Build a short, stable smoke route
For a one-screen dodge game, a smoke route could be: start a run; move left and right; avoid one obstacle; confirm survival time increases; collide deliberately; restart; verify the next run starts with a clean timer and score. This sequence touches controls, obstacle collision, score, game-over, and recovery. Keep it short enough to run whenever a change is made, not only before release.
Add tests where the change can cause collateral damage
Pair each change with a nearby behavior. After adjusting jump height, test both a low ledge and a taller intended jump. After changing the game-over panel, test keyboard or pointer restart and make sure input does not also activate the control beneath it. After modifying a saved score, close and reopen the game if persistence is part of the design. This is more efficient than replaying every level without a reason.
Questions about game regression testing
Do small game changes need regression tests?
Yes when they can affect a core interaction or nearby behavior. A short smoke route often catches more useful problems than a large checklist nobody reruns.
When should I expand the regression checklist?
Add a lasting check after a defect is fixed or a new mechanic creates a distinct risk. Keep one-off exploratory observations in notes unless they become recurring release conditions.