to have the topic in here and don't let it die.

[Isaac Uriel Arenas Caldera]

For multi strategy Articles, I have found the following one:

Creating a Multi-Currency Multi-System Expert Advisor - MQL5 Articles

[Isaac Uriel Arenas Caldera]

Have the next questions:

Do you consider an EA that has the same strategy with different params in different pairs a multi strategy system?

How important is correlation?

Should the percentage on each strat should be equal?

[Andrew Pun Magar]

I personally believe that a strategy is suitable for one pair/asset class in certain sessions with parameters that can work wonderfully. So one strategy for one asset class is the way I go. Every pair is different and reacts differently, there are correlation but markets are not always in equilibrium. 

[Smart Algotrade]

On your three questions: Same strategy with different parameters across pairs is closer to parameter diversification than true multi-strategy - it still shares the same underlying edge/failure mode, so if that edge stops working it tends to stop working across all the parameter variants at once. A "true" multi-strategy setup usually means genuinely different entry logic/timeframes/market regimes so the components can fail independently.

Correlation matters more than most people budget for. Even on different pairs, two strategies that both do trend-following on the same session will tend to win and lose together, so you get less diversification benefit than the pair count suggests. It's worth checking correlation of the equity curves/trade timing of the sub-strategies themselves, not just the correlation of the instruments they trade.

Equal allocation isn't usually optimal. If one sub-strategy has much higher volatility or drawdown than another, an equal split lets it dominate total portfolio risk. Sizing each strategy's allocation inversely to its own volatility (a simple risk-parity approach) keeps each one contributing a similar share of risk instead of a similar share of capital.

[Daniel-gheorghe Muresan]

Isaac Uriel Arenas Caldera #:

Have the next questions:

Do you consider an EA that has the same strategy with different params in different pairs a multi strategy system?

How important is correlation?

Should the percentage on each strat should be equal?

1. Using the same strategy with different parameters it means that you use multiple systems.(e.g. Turtles used S1 = 20/10 and S2 = 55/20 on the same pair/asset);

2. The more you diversify the correlation has smaller importance. If you trade only several markets trading AUD USD and NZD USD won't help because these pairs will act almost the same. The first step in diversification is to trade forex, metals, commodities, stocks, crypto.

3. You should risk the same on every trade. No trade, market, strategy is more important than other. Also you should risk small per trade, under 1% from your account balance.

[Andras Feher]

I've been working quite extensively with multi-strategy EAs lately, and I've gradually moved away from trying to create one "perfect" strategy.

What I've found more interesting is combining genuinely different strategies, each with its own entry logic, filters and trade management, while controlling risk at portfolio level.

One of the biggest questions for me is how much diversification is actually enough. Having 10 strategies doesn't necessarily mean much if several of them react to the same market conditions and start losing simultaneously.

Because of that, I'm now paying much more attention to:

- correlation between individual strategy equity curves

- maximum combined drawdown

- performance across different market regimes

- whether one strategy dominates the portfolio

- stability across parameter ranges rather than the single best optimization result

- differences between backtest and live execution

I've also started validating my newer systems live. The live histories are still very young — around one month — so I definitely don't consider that enough data to prove anything. For me, the interesting part will be comparing the live behaviour with the original research after 6–12 months.

For those already running multi-strategy portfolios:

How do you decide when two strategies are too correlated to justify keeping both?

And do you allocate equal risk to each strategy, volatility-adjust the allocation, or dynamically reduce exposure when several strategies begin behaving similarly?