Preface: how this book is made

Every number in this book comes from the dataset and is filled in when the book is built. Sentences that interpret a number, such as the claim that the pause button is the most common mechanic or that RimWorld's core mechanics are rare, carry checks that run against the data on every build. If the catalog grows and one of those claims stops being true, the build fails until the sentence is rewritten. The edition you are reading names the version of the dataset it was made from.

The dataset is public. Every game, mechanic, and technique is a record with its sources, its status, and a history of who created and reviewed it. I built the first version with AI agents doing most of the reading and drafting under rules that the build enforces, and people and agents now extend it through the same rules. Records marked "sourced" cite their evidence; records marked "verified" have been checked against primary sources by someone other than their creator. Every record this book relies on can be opened online to see which it is.

The catalog favors games that are well documented in English, so mobile management games and some Japanese and Chinese simulation traditions are thinner than they should be. The original rubric scores, such as how legible or how emergent a mechanic is, were assigned by one model rater against written anchors. Independent second ratings now cover 60 scores, and the rest still need another reading. Where a finding depends on those scores, the chapter says so.

Dataset b53bf4d5579d.