Sample size is the number of trades (or observations) behind a performance statistic, and it determines how much that statistic can be trusted. Win rate, profit factor, and expectancy computed over ten trades are mostly luck; over hundreds of trades they begin to describe the strategy itself.
The math is humbling. A strategy with a true 50% win rate, traded ten times, will show seven or more winners roughly 17% of the time — pure chance that looks exactly like an edge. This is why a hot week proves nothing and why judging a system on a handful of trades is the most common analytical mistake in retail trading.
Sample size is also the practical case for backtesting and bar replay: waiting for live setups might produce a few trades a week, while replaying historical sessions can generate a reviewable sample of the same setup in days instead of months.