If you at any point become interested in Mike Cook’s Next Level: Making Games That Make Themselves, skip to the end where I tell you how I read this book for free, then go read it. I am confident you and the author would approve of my method.
Next Level is about randomness. Specifically, it’s about the variety of ways developers program video games to create random yet intentional content. Levels, backstories, characters, items, quest lines: all can be constructed with rolls of the dice, as it were. Early in the book, the techniques mirror meat space randomizers from dice to shuffling to tearing and rearranging paper. The implementations, however, rely on a computer, and by the time we're talking about noise, we're fully in the digital realm with algorithms outside the scope of the text. This is a book for exploration and enjoyment, not technical implementation.
Cook structures his book in pairs of chapters. The first will introduce a theoretical concept: cut-and-paste components, for example. The second will show how that concept is implemented in an exceptional and influential video game: Spelunky levels, for cut-and-paste.
As a celebration of brilliant developers and their games, Next Level delights. Cook is excited about his material and knows precisely the amount of detail needed to help the reader understand what he’s so fired up about. This book has time for Rogue’s room placement biases, Dwarf Fortress bugs, Elite’s ambitions that were too extreme for its publisher, and much more. If you enjoy roguelikes or procedural exploration games like Minecraft or Caves of Qud, you will enjoy this book. I was prepared to slog through the last few chapters that broke from the theory-example pattern to talk about the Elephant and the future. That was unwarranted doubt. A standout for me toward the end is the discussion of recreational memory and storage hacking of running games called game corruption. Cook connects it to the musical movement plunderphonics; I think the found material ethic is quite reminiscent of musique concrète. As for the Elephant, one of the gifts of a well-written book is forgetting for a few hours that generative AI in the sense we now popularly use the term exists at all. Mike Cook gives the topic the chapter it is due but mostly ignores it, and so will I.
Next Level says in an early chapter that it is not about tabletop games. All the same, Prismatic Wasteland’s Random Blogwagon is on, and I churn everything game-ish through the TTRPG filter. Fortunately for me, there was plenty to catch from Next Level. The book opens by talking about dice, their history, and the emotional quirks of rolling method that share some statistical traits and diverge in others (2d10 vs 1d20, for instance). Check out the concise and enjoyable Goblin Guide to Advanced Die Rolling for similar discussions.
The most striking theme for me throughout Next Level is the message that procedural generation is not a replacement for hand-crafted design. It is hand-crafted design. That is, it’s not just that an individual would be incapable of manually writing the history of every game world in Dwarf Fortress. The game would be missing some of its essence if the dev tried. The random generator itself reflects the developer’s desires, beliefs, and goals. It carries their agency even as it acts in unpredictable and sometimes maddening ways.
Because procedural generation carries the author’s agency, it can be directed to different ends. Cook likes to compare and contrast games that use similar tech to highlight the impact of different implementations. Rogue uses randomized scrolls and rooms to create caution and claustrophobia; Elite uses randomized worlds and jobs to create freedom and wonder. Dwarf Fortress uses a detailed and expansive history generator to create traceable cause-effect from the start of the world state to the present; Caves of Qud starts with traits and events for its prominent characters then uses those facts to fill in a spotty history afterward. The comparison between these large world simulators calls to mind different methods of character building: chiseling a backstory for a clear concept as opposed to starting with traits and “discovering” the concept in the course of play.
In Next Level, randomization answers any number of questions. What enemies should appear? What climate should the player encounter next? How did the characters get to where they are now? What kind of person is this NPC? What jobs are available? And dozens more. Every time, letting go of the handlebars unleashes something specific. Platform placements can surprisingly challenge. A room can provide a welcome gift. A certain NPC combined with a quest can suggest a joke directly planned by no one. It’s worth knowing why we’re using randomization when we choose to pull out a rolling table.
Incidentally, the various forms of randomization in the book also suggest that dice are far from the only way to answer those questions, with or without a computer. But I’m a boring dice adherent and not the person to advise on novel randomization methods.
Cook would probably love to hear what you come up with, though. He is aware of and generous toward his community, both in the text and in endnotes. His closing words encourage all of us to join the play space of procedural generation. It doesn't need to be good. It doesn't need to sell. Just make something that makes something.
How did I read it for free?
I got it from the library. That sounds like a dodge, but the trick is that the library bought it because I asked for it. My library district, and possibly yours as well, lets patrons request new books to be brought into the collection. They buy them and let the requester have the first check-out. I can do this before the book is published. I can request academic books on archaeology, a newly localized manga collection, and a book on procedural generation in the same month. See if you have this power, too, at your local library. I guarantee you’ll be able to find something you can bring to your gaming table to make it more interesting, like this book. Don’t say I forget to pay my taxes.
Ok, fine. Behold, the Weapon Tiers. Good for Mark of the Odd and its ilk, the Weapon Tiers provide a nice curve that can slide weapon effectiveness up or down depending on quality and circumstances. What’s better, it can use any single die size.
1d, odd: 1 even: 2 => 2d keep 1 lowest => 1d => 2d keep 1 highest => 2d => 3d keep 2 highest