Freeze the official banzuke for each tournament before joining it to results. Preserve former and new rank, number, side, wrestler ID and pair state. A spelling repair and a redefinition of the comparison are not the same kind of update.
East or west is not a permanent wrestler attribute. A wrestler’s position can change at the next tournament. Joining historical results to the newest banzuke would overwrite the side that applied when those bouts occurred, so tournament ID and banzuke version are mandatory.
One versioned row per banzuke position
Store tournament ID, division, rank, number, side, wrestler ID, display name, retrieval time, official URL and dataset version. Generate a pair ID from the same rank and number. If its counterpart is absent, retain an unmatched reason instead of filling the slot from another line.
After joining bout data, validation checks total ranked wrestlers, counts by side, matched pairs and unresolved identities. Publication stops if a result participant cannot join to the banzuke, one wrestler receives two sides in the same basho, or a side has been inferred from website display position.
Classify the reason for correction
Separate a new official-banzuke capture, following an official display update, a name or ID transcription repair, pair-matching repair and analytical definition change. Updating a shikona to the official display is presentation maintenance. Discovering that a historical row joined to a different person is a substantive correction.
Expanding the analysis from exact same-number pairs to all wrestlers at the same rank is a definition change. Preserve the old chart, publish old and new sample counts, and explain why the result moved.
Current public baseline
- 3 September 2026, v1.0: side is stored as a tournament-specific banzuke attribute.
- 3 September 2026, v1.0: exact-rank-and-number pair, vacancy and unresolved states are distinct.
- Next checkpoint: after the next banzuke announcement, add a new snapshot and run a difference check without overwriting previous tournaments.
Changing page decoration or copyediting does not advance the numerical release. The dataset version changes only when side assignment, cohort definition or an aggregate value changes.
Follow one repaired row back to the ring
After correcting a side, return to that wrestler’s fifteen days. Watch how the neighbouring same-number wrestler meets opponents: are they the same names, or do they arrive in a different order? With a visible correction log, the banzuke stops looking like a static list and starts to feel like a map rebuilt for every basho.
Missing and Excluded Records Are Not Ordinary Results
Absences, defaults, no record, unclassified outcomes and incomplete observation are not converted into zero or a standard bout. Count them as missing data or exclusions, and state whether each entered the denominator. An unverified value stays in a review row rather than being imputed from a convenient pattern.
Freeze the Question Before Reading the Result
The research question is whether the defined metric differs under the article’s stated conditions. “There should be a difference” is a hypothesis, not a conclusion entered before counting. Preserve a contrary result, and treat a small difference, heavy missingness or incomparable conditions as grounds to hold the conclusion.
Sources
- Japan Sumo Association banzuke, daily bouts/results and tournament scorecard (checked 3 September 2026).
Source links used for verification
Each link was used for the corresponding claim in this article; inclusion is not a blanket endorsement of the whole page.
Article verified:
- Japan Sumo Association banzuke tablePrimary source (as checked for this article)
- Japan Sumo Association daily bouts and resultsPrimary source (as checked for this article)
- Japan Sumo Association tournament scorecardPrimary source (as checked for this article)