Correlate each period's return with the previous one. Positive serial correlation means winning runs and losing runs cluster; negative means results alternate more than chance would produce.
It matters in two ways. Positive autocorrelation makes drawdowns deeper and longer than independence-based estimates predict, so a monte-carlo-reshuffle that assumes independence will understate the risk. It also inflates ratios: annualising a monthly Sharpe by multiplying by the square root of 12 assumes independence, and positively correlated returns make that conversion too generous.
Strongly positive serial correlation in a reported record is also a red flag for smoothing - illiquid positions marked slowly, or losses recognised late. Genuine liquid trading records show correlations close to zero.
Related: monte-carlo-reshuffle, sharpe-inflation, stat-stability, drawdown-duration