I have read from more than one source that it's reasonable to expect MaxDD in real trading to be twice of what you find in BackTesting. Disregarding the fact that MaxDD is highly path-dependent (another topic),

- Have you found this to be the case?
- What measure do you use to shut down a system?

To me, if I saw a live MaxDD that is twice what I found in testing, I think I would shut it down. It's almost implying that I've either curve-fitted too much or not tested on a sufficient dataset (not enough stocks or not long enough).

Testing over a twenty year time-frame, it is shocking to find many systems that work for 10-15 years and then crapping out - maybe due to a bull-market bias in US equities. In that respect, at least the current bear market gives us a more "interesting" data to play with...

What do you think?

- Ken