Hi. Basic questions about reTraining a running strategy:

Should reTraining avoid Sun, Mon, Wed, Fri for the same reasons Download should?

The description of the Z Systems contains recommended reTraining intervals. How are these determined? How should we determine the best reTraining interval for our own strategies?

How can we determine how sensitive a strategy is to reTraining interval? I.e., similar to optimizing, might reTraining interval be a sharp peak, and being off by 1% either way cause significant drop-off? Or can we determine if it's a broad peak, and anywhere from (say) .5x to 2x of the recommended interval is OK?

Similarly, and back to the Z Systems, if instead of the recommended intervals we reTrained every month what would the expected result be? Would it be different for each Z based on the different algorithms employed? Are there "rules of thumb" or other such based on type of algorithm(s) employed in the strategy?

Thanks.

I believe this already is discussed in the manual under "retraining" on the trading page. It does not matter on which day of the week you train. But you need a server connection if your history is not up to date.

Thanks jcl. Yes, the Re-Training section covers much - guess I blew through it too fast the first time... Sorry! So much to learn...

The manual doesn't really discuss the sensitivity aspect, though. Is there a way to determine the "peaked-ness" of the reTraining interval value?

The manual says reTraining too often "can cause a slight reduction in profit". Is this a kind of overfitting bias?

Thanks.

No, it's due to the mentioned continuation of series and trades with different parameters. This is however theory. We have not made experiments how this really affects profit.

Hmmm... Thinking aloud... If it's not introducing additional bias, Would a "slight reduction in profit" be a valid "insurance payment" for possibly earlier catching & adjusting to a changed trading environment?