Question regarding possible owners vs preferred owners

I'm having trouble differentiating when to use the 2.

I know that possible owners is used in 2-node clustering.
I know that you use preferred owner for failback in 3+ node clustering.

When do you use possible vs preferred in 3+ node clustering? Why would you only want to use possible vs only using preferred and in what situations would you mix the 2?

Any help is much appreciated. Thanks.