During development and training of a strategy that uses several Algos and Assets, I set Hedging = 2. I am now trying to tune the money management part of the strategy. I am US based and will need to set the NFA flag. I am surprised (and sad) to see that changing Hedging to 4 or 5 nullifies the strategy. A somewhat steady profit (R2=.6) is turned into a loss.

I understand that system performance could be affected by the Hedging setting. But based on http://zorro-trader.com/manual/en/hedge.htm I expected the 2 equity curves to be "relatively similar".

What could be going on here?

Probably easier to open an account in a country that allows hedging. Australia?

I am a patriot.

Hard to say what might be going on with your strategy. Can you post it, or at least give us the general idea of how it works? Any intra-bar trades that could be affected by the TICKS flag?

I also think it's the intrabar trades - if they have a large effect on the result, the hedging simulation can be very inaccurate. You can find details under "Remarks" in the hedge description.

In such a case switch off virtual hedging for the test.

"In such a case switch off virtual hedging for the test."

I understand from the manual that there is a known risk of performance difference when using Hedge>=4 over Hedge 2. http://zorro-trader.com/manual/en/hedge.htm

If I use Hedge 2 in the backtest, then change to hedge 4/5 for Trade mode, at what point will I measure this risk? I guess I will just blindly realize the risk in live trading. That is not a way to get good sleep at night.

Alternatively, can Zorro's backtest for Hedge 4/5 be made accurate?

Yes, it can, but it is not worth the effort. So I guess you must live and sleep with the risk.

Reason for the inaccuracy is testing speed. Every portfolio component is simulated separately, which leads to different intrabar opening and closing of pool trades as in real trading. This could of course be changed, but then testing would be much slower - and we didn't want to implement two different simulation modes just for this minimal issue. The advantage of accurate simulating virtual hedging in the backtest is normally negligible, compared to other simulation inaccuracy issues. In training mode virtual hedging is not used anyway.