I'm analyzing simulated portfolios generated using the stationary bootstrap method proposed by Politis et al. (1994). This method is expected to be robust to the choice of average block size, as it employs a random block with mean m-months. However, I've observed that the selection of block size significantly influences which portfolio appears to outperform.
Specifically, when using a block size of m=5, Portfolio 1 demonstrates superior performance, whereas with a block size of m = 60 Portfolio 2 appears to outperform. This discrepancy raises questions about how to interpret these contrasting outcomes.
Given the absence of an optimal block length selection for two time-series, I'm seeking guidance on how to interpret these results?
