Hello all,

If a model works better in anchored WF rather than on rolling WF - meaning that in rolling WF it generates also profits but drawdowns are worse - what does it tell us about the model? is it bad and we should reject it outright or not?

Thank you

If anchored works better, it means that the oldest data is still significant for the model. As time rolls on with anchored, the old data will still have an impact, but it will be lessened by the addition of new recent data. But it also means that the new recent data will not have as much impact, and the model will not be as responsive to market changes.

Is this good or bad? The answer is really "maybe." I personally always use unanchored, and some of those strategies work great with unanchored, but look terrible with anchored (the opposite of your question, but the point is still the same). Yet I still trade them.

There is no rule that says a model has to perform well with both types of walkforward. You might decide to make that a rule, if your experience tells you it matters. My experience says it does not matter, but others could easily have the opposite experience.