Backtesting & Statistics
A strategy that genuinely makes money looks like a losing one about a quarter of the time over twenty trades. This paper shows you how to build the only benchmark that matters — random entry on your own data — and how many trades you need before a number is a result rather than a mood.
This paper is locked
Read it in the browser and keep the PDF with Premium, or buy this one paper.
Twelve winners out of twenty — is the strategy working, or was that a good fortnight? Almost nobody can tell, because two numbers are missing: what randomness would have produced on the same data, and how many trades it takes before a difference from randomness is visible at all.
Run the identical trade — same stop, same target, same instruments, same bars — but pick the entry at random and the direction by a coin flip, thousands of times. Here is why the result comes out above zero.
Build the baseline on the same instruments, the same period and with the same stop and target. Change any of those and you are comparing your signal with a different bet.
This paper is locked
