The Coding War Games is the empirical foundation of DeMarco and Lister's argument in peopleware. It was a multi-year productivity study conducted through the atlantic-systems-guild in which participants from different organizations completed the same standardized coding and testing exercise, allowing controlled comparison across hundreds of programmers and dozens of organizations.
Study Design
The exercise was a defined programming task — participants implemented a specified program from scratch, then tested it against a battery of test cases. Participants were drawn from real software organizations: they completed the exercise individually, on their own time, using their own tools and environments. Their organizations' managers were asked to complete a companion survey about the workplace conditions — office size, noise levels, privacy, interruption frequency, and related factors.
This design was a significant methodological innovation. Rather than relying on self-reported productivity estimates or project-level metrics (which conflate individual contribution with team and project variables), the Coding War Games produced directly comparable performance data: time to complete, defect rate, and test passage rate.
Key Findings
The most cited finding is a 10:1 variation in individual productivity — the best performers completed the exercise in roughly one-tenth the time of the worst, with correspondingly lower defect rates. The magnitude of this variation had been noted by others (including fred-brooks and gerald-weinberg) but the Coding War Games provided unusually direct evidence.
Crucially, DeMarco and Lister then asked: what predicts performance? Their analysis found that the standard candidates — years of experience, salary, programming language used, level of formal education — showed weak or negligible correlation with performance. What did correlate significantly was workplace environment quality, as reported by the companion organizational surveys.
The specific environmental factors that distinguished top performers included:
This is the empirical anchor for the office-environment-effect argument and the flow-and-interruption-cost analysis.
The Matched-Pair Analysis
One of the study's most compelling analytical moves was examining matched pairs: pairs of participants from the same organization who performed very differently. Within each pair, salary, experience, tools, and programming language are controlled — both people work in the same environment, use the same methods, and have similar backgrounds. Yet performance still varied substantially within pairs.
DeMarco and Lister used this to argue against purely individual explanations. Even when you control for organizational context, individual variation remains significant — which is part of why they argue so strongly against interchangeable-parts thinking about programmers.
Limitations and Context
The Coding War Games data comes from a standardized exercise rather than real project work, which means it may not fully capture the collaborative and architectural aspects of software development. The study also accumulated data over time across different administrations, which introduces methodological heterogeneity. DeMarco and Lister acknowledge these limitations but argue the pattern is robust enough to be taken seriously.
The coding-war-games-study as a specific event refers to the ongoing data collection that fed the analysis in peopleware and subsequent work. The data continued to inform DeMarco and Lister's consulting and writing into the 1990s and beyond.
Significance
The Coding War Games gave DeMarco and Lister a rare commodity in software management writing: actual data. Most software management advice is based on anecdote, consulting experience, or reasoning from first principles. The ability to say "we measured this across hundreds of programmers" gave the peopleware-thesis a credibility that purely argumentative books lack.
The 10:1 productivity variation finding has become one of the most widely cited statistics in software engineering debates about individual differences, hiring, and workplace design — often cited without reference to its empirical source in the Coding War Games.