A solo RPG dev nails a problem every systems-heavy game faces: changing a scaling formula that looks fine on a graph, then finding three weeks later that a mid-tier item dominates a level band you never manually checked. The curve was fine — the problem was the interaction with crit, attack speed, and enemy resist.
The thread surfaces a few approaches:
- Spreadsheet tabulation — the default. Plot every stat and average across 100 levels, scan for visual outliers. Most people do this, and it catches single-variable drift but misses cross-system interactions.
- Combat simulation + statistics — one dev tried four iterations: a combat generator, statistical TTK averages, optimization (failed — no suitable objective function), and finally simulating many hero variants against bins of actors. The hard-won lesson: the fewer variables the better. Derive HP from weight. Shrink the space before you simulate.
- Alpha/beta + postlaunch patches — accept that no automated tool catches everything, use playtest metrics, patch after ship.
The most transferable insight is from jube_dev’s four-attempt journey: reducing the number of independent variables by computing derived stats from formulas is the only systematic way to make the remaining combos testable. The same balance math that works at parameter level breaks at interaction level, and no off-the-shelf tool handles that — everyone builds their own spreadsheet.