How MyRigCheck works

Your Rig Power Score is the average of two calibrated hardware scores — one for your CPU, one for your GPU — normalized to a 0–100 scale where a top-tier modern build lands at 100. The score is then bucketed into one of five tiers: Elite (88+), High (70+), Mid (48+), Low (28+), or Minimal (below 28). It is a quick way to compare rigs against the 350,000+ titles in our catalog without running each one.

FPS estimates are computed from your hardware scores, the game's documented minimum and recommended requirements, the resolution you are checking, and a performance-scaling profile that adjusts for resolution cost. Results are calibrated against observed benchmarks for popular titles; for less-tested games the estimate is approximate. We surface a confidence percentage so you can tell the difference between a 90%-confidence estimate and a 60%-confidence one.

We estimate the relative load your CPU and GPU are under at each resolution. The component that hits 100% utilization first is flagged as the bottleneck. If both are near their limit, we report the system as balanced. Laptop thermal adjustments are applied automatically when you mark your rig as portable.

When you grant permission, we read your browser's WebGL renderer string and User-Agent to identify your GPU. CPU detection uses the User-Agent CPU hint where available. The detected hardware is then matched against our reference database of canonical SKUs. Hardware we cannot match falls back to manual entry.

Game metadata, genres, themes, and platforms come from IGDB. Steam review counts, peak concurrent users, and pricing come from the Steam Store API. We re-ingest and re-index our read layer weekly; the data-refreshed timestamp in the footer tells you when the last build completed. The full data provenance is at /data-sources.

Hardware scores are derived from the curated reference set of 265 SKUs (data/benchmarks/*.json) by mapping each component to a 0-100 score via percentile-rank against the set, then calibrating against observed community FPS reports. The calibration loop:

1. Bootstrap the calibration corpus (10,000 iterations, 95% CI).
2. Fit a per-game residual (observed − predicted) on the calibration set.
3. Re-derive the FPS estimate for each game using the residual as a prior.
4. Cross-validate on a held-out 20% of the corpus.

The estimator does not currently ingest driver-version data, so estimates for rigs on drivers within 14 days of a major GPU driver release may shift by 10-20% in either direction.

The hardware reference set is a static JSON corpus of 265 clean SKUs (CPUs and GPUs). The set is JOINED into the read layer at build time, not seeded from a third-party dataset. Each entry contains a 0-100 score, the SKU's official name, the manufacturer's spec sheet URL, and the calibration date. The full set is exposed as part of the public dataset at /api/dataset.json.

Manufacturer spec sheets are the primary source for any hardware-specific claim (TDP, core counts, memory bandwidth, base/boost clocks). They are cited by SKU in the public dataset's hardware[] array.

Rig data is user-submitted (anonymized, never sold) via the /check form. We do not publish per-rig data, but we do use it for rig-coverage statistics (e.g. "12% of submitted rigs can run this game at 60 fps"). The community FPS report on every game page feeds the calibration loop.

Estimates do not account for every game build, every driver revision, or every background process. The model under-estimates on games with heavy CPU single-thread dependence and over-estimates on poorly-optimized ports. We do not currently measure frame-time variance, 1% lows, stutter analysis, VRAM headroom under load, network latency, or ray-tracing / DLSS / FSR scaling behavior. We do not currently support console platforms. If you spot a result that is badly wrong, the community FPS report button on each game page lets you submit observed numbers — your reports feed the next calibration run.

The full limitations list — including the known failure modes and what we explicitly do not measure — is at /accuracy.

• IGDB API — game metadata, genres, platforms, themes, modes.
• Steam Store API — system requirements, reviews, peak concurrent users, pricing.
• Manufacturer spec sheets — Intel ARK, AMD product pages, NVIDIA product pages — cited by SKU in the public dataset.
• /data-sources — full data provenance.
• /accuracy — calibration method and known failure modes.

This methodology and the underlying dataset are semantically versioned. The current version is 1.0.0, released 2026-07-25. The next scheduled review is 2026-10-25. The full public changelog is at /CHANGES.md.