Git-Based Software Metrics for Evidence-Based Developer Reputation: Comparative Evaluation and Smart Contract Feasibility

This thesis studies software quality metrics as verifiable evidence for developer reputation systems, with emphasis on authenticating who changed the code, what was changed, and how those actions can be audited. Building on prior work in software metrics, maintainability, reputation systems, and smart-contract evidence models, it evaluates multiple metrics instead of relying on a single indicator. A reproducible dataset of ten open-source repositories is constructed, covering 80 stable release tags and 70 release windows. The work compares snapshot size, churn, activity cadence, and ownership/concentration using stability, interpretability, resistance to gaming, long-term suitability, and smart-contract feasibility. The analysis includes duration normalization, bot handling, human-only developer evidence, and short-window robustness checks. This extends previous metric research by combining theoretical metric selection with a practical, replayable pipeline linked to contributor identities and versioned code changes. The results show that snapshot size is mainly contextual, churn is useful but volatile, and cadence is interpretable but easier to manipulate. Ownership/concentration is the strongest candidate because it reflects maintainer responsibility, connects changes to contributors, and supports auditable implementation. The thesis therefore proposes a more robust method for selecting a metric for a future proof-of-concept system where computation is performed off-chain and compact evidence records can be stored on-chain.