Expected longest streak grows with sample size. A rough estimate for the longest run of losses in N trades at loss rate q is log(N) / log(1/q). With 500 trades and a 55% loss rate, that is log(500) / log(1.818) = 6.2 / 0.598 ≈ 10 consecutive losses.
Ten in a row at 1% risk each is roughly a 9.6% drawdown arriving with no warning, purely from ordinary variance in a profitable system. Anyone who has not planned for it will conclude the edge is gone at trade seven, which is exactly the point at which the sample says nothing of the kind.
Plan the response in advance: a drawdown-throttle that cuts size, a review trigger that checks process compliance rather than results, and a pre-written statement of what would actually constitute evidence of a broken strategy. See streak-probability.
Related: streak-probability, loss-rate, drawdown-throttle, risk-of-drawdown