Zugwise: Domain-Aware Game Reconstruction for Digitizing Handwritten Chess Scoresheets
Summary
- Zugwise is a domain-aware game reconstruction system for digitizing handwritten chess scoresheets by framing the task around plausible human move generation rather than basic handwriting recognition.
- The system repurposes computer-chess machinery like quiescence search and static exchange evaluation to detect silent, legal-but-wrong move errors that standard legality-only pipelines miss.
- Integrating a Needleman-Wunsch merge of both players' sheets and learned signals such as a Maia-style human-move prior significantly increases the correct fix top-rank rate to 78.0%.
- Zugwise runs entirely in the browser, offering a free, private, and offline-capable application alongside a three-way adjudication protocol for domain ground truth.
Why it matters
By shifting from simple text recognition to domain-aware game reconstruction and plausibility checking, Zugwise demonstrates how incorporating specialized search mechanics and priors can dramatically improve data digitization accuracy in complex human-recorded domains.