Zero Information Points: The Cricket Desk Report That Refused to Invent a Match

The file opened at eleven at night on a Kolkata data desk. The structure was complete — eight dimensions, more than thirty cells, each holding a properly formed sentence. The trouble was that every sentence said the same thing: insufficient information. The list of information points was empty. Title N/A, source N/A, article type “Unclassified”, domain label “cricket_world” — which does not match our standard “Cricket” tag. Four people on the desk offered identical advice: fill something in, the template cannot go out half-done. The file that was eventually signed off two hours later contained no invented team, no invented player, no invented number. Its conclusion was a single line — input insufficient, analysis withheld.

No journalism school treats that as a success story. But a data desk is really tested where the pressure to fill gaps runs hottest.

This is not a report about a match. It is a report about a pipeline. The desk had ordered a Stage-1 deconstruction of a cricket article — the process that pulls atomic facts out of a source text, the units we call information points. An information point is something that can be checked independently: a score, a date, a contract figure, a ruling, a direct quote. What came back had none. Headline blank, outlet blank, time sensitivity “not assessed”, entities described as “identify from the information points above” — while there were no points above.

Two failures produce this. The source page may be JavaScript-rendered, so the scraper receives an empty shell. Or a paywall and a block get in the way. Either way the system cannot tell a thin article from an article that never arrived. Our house rule is blunt: treat it as a pipeline failure, not as a thin piece. Confidence sits at high because three signals arrived together — zero information points, an “Unclassified” type, and a non-standard domain label. Any published cricket report, whether a scorecard, an auction story or a boardroom fight, leaves behind at least one team, one player or one institution. Leaving behind zero names is not news. It is a broken pipe.

The first gate of cricket analysis is format, and that gate was shut. A Test average, a T20 strike rate and an ODI economy rate are three different languages; putting them in one sentence is not a translation error, it is a lie. A powerplay rate of 9.1 an over means one thing in T20 and something else in the first hour of a Test. Until format is fixed, six of the eight dimensions — player, team, league, governance, risk, transmission — stay empty. That is what the desk did. No dimension was filled by inference.

The strictness can read as posturing, so I keep one case in mind where the data was populated and still lied. In 2026 in Kolkata I built a ball-by-ball pressure index for the IPL. Kolkata Knight Riders carried a powerplay run rate of 9.1 — handsome on the table. The index showed that 41 per cent of their boundaries came off edges or mistimed shots, and that behind Sunil Narine’s opening cameos the phase-adjusted rate settled at 7.3. A raw total tells you how much happened; a phase-adjusted rate tells you how much will hold. I ignored the desk’s request for a quick preview and gave the model three nights — I built that index three times before it stopped lying to me. When KKR stalled through the middle overs against Mumbai Indians in the playoff, the model had already flagged it. The editor ran it as a sidebar. The lesson was that populated data without phase correction is still incomplete — so a story built on empty data is incomplete by an order of magnitude.

The second case concerns a borrowed metric. On 1 July 2026 in Moscow, Spain completed 1,029 passes against Russia, held 74 per cent possession, and generated 0.8 xG. The match finished 1-1 and Russia won on penalties. Possession without penetration is noise. That same year I tried to push xG and PPDA into cricket’s structure — expected runs per phase, pressure without wickets, fielding-intensity proxies. xG borrowed into cricket logic and came back with an accent: in football a pass is an event, in cricket a ball is an event whose value is set by wickets in hand and overs remaining. Since 2026 the Impact Player rule has rewritten team composition itself; a model built without that distortion is a new letter written in an old language.

The biggest danger of an empty payload is analytical rather than cricketing. A complete eight-dimension frame sitting on zero evidence is exactly the pair that breeds hallucination. A filled template reads like proof; a figure with three decimal places reads like truth. In the risk matrix, five of six channels were marked unassessable, and the one channel that could be rated was this — credibility contamination. A model that cannot embarrass itself cannot protect anyone else either.

There is a commercial layer to this, and in the current cycle it is the loudest one. Transfer windows and auction cycles produce dozens of “sources say” claims a day — who will be bid for, who will be retained, which agent was spotted in which city. Most carry no information point at all: no date, no document, no confirmation. Franchises bid on narrative, and newspapers turn that narrative back into news. Surplus value lives in the gap between credit and contribution, and that gap can only be measured when the data exists. Auction analysis built on an empty payload stops being a guess and becomes mispricing advice.

Now the other side. Saying “insufficient information” is not a virtue in itself, and I see that risk most clearly on my own desk. In 2026 I bolted distance-covered pressing proxies onto the xG dashboard, the desk cut them, and the explainer ran late. The model was overbuilt and out of time — correct, and useless. A null result carries the same temptation: dressing a broken pipeline in the robes of principle. So the null needs a band too — “pipeline failure, high confidence” and “the article was genuinely empty, low confidence” are different claims, and merging them into one sentence is sloppiness. The second point matters more. Abandoning the eye test as evidence would be my largest error. The regional commentary booth usually catches the pattern first — the drift on a spinner’s ball, the backlift of a batter. The model confirms later; it rarely discovers. Write what the eye saw as a hypothesis with a test attached, never as a verdict.

One rule is worth adopting from here. Any Stage-1 output with fewer than three information points and no named entity should be auto-rejected and logged as a failure, with the changelog kept public — which cutoff was tried, which was discarded. The next time a desk declares some young player the next big thing, ask how many information points sit inside that claim. If the answer is zero, it is not a story. It is an empty file waiting to be filled.