Do we have any particular documentation or examples around loss?

Hello Nixtla,
I am testing the Time GPT model for monthly price prediction, for three months ahead on stocks.
 And I'm looking for a way to do basic research on the quality of prediction and measuring LOSS, but I can't find documentation that helps me.
 I saw the "Fine-tuning with a specific loss function" section, but without basic research how can you determine which loss function to use? 
I would like to build a presentation that shows the priority of Time GPT over other models using BackTest and historical LOSS measurement. 

Please help.

Ofir 

my user name - 
<mailto:orcafm@gmail.com|orcafm@gmail.com>

this is the link of the current docu: https://docs.nixtla.io/docs/fine_tuning_with_a_specific_loss_function We can also mention them that we are improving it and will be live next week