Game Systems Design

The function layer: mechanics, balance mathematics, and the models that govern them.

The common failure is structural: the systems are never written down as systems. Difficulty lives in a designer’s intuition, the economy is whatever the content added up to, progression is a feeling. When the game drags or snaps, there is no instrument to move, only opinions to hold. Content then multiplies on top of ungoverned mechanics, and every new piece makes the tuning problem combinatorially worse.

The work is to build the function layer as engineering. Rating models place a player precisely and keep placing them as they change. Spaced-repetition curves schedule difficulty against how memory actually decays. Economies carry stated sources, sinks, and the equation that keeps them honest. Telemetry is designed per mechanic before content ships, so every patch is an experiment with a readout: agents sim headless runs at volume, the numbers cull candidates, and a person keeps the survivors. The firm has built this layer with its own hands: a full Glicko-2 rating engine implemented from the original paper across six skill dimensions, an FSRS-5 spaced-repetition scheduler, thirteen practice opponents defined by roughly twenty tunable parameters each, and a 3,733-entry opening book in which every line is legality-checked in CI.

What a studio keeps from this work is an instrument panel it did not have: when the game drags or snaps, there is a lever to move instead of an argument to win. Mechanics governed by mathematics stay tunable as content multiplies, and the tuning becomes an operation the team runs on its own build, patch by patch, long after the engagement ends.