Score the experience with evidence
Use four dimensions and attach an observation to each. Clarity: did a first-time player understand the goal without coaching? Control: did the jump respond when expected? Challenge: did a miss feel attributable to timing or to an unreadable rule? Recovery: could the player try again quickly? A simple “clear / uncertain / blocked” rating is often more actionable than a precise number that implies unsupported measurement.
Choose the next revision from the largest risk
Do not average every observation into a vague request to improve the game. Sort issues by consequence: can the player start, perform the core action, recognize success or failure, and continue? Fix a blocked core action before balancing the final platform. For example, if landing feedback arrives after the camera has scrolled away, adjust feedback timing before changing platform spacing.
Questions about evaluate AI game prototypes
Should I judge an early game prototype by its graphics?
Treat visuals as evidence only when they help players read the rule, target, hazard, or feedback. Cosmetic finish is secondary while the core action is still uncertain.
What is a useful first prototype review question?
Ask a tester to explain the goal and demonstrate the main action after a brief uncoached attempt. Their explanation reveals whether the prototype communicates the idea.