Hello all,
When performing backtesting constant USD seems that is the right approach, rather than using constant shares/contracts count as price varies. However, is it appropriate to add net profits to the amount of initial capital applied to the strategy or leave it out and use the same initial capital amount through the backtesting? The results can vary substantially depending on the type of a strategy one is testing (profits follow profits or profits follow loss). Any suggestions, references to the books/articles that talk about it are welcomed.

hi there alko

I don't know the answer, but I can think of someone who may be able to help.


@kevinkdog wrote a book on building algo systems and can probably offer his opinion about your query.

Hello all,
When performing backtesting constant USD seems that is the right approach, rather than using constant shares/contracts count as price varies. However, is it appropriate to add net profits to the amount of initial capital applied to the strategy or leave it out and use the same initial capital amount through the backtesting? The results can vary substantially depending on the type of a strategy one is testing (profits follow profits or profits follow loss). Any suggestions, references to the books/articles that talk about it are welcomed.

Thanks for the mention @xplorer

When I test, I always test with 1 contract. The reason is that I want to see what the strategy does over the whole period with that 1 contract. You can test with position sizing built in, but it may be hard then to distinguish where the performance comes from - is it from the actual trading edge, or maybe the edge has eroded over time, but that is masked because you are trading many more contracts (from reinvesting profits)? Sometimes it is hard to tell whiat is occurring in a combination curve.

So, when I test, it is with one contract. If that strategy passes my tests, then I look at adding it to my portfolio, and incorporate position sizing at that point.

But, you could make an argument for the flip side (testing with position sizing included), as my friend (and champion trader Andrea Unger) does. His point is that a good strategy with one contract might not be as good with position sizing as a weaker one contract strategy. For example, take two systems that average nearly the same profit over time, with "A" a bit better. But, "B" has small stop losses, and "A" has large stop losses. When you add in position sizing, "B" might be better because it will likely perform better with position sizing techniques.

Hope this helps!