Every power question you ask MCPower is answered by running thousands of small simulations: the engine builds synthetic datasets from your design, analyses each one, and counts how often the effect shows up. The part that makes that fast, reproducible, and identical whether you run it from Python, R, the desktop app, or the browser is a single compiled native engine doing all the numerical work.

You never touch the engine directly — the package or app in front of you is a thin layer over it. These pages lift the lid for the curious reader who wants to know how the numbers are produced and why they can be trusted.

• How it's built — one engine, four ports, so a result in Python is the same calculation as in R, the app, or the browser.
• Why it's fast and reproducible — automatic multi-core simulation, a reproducible random-number stream, and the one knob you actually control: simulation count versus precision.
• How we know it's right — the evidence that the engine generates the design you asked for and solves it the way a trusted reference does.

For the bigger picture of what MCPower is and who it's for, see About MCPower.