how to benchmark for optimal hist len and forward window

Gents, I'm working with 1-min data like from this example. So far I see good results for the same parameters.
https://facebook.github.io/prophet/docs/non-daily_data.html

But is there any reference where I can find an optimal methodology on how to benchmark ideal history period length and forecast window size to get most reliable models? For what I understand, one way would be to just combinatorically test various period sizes and then compare RMSE and similar metrics. May be someone can share his experience or direct me to an article? I've been reading some articles, but most of them don't further than basics.

Thanks.