Data and code for "Conditional predictive value of revision history in a traffic incident archive: An empirical comparison"

The deposit contains the two unchanged public Sydney V5 input files, deterministic audit/cohort/feature reconstruction, the original baseline and controlled-comparison code, frozen inputs and protocols, saved predictions and analyses, all retained unfavorable/failure records, table-transcription code, and the original bounded diagnostic plus its replay launcher. The computational replay uses already known results and does not constitute independent validation. The current literature note records completed readings of the published Li (2015) and Gao (2026) articles. Some reported implementation details and the correspondence of the fixed Gao code commit to its published run remain unresolved; a missing full-text access gap is no longer claimed. Both author-supplied publisher PDFs and their full-text derivatives are excluded. Licensing: study-derived data, tables and study-written documentation are released under CC BY 4.0. Study-written code is released under the MIT License (copyright 2026 YuanQiu Zheng). This is a file-specific allocation, not a choice of either license for every file. The source Sydney V5 dataset retains its own documented CC BY 4.0 terms, and third-party materials retain their applicable terms.