I haven’t actually finished any of my WiPs yet, but I’m happy to share my system, such as it is.
I’ve discovered that, however good my writing may or may not be, my ability to write is much more advanced than my game design skills. I recently scrapped an incredible amount of words because I had been writing a game with a poor design, and I’m now rebuilding that game from scratch.
So, instead of using a loose plot outline as I’ve done for previous works of fiction, I’ve taken to running a really specific outline for the entire game, including major plot points, pivotal choices, end states, and a rough idea of how we get from the choices to the end states.
My two favourite resources currently for helping to figure this all out are the CoG full game design guidelines, which helped me figured out a lot of the flaws in my game design choices, and also this blog article, which talks about writing plots in general.
I like to use paper for charts and names and that sort of thing, but I like to plot where I’m typing because I can adjust and fix as I go along, or expand.
Once I’m actually coding, I follow this article’s advice (same blog, different article) about writing basically what you’re going to write. Then I do the actual narrative, stats, and coding all together. (For those who’ve asked me how I wrote 120,000 words in a month for my CS Comp entry earlier this year, that was a good part of how.)
And I don’t write scenes in order. One reason is some older advice in a CoG blog article. (I’d be curious to see if that approach is still recommended, too.) I also find my first chapter is more useful at presenting the rest of the game when some of that ‘rest of the game’ is already finished.
My first rule of thumb now, though, is worrying about game design before I think about narrative. With that CS Comp piece, I made the mistake of whirling through without a finished design for the game, hence the current dismantling. There were parts of the game design that weren’t working, no matter how much editing I put into fixing the problems. The current method seems to be a lot more effective thus far. We’ll see!