Commit-level time series for stable and unstable software periods

This dataset contains the raw commit-level time series used in the study Long-Range Correlation in Code Commit Dynamics as a Novel Indicator of Software Product Stability: A Detrended Fluctuation Analysis Study (Mitevski in review). Each record captures the number of lines of code added in a single commit event. The data were extracted from the complete GitHub activity log of a small, globally distributed closed-source software organisation with more than five years of continuous operation. Two non-overlapping 712-day periods were identified and labelled by the lead software engineer as stable and unstable based on the frequency of software crashes — instances in which the software became unresponsive — as reported by Google Analytics and equivalent monitoring services. The labelling criterion is grounded in observable, user-facing behaviour rather than developer judgment of code quality. The dataset consists of two CSV files: unstable_period_code_additions.csv — 3,129 commit events with number of code lines added across 712 days (mean: 4.39 commits/day) stable_period_code_additions.csv — 976 commit events with number of code lines added 712 days (mean: 1.37 commits/day) Each file contains two columns: time_point (UTC-aware ISO 8601 timestamp of the commit event) and additions(number of lines of code added in that commit).