Combining Multiple Pairs

New comment
 

For the past weeks I've been developing an EA based on SMC, multitimeframe liqquidity-sweep:

Sweep

Stats are good IMO, DD < 10%, good sharpe ratio, etc...

But I question myself about this being more stable? Such as diversifying.

My main problem is choosing a variety of pairs, i.e:

  • EUR/USD
  • GPB/USD
  • USD/JPY
  • XAU/USD

But I'm not sure how can I choose a better or more "long-termed low-dd" combination of pairs without backtesting a lot (for the computational cost)

Any ideas about this?

 
There are tools that you can place on a chart and measure the ATR for a lengthy period. That may solve the problem.
 

Do you have an example?

Like checking how the ATR behaves through different periods of time and pairs?

 

In my opinion, the choice of partners should be based on the logic of the strategy itself rather than diversification alone.

For example, if the strategy is range-based, it is best to focus on currency pairs with relatively stable and moderate ATR values, such as the EUR/USD pair.

 
Hmmmmm, haven't thinkf of that, I'll try it.
Abdulrahman Saad #:

In my opinion, the choice of partners should be based on the logic of the strategy itself rather than diversification alone.

For example, if the strategy is range-based, it is best to focus on currency pairs with relatively stable and moderate ATR values, such as the EUR/USD pair.

 

Also maybe currency strength?

You need pairs with enough liquidity, so just for validation. Not sure if that might help.

 
Adrian #:

Also maybe currency strength?

You need pairs with enough liquidity, so just for validation. Not sure if that might help.

Yes, assuming that you have a way to measure actual liquidity. In the realm of FX and CFD's, it's tough.

The "least worst" way that I've found is using CME futures volume and O/I as a proxy of sorts─for the majors anyway. The CME publishes delayed data on its website for free.

 

Abdulrahman's point (pick pairs from the strategy logic, not diversification for its own sake) and Adrian's currency-strength angle are pointing at the same thing, and I think it's the core of your question: diversification only lowers drawdown if the return streams are actually de-correlated. Adding pairs that move together doesn't diversify, it leverages the same bet.

Look at your list - EURUSD, GBPUSD, USDJPY, XAUUSD. Those aren't four independent markets. Long EURUSD and long GBPUSD are both short-USD; USDJPY is the mirror; XAUUSD is heavily USD-driven too. When the dollar trends hard, a single SMC/liquidity engine tends to fire the same direction across all of them in the same sessions, so instead of averaging out, the drawdowns line up and compound - the opposite of what you want from a basket.

You can check this without paying the multi-pair backtest cost you're worried about. You already have the single-pair backtests - export the trade-by-trade PnL (or daily equity change) for each and compute pairwise correlation of those streams. That's almost free, and it tells you which pairs add independent edge and which just multiply one exposure. Choose the combination that minimizes average pairwise correlation, not the one with the best individual Sharpe.

And since your signal is one engine across pairs, decompose risk by currency rather than by symbol: track net USD (and net Gold) exposure across the open book and cap that. A currency-strength view is a cheap proxy for it, which is where Adrian's idea earns its place. Ryan's CME volume/OI proxy is a good add for the liquidity-filter half of the question.

New comment