Connect-4

Does grouping symmetry-equivalent Connect-4 positions change apparent test accuracy?

After qualified ingestion and an immutable evaluation plan, recompute confusion matrices, classwise metrics and uncertainty from held-out predictions and frozen labels; fit preprocessing inside training partitions. Successful reproduction would not validate original measurements or causal interpretations.

What you could produce

• Exact input manifest with data-file digests and a schema/rights audit.
• Frozen evaluation protocol with split or group identifiers and an explicit baseline.
• Bounded analysis report with recomputable metrics, uncertainty and unresolved findings.

Before you use it

• Resolve exact data files and digests through qualified isolated ingestion.
• Audit data schema, provenance and license conditions before designing a bounded analysis.
• Freeze preprocessing, splits, metrics, runtime requirements and budget before execution.

Limits to keep in view

• No dataset contents, dependency compatibility, runtime cost or scientific result was checked.
• Grouping, timing or identifiers needed by the proposed protocol may be absent; report that limitation rather than inventing split labels.
• Large or specialized sources may require a predetermined subset or a broader qualified runtime; no resource fit is established.
• The qualified self-contained Python 3.13 pilot is not evidence that this dataset or its usual analysis packages can run.

Source and permission context

Attribute the supplied creators and UCI dataset DOI/source. Preserve license links and identify changes before any permitted data reuse.

Catalog listing reviewed. This review covers the description and source links displayed here.

Reviewed the original description of Connect-4: Enumerated game positions for studying categorical prediction and symmetry-aware evaluation. Its public UCI identity, supplied creators, scoped license observation and per-record page evidence match the stored official metadata. Listing approval concerns that description and source links; the recorded dataset license is not treated as a license for every linked artifact or as metadata rights clearance.

Reviewed 2026-09-14. Copying or adapting source files remains subject to their own terms.

Before copying source material

• Data files and embedded or third-party notices were not downloaded or inspected.
• The API's last-updated date is metadata, not an immutable dataset version or content digest.
• Any reused generator code needs its own version and code-license check; a dataset license does not establish software reuse rights.