Empty Shell: When a Martial-Arts Sports Report Comes Back with Nothing

A sports desk's morning normally begins with checking last night's results. Here the opposite happened: after running the first stage of the data pipeline for a martial-arts piece, what came back was an empty shell. Title field: N/A. Source: N/A. Type: Unclassified. Core viewpoints: blank. Information points: an empty list. Only one field carried a value, the domain label reading martial_arts. If that single word is used as the basis for analysis, whatever gets produced is not reporting but fabrication. The arithmetic of five fields is simple. Of the five — title, source, type, core viewpoints, information points — the first three are either N/A or Unclassified, and the last two are empty lists. Not one piece of raw material needed to build a report is present. No source means source credibility cannot be assessed. No type means there is no way to tell whether this is news, commentary, instructional material or academic work, and each of those four demands a different analytical frame. This is where the biggest confusion sits. Many assume the first stage did half the work and the rest will follow. But the next three fields — entities, time sensitivity and source quality — are inferred at exactly that stage from the information points. With no information points, there is nothing to infer from. Their emptiness is therefore inevitable, and it is not half-finished work but zero output. The easy path here is to seize on the word martial_arts and write an analysis of the combat-sports market, training methods or international federation structures. That is not analysis, it is construction. Building a report out of a label means handing readers claims they cannot check. In sports journalism that is the cardinal sin, because sports readers are trained to cross-check results, times and names. One likely cause: the body text never came out of the source at all. Scanned PDFs, text locked inside images, JavaScript-rendered pages or broken encodings leave the extractor holding an empty string, while the schema still returns a well-formed but empty JSON. The pipeline does not fail; it merely returns nothing. A second possibility is a schema mismatch. The JSON the model returned did not use the field names the consuming code expected, deserialisation failed silently, and every value fell back to the default N/A or an empty list. A third is truncation: the text was too long so the input was clipped, or the response ended before the viewpoint section was written. A fourth is object confusion: the result of a metadata or summary request was mistaken for a body-text extraction. Diagnosis is short. Look at the token count of what was sent to the model and at the raw output that came back. If the input is close to zero, the extractor is at fault; if the input is fine but the output is an empty JSON, the schema or truncation is. The stakes go beyond one blank report. An empty shell is more dangerous than no data because it looks like a successful result. An editor who sees five blank fields knows something broke. An editor who sees the label filled in may conclude the analysis has arrived. That false assurance is the real damage. The harm then cascades. No information points means no entity list, so no athlete, organisation or competition can be identified. Time sensitivity cannot be set, so there is no way to say when which claim expires. Source quality cannot be scored either, which puts a verifiable fact and a marketing claim on the same footing. Telling those two apart is the core of sports reporting. Any one of three minimum inputs moves things forward: the full article; or title, outlet, publication date and body; or the raw input and output of the first stage verbatim. With the third, it becomes immediately visible whether the failure was in extraction or in serialisation. Once the body text is available, the analysis can be built in seven steps. First, an inventory of every falsifiable claim with its evidence type — data, quotation, case or bare assertion. Second, the argument structure: premise to inference to conclusion, with every leap and unstated assumption marked. Third, the author's position and purpose, target reader and persuasion technique. Fourth, an entity map of people, organisations, events and concepts and their role in the argument. Fifth, time sensitivity: which conclusions have a shelf life and when they expire. Sixth, a source-quality score covering primary versus secondary status, verifiability and potential conflicts of interest. Seventh, counterexamples and disputes: what the piece avoided or suppressed. The lesson for a sports desk is plain: the pipeline has to be able to stop. Treating an empty result as a success and passing it downstream means building a report on sand. Martial arts or cricket, a result cannot be published unchecked — and the same rule applies to data.