Why does sample size dominate everything?
Sample size dominates because trading results are noisy enough that short runs carry almost no information. The spread of outcomes across a small number of trades is wide enough that a strategy with no edge whatsoever routinely produces long profitable stretches, and a genuinely good strategy routinely produces long losing ones.
The question is never whether you are up. It is whether being up this much, over this many trades, is unlikely enough under the assumption that no edge exists. Those are different questions, and only the second one has an answer worth acting on.
This is also why the trader's instinct is systematically wrong here. Human pattern recognition is tuned to find structure in small samples, which is useful for spotting predators and useless for evaluating strategies. Every trader has experienced a run that felt like confirmation and was not, and the feeling is identical either way.
How many trades do you need to know if a strategy works?
The number of trades you need depends entirely on the size of the edge relative to its variability, and the relationship is unforgiving: halving the edge roughly quadruples the trades required.
A strategy averaging 0.5R per trade with a standard deviation of 1R needs far fewer trades than one averaging 0.05R with the same spread. In the first case a couple of hundred trades might settle it. In the second, you may need tens of thousands — which is more trades than most retail traders will take in a career.
That has a hard practical implication people rarely draw. If the number of trades required to confirm an edge exceeds what you can realistically accumulate, then the strategy cannot be proven by forward testing in any useful timeframe. The honest response is either to accept it on other grounds — mechanism, out-of-sample backtest evidence across many instruments — or to look for a larger effect.
Backtesting exists largely to escape this constraint. Ten years of history across thirty markets accumulates the sample a live account never could, which is the whole reason to test at all rather than just trading and watching.
What does statistical significance mean for a trading strategy?
Statistical significance means the observed result would be unlikely if the strategy had no edge — nothing more, and specifically not that the strategy will keep working.
It is a statement about one direction only. A significant result says luck is an uncomfortable explanation. It does not say the effect will persist, that it is large enough to be worth trading after costs, or that the market will keep behaving the way it did.
There is also a trap specific to this domain. Significance testing assumes you asked one question. If you tested forty strategies and are reporting the one that came out significant, the calculation no longer means what it says — roughly one in twenty will clear a 5% threshold by chance alone. Multiple-comparison corrections exist precisely for this and are almost never applied outside academia.
The practical version, stripped of the vocabulary: be much more sceptical of a good result that emerged from a search than of one you predicted in advance.
What should you do when a result is inconclusive?
When a result is inconclusive, the right response is to size accordingly and keep gathering evidence — not to keep trading at full size and hope the sample catches up.
Most results are inconclusive, and that is the useful information. A tool telling you your result is indistinguishable from variance has saved you from scaling into something that was never there, which is worth considerably more than the confirmation you were hoping for.
There are three reasonable moves. Trade it small while the sample accumulates, accepting that you are paying for information. Look for a larger effect that can be confirmed in a realistic number of trades. Or test the same idea across more instruments and periods, which grows the sample without waiting.
The unreasonable move is treating an inconclusive result as a soft yes. Inconclusive means the evidence does not distinguish your strategy from a coin, and coins do not become edges by being traded confidently.
Does a losing streak mean my strategy stopped working?
A losing streak almost never means a strategy stopped working, because streaks of a length that feel alarming are entirely ordinary for strategies with a genuine edge.
A system with a 45% win rate will hit eight consecutive losses reasonably often across a few hundred trades. It feels like a regime change from the inside and looks completely unremarkable from the outside, which is a large part of why traders abandon working strategies.
Distinguishing genuine decay from ordinary variance requires comparing recent results against the distribution you would expect, not against how the drawdown feels. If your expected worst streak is nine and you are eight in, nothing has happened yet.
Real decay does occur — edges get competed away, market structure changes, a broker's execution shifts. But the evidence for decay has to be a departure from the expected distribution, measured, rather than the observation that recent trading has been unpleasant.