Skip to content

Theory before math

Written 2026-10-27 from the project's own design docs.

The principle: settle what a system is (its parts, its rules, how it connects to everything else) before solving its numbers. Treat every number written before then as a first guess, and judge the final numbers in play.

In The Crawl the founder put it plainly: "The scaling and impact of player attributes is math, and we don't need to solve any math until we've laid down all our theory."

Why it matters

  • Numbers tuned on the wrong structure are wasted. If the shape of the modifier model changes, every formula built on the old shape has to be redone. The Crawl's planned order was explicit: finish the theory, rebuild the combat code on the new model, and only then do the "pre-final" math.
  • Theory decisions are the expensive ones. A balance number is a one-line change. "Does a condition drain only health, or any pool?" changes the shape of every condition.
  • Early numbers still have a job. They make the theory concrete enough to test. The Crawl's first damage, stamina and exposure numbers existed so tests and simulated scenarios could run, not because anyone believed them final.

Practices that go with it

Label first guesses as first guesses

Both projects mark untuned numbers as such. Every physics and scoring number in Fboard's first build was recorded as "a first guess"; The Crawl's combat numbers are "first tuning". Writing that down stops a placeholder from quietly becoming canon.

Balance in two layers

The Crawl adopted this as law (Conditions and environment):

  1. Within a mechanic, by math. Things of the same kind and the same strength must cost the same. For conditions, "strength" is the harm done untreated over a fixed 60-second window, so a poison and a bleed of equal strength are interchangeable, and tests enforce it.
  2. Across play styles, by scenarios plus human judgement. You can't reduce "is a knife fighter as viable as a rifleman?" to one formula. Instead, simulated scenarios report survival times and outcomes (how long an average character survives fire, cold or a deep wound untreated), and the designer judges whether the results feel right.

The first layer catches arithmetic mistakes. The second catches design mistakes. Neither replaces the other.

Let a person's judgement have the last word

In The Crawl, the founder judges the scenario results. In Fboard, the plan after the first build put a founder playtest first ("collect what feels wrong; it drives everything else"), ahead of any tuning pass. Measurements tell you what the numbers are; whether the game feels right is a judgement call, and that's the designer's.

Feel is part of the theory

Fboard's "feel pass" chose its foundation before tuning: full real-world scale for time, speed and hand size, with the older arcade physics kept as a toggle. That choice (realism as the base) came first; the individual thresholds that followed are still marked as first guesses to tune from feedback.

Keep decided and open apart

Both projects keep what's decided separate from what's open, and record each decision with its date. That makes it obvious which numbers are allowed to move and which rules aren't.

The trap it avoids

Tuning feels like progress, and it's visible: tables of damage numbers look like a finished system. But numbers polished on a structure that's about to change are lost work. The Crawl deliberately stopped writing combat code until player and AI controllers were discussed, rather than polishing systems whose shape was still moving.

Where this came from