step 1: I play random scenarios in my head (imagine). Fast. Not always reliable.
step 2: Represent the game*. Pretty much any visible medium on hand can be a go-to. You can create hard evidence something works. It is slow.
step 3: Experiment**. Determine what really works by asking questions, testing on the physical game, and/or imagine some more. What is boring? What parts are engaging? Always ask 'why'?
The type of experimenting that I mean is tweaking behavior to outrageous extents, altering anything important to see results, go as far as necessary to see what breaks an idea, making it unappealing, and then trying to bring it back to discover any nuances.
Recur until game.
*step2 requires a healthy dose of step1, unless you have a physical object that is for all intents and purposes your subject, which may imply the scientific method is involved during step3.
**The actual experiment would involve more creativity, requiring steps 1 and 2.
The Scientific Method applies during step3 when your idea coincides with real physics or anything else that's real. In which case you will spend more time on step2 than if someone were to say, try to figure out what a world made of strawberry jam would be like.