game-prototype

Use the format: "We believe that [mechanic] will produce [player behavior/emotion] when [specific condition]." Then define falsification criteria describing observable evidence that would disprove it, and map the hypothesis to one MDA aesthetic such as Challenge or Sensation.

Three to five external testers are sufficient for a binary kill-or-promote signal at the prototype stage. If results are ambiguous with five testers, the mechanic likely lacks strength, since a strong mechanic produces obvious behavioral signals even in small samples.

Use paper prototypes for turn-based mechanics, economy systems, and card game loops where feel is irrelevant. Reserve digital throwaway prototypes for real-time input-to-feedback questions like grappling hook momentum or platforming feel.

No. Prototype code is treated as disposable and never graduates to production. When a hypothesis is validated, the real implementation starts from scratch with proper architecture, since prototype code lacks error handling, edge cases, and scalability.

Kill the prototype if fewer than 2 of 5 testers exhibit the target behavior, the core interaction fails to produce the intended feeling, the mechanic requires extensive explanation, the fun came from an unintended secondary interaction, or players found a degenerate strategy.

Solo developers should wait 48 hours before self-testing, record and review their own sessions as an outside observer, and still recruit at least 3 external testers. Remote testing via screen share or private itch.io links works when in-person observation is impossible.