| name | 09-game-test-and-player-evidence-plan |
| description | Use when planning game QA, playtesting, balance, deterministic simulation, save migration, multiplayer network conditions, device performance, accessibility, localisation, security, certification, and release-candidate evidence. |
| metadata | {"portable":true,"compatible_with":["claude-code","codex"]} |
Game Test and Player Evidence Plan
Use When
- A game milestone needs technical tests and observed-player evidence.
Do Not Use When
- Programme testing policy is the task; use
01-test-strategy.
- Do not substitute opinions for correctness or device evidence.
Required Inputs
| Artefact | Source or provider | Required? | Behaviour when missing |
|---|
| Requirements, architecture, build manifest, targets and gates | Product, engineering, QA and release | Yes | Stop and issue a readiness gap. |
| Research, learning, culture, wellbeing, ads/privacy, network, accessibility, localisation and threat cases | Specialist owners | Conditional | Mark coverage not assessed. |
Workflow
- Build a risk matrix across unit, integration, simulation, content, device, network, security and access layers.
- Specify seeds, clocks, tolerances, replay and save corruption/migration/interruption.
- Test latency, loss, jitter, reorder, duplication, disconnect, crash and version mismatch online.
- Profile release builds on named hardware for pacing, CPU/GPU, memory, loading, storage and thermal behaviour.
- Separate correctness, comprehension, appeal, balance, learning, cultural acceptance, wellbeing/stopping, ad interruption/trust, commercial and release questions; no evidence class substitutes for another.
- Test critical verbs with alternative input and redundant feedback.
- Test every ad placement for expectation, disclosure, focus/input, decline/close, reward, cap/pace, save/recovery/stopping/exit, child treatment, accessibility, no-fill/offline/SDK failure and kill switch on real devices.
- Stop on missing build identity, oracle, consent or accountable reviewer; recover with an evidence plan.
Outputs
| Artefact | Consumer | Acceptance condition |
|---|
| Game test and player evidence plan | QA, design, engineering and release | Cases map to requirements, environments, oracles, evidence and decisions. |
Evidence Produced
| Evidence | Reviewer | Acceptance condition |
|---|
GTEST-* matrix | QA reviewer | Build, device, scene, seed, network and outcome are reproducible. |
Capability and permission boundaries
Read and search are allowed; planning is read-only. Shared-service, player-data or security tests require authority and consent.
Degraded mode
Return the narrowest qualified plan, mark unavailable checks not assessed, and retain blocked gates.
Decision Rules
| Choice or condition | Action | Failure or risk avoided |
|---|
| Oracle absent | Redesign or classify exploratory. | False confidence. |
| Capture lacks identity | Reject release evidence. | Irreproducibility. |
| Core accessibility blocker | Block or approve explicit exclusion. | Menu-only claims. |
Quality Standards
- Retain failed-path evidence; distinguish preference, comprehension, balance and correctness.
Anti-Patterns
- Tautological tests. Fix: assert invariants.
- Editor-only testing. Fix: packaged builds.
- Average FPS only. Fix: frame pacing.
- Localhost multiplayer only. Fix: hostile profiles.
- Anecdotal playtests. Fix: question, cohort and disposition.
- Ad renders so it passes. Fix: test interruption, accidental tap, coercion, privacy, regret, trust and all failure paths.
References
