Reply to thread

lindsayf,

Don't confuse validation of a sample with the journey of trading. 100 trades - depending on variables, will probably be enough sample to validate a positive expectancy. But that is not what I'm saying. Probability Theory, not backtesting, will dictate a losing streak. How ever you choose to back test and use Monte Carlo, forward test, back test, in sample, out sample, blah blah blah, the end result is probability theory is the absolute answer that should be anticipated. It may never occur but being in readiness for it will allow you to be realistic.

Many people on this forum promote in sample/ out sample testing. I think that's all bollocks.

1. Understand WHY your system makes money
2. Understand, via probability theory, what can really go wrong

Then you have realistic outcomes to consider.

In sample/out sample is rubbish unless you need to rely on some form of optimization or curve fit to make a method work. To me that's a recipe for failure.

But, getting back to the question at hand, a sample of 100 trades may suggest 10 losses in a row, but probability theory will suggest 21 is possible. You need to go with expecting 21.

I'm not sure if this makes sense. Happy to try to clarify better.