How do you handle drawdown rules across a portfolio of EAs (prop firm style)?

New comment
 

I've been running several EAs together on the same account instead of one single strategy, mainly to smooth the equity curve, but it's forcing me to rethink risk management completely. A single trailing stop or fixed lot size per EA doesn't work anymore, because the drawdown that actually matters is the combined one, not each EA's individual number.

Right now I'm doing something fairly simple: each EA reports its floating result to a shared variable/file, and a master risk block starts cutting position sizes proportionally once the aggregated equity drawdown crosses a threshold (testing 5% and 8% steps). It works, but it feels crude, especially when trying to respect daily loss limits like the ones prop firms impose.

For those running multi-strategy portfolios: how are you measuring correlation between EAs in real time, not just from backtest reports? And once combined drawdown gets close to the limit, do you cut size, pause the weakest EA, or something else? Curious what's actually working for people trading this way instead of a single EA.

 
Cristian David Castillo Arrieta:
[O]nce combined drawdown gets close to the limit, do you cut size, pause the weakest EA, or something else?

I suggest calculating your average consecutive loss count (as a quantity of losing trades). The Tester can readily spit that number out but if you don't like the Tester, you're left with real-time tracking and calculation. Then when that threshold average is exceeded, reduce all of your next trades' position sizes to something like 0.01. When you get back in the black (profit) and only when you get back in the black, increase all of your next trades' position sizes back up to the usual sizes.

You could write some code to simulate trades and stay flat after exceeding your average consecutive loss count, but then you have to calculate your average spread, slippage, etc. which will never be 100% as accurate as live trading a reduced position size.

 
Cristian David Castillo Arrieta:

I've been running several EAs together on the same account instead of one single strategy, mainly to smooth the equity curve, but it's forcing me to rethink risk management completely. A single trailing stop or fixed lot size per EA doesn't work anymore, because the drawdown that actually matters is the combined one, not each EA's individual number.

Right now I'm doing something fairly simple: each EA reports its floating result to a shared variable/file, and a master risk block starts cutting position sizes proportionally once the aggregated equity drawdown crosses a threshold (testing 5% and 8% steps). It works, but it feels crude, especially when trying to respect daily loss limits like the ones prop firms impose.

For those running multi-strategy portfolios: how are you measuring correlation between EAs in real time, not just from backtest reports? And once combined drawdown gets close to the limit, do you cut size, pause the weakest EA, or something else? Curious what's actually working for people trading this way instead of a single EA.

Please take a look at this article: https://www.mql5.com/en/articles/21720

Basically it describes a reliable infrastructure to implement risk Management for a portfolio of EAs. I hope it helps.
 
I've been dealing with exactly this problem while building multi-strategy systems.

For me, the key is to stop thinking about risk at individual EA level once several strategies are running on the same account. What matters is the combined exposure and how the strategies behave together during adverse market conditions.

I prefer reducing portfolio exposure before reaching the hard drawdown limit rather than waiting until the limit is almost hit.

Another thing I think is important is correlation. Two strategies can look completely different logically but still produce losses at roughly the same time. In that case the diversification is much smaller than it appears.

I've increasingly moved toward treating the whole system as a portfolio: individual strategy risk + combined exposure + correlation + account-level drawdown protection.

I'm currently validating my newer multi-strategy systems live as well. The public histories are still only around a month old, so it's far too early to draw long-term conclusions, but comparing actual portfolio behaviour against the backtests has already been very useful.
 
keep Capital more and lot size small for all ea , if it all loss so all profit also, and try mix pairs , never put all egg in one basket 
 
Cristian David Castillo Arrieta:

I've been running several EAs together on the same account instead of one single strategy, mainly to smooth the equity curve, but it's forcing me to rethink risk management completely. A single trailing stop or fixed lot size per EA doesn't work anymore, because the drawdown that actually matters is the combined one, not each EA's individual number.

Right now I'm doing something fairly simple: each EA reports its floating result to a shared variable/file, and a master risk block starts cutting position sizes proportionally once the aggregated equity drawdown crosses a threshold (testing 5% and 8% steps). It works, but it feels crude, especially when trying to respect daily loss limits like the ones prop firms impose.

For those running multi-strategy portfolios: how are you measuring correlation between EAs in real time, not just from backtest reports? And once combined drawdown gets close to the limit, do you cut size, pause the weakest EA, or something else? Curious what's actually working for people trading this way instead of a single EA.

Approach in this context varies; there is no absolute right or wrong...

One solution, from my perspective, is to treat the system as a single combined portfolio. Assuming the maximum daily loss you are willing to tolerate in the worst-case scenario is 5%, and you have 5 strategies or Expert Advisors (EAs), you can set a maximum risk limit of 1% per EA. Whichever EA exhausts its allocated daily limit is paused for the remainder of the day. It is a very simple principle: the failing one steps out, and no other EA bears the burden of its losses—risks are strictly isolated, meaning EA 2 is not expected to compensate for EA 1's loss, and vice versa.

However, if an EA proves to be consistently unprofitable—let's assume EA 3—you should halt its operation, remove it from the pool, and redistribute its risk allocation equally among the remaining EAs. Alternatively, you can allocate its share to a profitable EA experiencing a strong performance streak to capitalize on that momentum. Rather than allowing capital to drain through a losing EA, pausing it and reallocating its share—or spreading it across the remaining EAs—reduces overall market exposure and optimizes risk distribution.

 

I wouldn’t pause the weakest EA based only on recent P/L; that can disable it just before its market regime returns. I’d calculate rolling correlation from each EA’s hourly equity returns, group correlated EAs into one risk bucket, and enforce the daily-loss limit at account level. As drawdown rises, reduce or block new entries first. Force-close positions only at a predefined hard limit.

New comment