RidTheWann Performance — 0.5.0-beta (candidate; publication pending)

Measure. Diagnose. Prove.

Release scope (0.5.0-beta candidate; GitHub pre-release publication pending; 0.1.0-alpha through 0.4.0-beta records frozen): single-device validated only (see docs/methodology.md §7 and docs/limitations.md); installer is unsigned (expect a SmartScreen "unknown publisher" warning); PresentMon.exe is user-supplied (live capture is Windows-only via the GUI or the rwp capture CLI). Release notes: docs/release/0.5.0-beta.md (candidate), docs/release/0.4.0-beta.md, docs/release/0.3.0-alpha.md, docs/release/0.2.0-alpha.md, docs/release/0.1.0-alpha.md (frozen).

RidTheWann Performance (rwp) is an evidence-based gaming performance analyzer for Windows 11. It correlates frame timing with system state, detects reproducible failure patterns, and explains evidence without treating a heuristic score as probability. Streaming attribution is not shipped: no encoder telemetry exists and streaming_overhead never fires (see docs/limitations.md).

Measured example — single device, labeled (Windows 11 Build 26200 x64, Ryzen 5 5600G, 16 GiB, non-admin, release profile, 50 ms cadence, ~211 target threads, 12,000 ticks):

Sampler mean 18.855 ms/tick, rolling p99 28.352 ms/tick; system-only mode 0.016 ms mean. Probe-process CPU 23.75% of one core.

This describes one box, not hardware in general. Method and validation basis: docs/methodology.md. What RWP cannot claim: docs/limitations.md.

Product boundary

rwp ships capabilities in this invariant order, currently ending at Explain:

Measure -> Analyze -> Diagnose -> Compare -> Explain

Recommend and Optimize are roadmap direction, not shipped capability; nothing in rwp emits a recommendation or optimization output.

The engine is deterministic and offline-first. Capture is out-of-process. There is no DLL injection, game hook, overlay dependency, account, or background telemetry requirement.

Quickstart — CLI

rwp doctor --presentmon C:\Tools\PresentMon.exe
rwp capture --presentmon C:\Tools\PresentMon.exe --process-id 1234 --duration-seconds 60 --output-dir sessions/live
rwp import-presentmon capture.csv --process-id 42 --qpc-frequency-hz 10000000 --output-dir sessions/run
rwp analyze --session sessions/run --output-dir sessions/run-analyzed
rwp report --session sessions/run-analyzed --format md
rwp compare --baseline sessions/a --candidate sessions/b --same-scenario
rwp migrate --session sessions/legacy --output-dir sessions/migrated

Legacy reports or sessions go through rwp migrate (see docs/troubleshooting.md). Full command reference: docs/cli-reference.md. Error codes: docs/cli-errors.md.

Quickstart — GUI

Open RWP and follow the Home readiness checks. PresentMon remains user-supplied: when prompted, choose a supported PresentMon.exe from its official source; RWP does not bundle or download it. Open your game, select Refresh, choose the running application, and select Start measuring. The default preparation countdown is 5 seconds and the measurement duration is 60 seconds. Duration, telemetry interval, and other technical controls stay under Settings → Advanced. Results, verified local history, Before/After comparison, and failure remediation are built in. Manual Windows/PresentMon acceptance remains tracked in docs/acceptance/gui.md.

Project status

The authoritative milestone, release-gate, audit, and verification status is maintained in docs/project-ledger.md and the active autopilot state in docs/autopilot/STATE.md. The release plan is docs/autopilot/MASTER_PLAN.md.

Engineering constraints

• No DLL injection, hooks, or in-process overlay.
• No network access in normal operation (RC-10 evidence recorded).
• Missing measurements are represented as absence, never fabricated zeroes.
• Diagnostic evidence scores are not probabilities.
• Unknown is a first-class result.
• Local verification gates are mandatory: fmt, Clippy, tests, cargo-deny, cargo-audit, schema and docs checks (see docs/local-gates.md).

License

Licensed under either the MIT License or Apache License 2.0 at your option.