Why the usual answers are both wrong
"You're up, so you have an edge." A trader with no edge whatsoever will still finish a fifty-trade stretch in profit roughly half the time. Being up says almost nothing on its own; it is exactly what randomness looks like from the inside.
"Markets are random, it's all luck." Equally empty. It is a claim about you made without looking at a single one of your numbers.
The version of the question that can be settled is narrower and more useful: is your average trade far enough above zero that a run of good luck is an implausible explanation? That has a standard answer, and it needs only what your statement already contains.
The test
mean = your average trade, after costs
sd = how much individual trades vary around it
se = sd ÷ √n
95% range for your true average = mean ± 1.96 × se
If that range sits entirely above zero, this sample does not look like luck. If it straddles zero, your results are consistent with having an edge and consistent with not having one — and no amount of staring at the equity curve resolves that.
The number this page exists for
When the answer is "undecided", most tools stop there. That is the least useful place to stop, because it sounds like a verdict on you when it is a fact about your sample size.
So the headline is the sample size that would settle it, at your own mean and your own variability:
trades needed = (1.96 × sd ÷ mean)²
The shape of that formula is the whole lesson. It squares the ratio between how much your trades scatter and how much you make on average. Halve your average edge and you need four times the trades. Trade an instrument twice as volatile for the same average gain and you need four times as many again.
This is why a scalper with a small consistent edge can demonstrate skill in a few hundred trades, while a swing trader with the same annual return may need more trades than a career contains. Neither is better. They are simply different distances from an answer.
What a "yes" here is not
A significant result describes trades already taken. It is not a forecast, and three specific things sit outside what it can see:
- Continuation. The test says these trades were unlikely to come from nothing. It says nothing about the next hundred.
- A changed method. If you switched markets or rules halfway through the file, the average blends two different traders, and the interval around a blend is not meaningful for either.
- Why you are looking. You opened this page because this history interested you. Someone who tries ten strategies and tests the one that worked will find significance in it about as often as chance provides — and this page will agree with them about that one, because it only ever sees the file it is given.
None of that makes the number useless. It makes it a floor: if your sample cannot clear this bar, the more optimistic questions are certainly premature.
Frequently asked
My range crosses zero. Does that mean my strategy does not work?
No, and the wording on the page is deliberate about this. It means this sample cannot distinguish between working and not working. Those are different statements, and the difference is the entire point — "unproven" is not "disproven".
Why 95% and not something else?
It is the conventional threshold, which makes the result comparable to how the same test is reported everywhere else. It is a convention rather than a law of nature: at 90% the bar is lower and the required sample smaller.
Are the figures after costs?
Yes. Commission and swap come off each trade before the average and the spread are computed. Testing a gross edge would answer a question about a market you cannot actually trade in.
The required number is enormous. Is that a bug?
Usually not. A large figure means your trade-to-trade variation is large relative to your average gain, which is the ordinary condition of most discretionary trading. It is the honest arithmetic of your own distribution, and it is the reason why "I know it works, I've been doing it for months" is such a hard claim to support.