Every time you get an LG problem wrong, redo the whole game 5 times like JY said. LG is the most memorizable. With that bruteforce method, I got down to -0/-1. That is the quickest, most efficient way I think you can do it.
I would keep watching the videos and redoing the games until you get better at making worlds. Sometimes you will need to bruteforce, but the more you make worlds, the more you will become familiar with inferences, which will help you in the long run.