Hi,
I just enrolled in the course and have watched a couple of videos but I cannot seems to find anything regarding the lookback periods for any of the tests he performs. For example, why does Ernest take 2 years to calculate de Johansen Vectors (or the Hedge Ratios) and not 3 or 4 years (section 3) ? How did he come up with 2 years ? Why does he take 5 days for the bollinger band in the EURCHF strategy (Section 1) ? Why does he take 90 days in section 2 for the Hedge Ratio between GLD and GDX ?
Hi Stephane,
The lookback period is a parameter to be optimized. Typically I just pick a timeframe that is reasonable based on the frequency of our data. But strictly speaking, one has to optimize this in-sample.
Best,
Ernie
Hi Ernest, thanks a lot for taking the time to answer my question. Quick follow up question though: by “optimizing”, I understand that I would use the look back period that will generate the most profit (or best Sharp Ratio) but in the end, doesn’t it become data snooping ? Thanks again,
Hi Stephane, Yes, all optimization invites data snooping. You can verify whether this optimization works out-of-sample by optimizing parameters in the lookback period (in-sample), and testing whether the optimized parameters are still optimal out-of-sample.