Deflating a backtest

You test 200 variants of a Trend-Template breakout system on four years of daily data. The best has an annual Sharpe ratio of 1.6, skewness $-0.3$, kurtosis 5. Across the 200 trials the annualised Sharpe ratios have standard deviation 0.5. (a) Compute $S_0$ (per day and annualised) and the DSR. (b) Compare the cross-trial dispersion with what pure noise would produce over four years. What does that suggest? (c) Apply the McLean–Pontiff 58% post-publication haircut to the best Sharpe ratio. What annual Sharpe ratio would you plan for?