9.6 Pseudo Out of Sample Forecasts

[@gianirobruno4511]

could you please link an article about this topic?

[@mathieutorchia178]

if someone were to make 1000 models and test them using this method, wouldn't it be wrong to pick the best performing model? Isn't that the same as if I told 100 people to flip a coin, and that I would take the "best heads flipper" and make him compete at the World competition of coin flipping? In this case, the 100 people are the 100 models. there's bound to be one or two models that do an AMAZING job if I starting testing hundreds of models.