Case Study
A tabletop RPG sounds like one thing. It isn’t. It’s a lot of different deliverables that all have to agree with each other, and a lot of people with opinions about every one of them.
20 Below grew into a rulebook manuscript that feeds a web book, a reference document, and a print PDF. It has four apps, each shipped as both a web build and a desktop build: a character creator, a battle tracker, an encounter calculator, and a page-layout tool. There are three extensions for a virtual tabletop, each in its own repository. There’s a campaign setting in its own repository too, and a brand library of reference sheets, diagrams and dividers. Change one rule and it can land in eight places. If you miss one, a player finds it at the table.
The team is fifteen people: a communications lead, a language expert, an engineer, a web development consultant, four game designers, and seven playtesters. Everyone is looking at the same rules, and everyone sees something different in them.
Fifteen people who all care about a game will pull it in fifteen directions. If each of them edits the rules directly, the rules fork. If feedback goes through layers, it arrives late and watered down. And the rulebook has to read like one voice wrote it, because a rulebook that argues with itself can’t be played.
Then there’s the AI. An AI assistant is fast and tireless, and it’s confidently wrong often enough that “it said it’s done” can’t mean “it’s done.” Its risk is a quieter version of the team’s: output you trust because it sounds sure of itself.
Both problems have the same answer. Everyone gets to see everything and say anything. Decisions go through one person, and that person checks every change against the source before it ships.
A roundtable is only worth running if it changes the product. This one did.
Playtest feedback led to a complete overhaul of how new characters buy their starting gear with Wealth. It’s now its own procedure with its own reference page.
The designers pushed hard for social and mental attacks to work like physical ones. They got them. Presence and Psyche became walls alongside Soak in August, and every attack in the game now hits one of three tracks: Health, Poise, or Sanity. Their feedback also shaped the critical hit rules, the Ki die boost, the damage scale, the size of the Ki pool, how Fate Tokens are spent, Flaw levels, and the Adders and Limiters that tune every Gift.
Found a dead-end link in the character creator app that went nowhere, and it was fixed.
An ongoing, detailed design review of the character sheet and every reference sheet: type that smears, lines that crowd an edge, an arrow sitting a few pixels low. It’s why the character sheet is still being rebuilt instead of being called finished.
Figures are from the project’s version history and count game work only; commits that touch nothing but these portfolio pages are left out. The design work goes back further than the version history does.
A system with a lot of moving parts ships as one thing and holds together: one set of rules behind every app, and a print book and web book built from the same source. Fifteen people shape it, and it still reads in one voice. The team brings the problems, the AI speeds up the work, and the decisions and the final check stay with the hub. That’s the job: keep everyone informed, keep one version real, and don’t ship anything you haven’t checked.
“I found a link in the Character Creator that led nowhere, the kind of small thing that makes a new player lose trust in a tool. Feral had it fixed and released quickly, and thanked me for catching it. On 20 Below, a bug report from anyone on the team gets the same attention as a design decision, and everyone can see when it’s done.”
| Communications | Hillary McKey |
| Language Expert | Rilla Scarberry |
| Engineering | Louis Leboeuf |
| Web Development Consultant | Dillon Wyatt |
| Game Design | George Szabo, Shawn Beatty, Jordan Sorrick, Ty Irons |
| Playtesting | Seven playtesters, uncredited by request |