A practical book by Puma on moving from a surprising result to a better engineering decision. Learn to frame questions, measure fairly, preserve comparisons, reason about uncertainty and choose the next experiment.
Read online · Download the PDF · Downloads and reproducible companion
The expanded edition has 18 chapters, nine interactive examples, a mathematical reference and a 100-page PDF.
Follow the book in order for a complete scheduler investigation, or try a browser lab to explore distributions, matched comparisons, vectors, images, GIS and other worked examples.
The website needs no installation, login or backend. Browser-lab inputs stay in the page and are neither sent nor saved. For offline reading, download and extract the interactive-site ZIP, then open its index.html. The separate PDF is self-contained.
Download the companion and follow its included README for the Python environment, data snapshot and checks. The scheduler case uses measured Python-simulator reports and traces. Cohort and judge examples are synthetic. Each lab states its evidence and limitations; simulator measurements do not establish real-device or production-service performance.
The manuscript lives in source/chapters/ and source/references.md; docs/ is the generated GitHub Pages site.
Contributor instructions cover the Python/Pandoc build, Node tests, site checks and publication. Reading and using the supplied browser examples require none of those tools.
No distribution license is specified. Third-party works retain their original rights; see the sources and evidence notes.