macX public data & methodology

Find a game and an exact Mac

OpenAPI 3.1 specification describes the public endpoints for agent and application integrations.

1. Search /api/v1/games?q=Stray. Use the returned game slug; never guess an edition.
2. Search /api/v1/configurations?q=M3. Match family, size, year, chip, CPU cores, GPU cores and memory. If several match a vague user request, ask which configuration they own.
3. Request /api/v1/answer?game=GAME_SLUG&config=CONFIGURATION_ID.
4. Cite the returned canonical URL and original external evidence sources. Include resolution, preset, upscaling, frame generation, software version, test date and uncertainty in the answer.

Directory endpoints return 50 items per page with total, page and pages. Add page=2 to continue. /api/v1/methodology gives a compact machine-readable summary.

Reading an answer

• measured: eligible reports exist for the exact hardware. These are reported measurements, not a guarantee of current performance.
• estimated: no exact reports; a qualitative rating comes from the same family, introduction year, chip and CPU/GPU core counts. The target may differ in size or have equal to twice the source memory. Confidence is low. The closest eligible source configuration is chosen before rating, preferring the same size and nearest memory; settings are never combined across source Macs. Unknown or enabled frame generation is excluded from estimate sources. No cross-generation scaling or predicted FPS is provided. Source FPS belongs only to source hardware.
• insufficient_evidence: performance is unknown. This does not establish incompatibility.

Estimates never modify submitted benchmarks or the existing exact-Mac dashboard badges. Unmeasured answer pages are excluded from indexing; exact measurements get discoverable canonical pages.

How ratings work

Great: at least 60 average FPS. Good: at least 45. Playable: at least 30. Struggles: below 30. Public answers use these fixed thresholds; personal preferences do not affect them.

Hardware, graphics settings, storefront, measurement method, defaults status, macOS, game version, test type and scene remain separate. Median average FPS is calculated within each group. Repeated contributions from one member or one external source count once within that context; the latest is retained. This limits duplicate weighting but cannot prove that separately published tests are independent.

The existing defaults-first rating selector favors repeated contexts, non-generated frames, resolution and test recency. Tuning must improve the tier in the same context. Counter estimates remain distinguished from tool-reported averages. A representative group is shown, rather than averaging unrelated settings.

External sources must be entered with attribution, exact hardware, complete test context, publication date and a source locator. An administrator must approve them before public use. Only concise factual summaries are stored; source pages and videos are not copied. Sources are not automatically fetched. Review is not an independent reproduction of the test.

Missing versions and old tests are exposed, not silently treated as current. Tests over a year old receive a warning. Each answer scans up to 500 recent community and 500 external records for matching chip/core counts, prioritizing exact configurations; coverage limits are disclosed.

Access and responsible reuse

Limit: 60 API requests per minute per client IP across v1 endpoints. A 429 response includes Retry-After. Check X-RateLimit-Limit and X-RateLimit-Remaining. Retry with backoff, reuse directory results, and avoid scanning every possible game-and-Mac combination. Browser clients may read the API using CORS. Unsupported writes return 405.

Public results contain hardware, measurements and citations. They exclude account IDs, emails, saved nicknames, private review notes and credentials. Third-party sources retain their own rights; public access does not grant a license to reproduce their content. Link and attribute.

API version: 1. Methodology version: 2026-11-03.1. AI index · Crawler directives · Performance lookup.