A Novel Approach for Estimating Truck Factors
https://arxiv.org/pdf/1604.06766
Method
"we propose a novel (and automated) approach for estimating TF-values, which we execute against a corpus of 133 popular project in GitHub. We later survey developers as a means to assess the reliability of our results."
Population
"a corpus of 133 popular project in GitHub. In total, such systems have over 373K files and 41 MLOC; their combined evolution history sums to over 2 million commits."
What it does not show
Popular, actively maintained open-source projects in six mainstream languages only; says nothing about industrial codebases. It measures the concentration itself and no downstream consequence — it does not test whether a low truck factor predicts defects, slower delivery or abandonment. Only 53% of respondents fully endorsed the specific named authors.
Guilherme Avelino, Leonardo Passos, Andre Hora, Marco Tulio Valente
“87 systems (65%) have TF ≤ 2” — most popular, actively maintained open-source projects could be incapacitated by one or two departures. In 84% of valid survey answers respondents agreed the identified authors were the main authors of their system.
Tier II: Observational computational study over real commit histories of 133 projects with a developer survey used to check the automated metric against practitioner judgement. Descriptive, no comparison group.