It's time to separate signal from noise-by subjecting our parameters to rigorous stress tests, we identify the most robust and reliable ones. This step filters out the weak, ensuring only the strongest parameters make the cut. Through comprehensive backtesting across various conditions, we gain valuable insights into their real-world performance.
The structured robustness is carried out by ranking a stock universe based on the chosen parameter, dividing it into different equally sized slices (e.g., terciles, quartiles, quintiles), and analyzing the performance of these slices over various timeframes. Ensuring that an equal number of stocks are distributed across slices is imperative.
The rationale behind this approach is simple: if a parameter effectively contributes to performance, the highest-ranked group should generate better returns than the group below it, which in turn should outperform the next one, and so forth. Additionally, volatility should increase as one moves towards the lower-ranked slices.