Looking for some information on theorycrafting.

How many samples of data do you need to prove something?

From that data how would you figure out your damage range from your stat inputs so you wouldn't have to shoot 500 ruins at a training dummy each time your gear changed?

i.e if ruin 2 has a potency of 80 and I have 438 int and 300 det ruin 2 will do between X-XX damage.

That's all I have for now. I plan to expand as I answer these questions.

Edit: If there is already a resource with this information just point me to it and I'll read it. Thank you.

http://valk.dancing-mad.com/?page_id=179 likely has much of the information you're looking for.

As far as how much data it takes to "prove" a hypothesis if we want to be really formal about it, it's basically impossible to 100% prove something with a finite amount of data. The informal answer is that the number of samples you need to be "sure" about something can vary depending on the complexity of the underlying mechanic that generates your results.

There are ways to see if you need more samples. One way is to take some portion of the data and use it as a holdout set, and then train your model on the remaining data. I can go into more detail if you're interested, but that's the basic idea.