Placebo control
A placebo control tests a strategy against a version of itself with random entry timing but identical risk, costs and exit rules. If the random version performs as well, the entry signal is not doing anything.
In plain terms
A strategy is more than its entry signal. Position sizing, stop placement, trailing logic and exit timing all contribute to the result — sometimes all of it.
It is entirely possible to build a strategy with a meaningless entry rule and a genuinely good exit structure, backtest it, see a profit, and conclude that the entry rule works. The placebo control separates the two. Keep everything except the entry signal, replace that with random timing, and see what happens. If the result barely changes, the signal you spent weeks refining is decoration.
Why it matters
Borrowed directly from clinical trials, and for the same reason: the question is never "did the patient improve" but "did the patient improve more than they would have without the treatment". Most retail backtesting never asks the second question.
How Atlas Edge handles it
Every rigor check includes a Monte Carlo randomised-entry comparison, run as five seeded draws within each walk-forward window. The draws are seeded, which means the result is deterministic and reproducible — re-run the same test and you get the same answer, rather than a number that shifts each time.
The methodology is disclosed inline in every verdict. You can see exactly what was compared and how, rather than being handed a score and asked to trust it.
Answers to the usual ones
- What if my strategy beats the placebo only slightly?
- Then the entry signal is contributing only slightly, and the honest conclusion is that most of the performance is coming from somewhere else.
- Why randomise entries rather than exits?
- Because the entry signal is usually the thing being claimed as the edge. Randomising exits answers a different and generally less interesting question.