The Feed Returned Zero: An Audit of Cricket's Information Pipeline

At 11:40 pm the television was still showing a match, but my content-management screen was a wall of blanks. Title: N/A. Source: N/A. Type: Unclassified. One-sentence summary: empty. Information points: empty list. Time sensitivity: not assessed. Source quality: "to be determined from the source field of the information points" — except there were no information points above it. One cell held text: Domain Label: cricket_world.

A match was happening; the system built to give it language returned zero. I have seen gaps in scorecards before — rain, power cuts, a scorer's typo. This gap was different. The event had not failed to occur. The instrument built to catch it had gone silent. An empty record is never the story, but the reason for the empty record always is.

Context: from scorecard to query

In the 1980s, cricket news reached Bangladesh in three layers: a shortwave bulletin, a phone report from an agency, and a printed scorecard that arrived two days late. Whatever that scorecard omitted, nobody ever knew. The print desk died the day I learned to query the match. Since then, a match has been a set of rows — each with a primary key, a timestamp, a source tier.

Today's pipeline has six stages, and each fails with its own signature. Capture (ball-by-ball events, ball-tracking, workload sensors). Parse (HTML or JSON pulled from a site or API — where most deaths happen: JavaScript not rendered, a paywall, an anti-bot response returning 403 instead of 200, and a zero-length body enters the pipe). Extraction (claims and information points pulled from raw text). Schema validation (a field name mismatch or an enum value outside the whitelist and the whole record is dropped). Classification. Publication.

Core: four faces of zero

When the type is "Unclassified" and the summary is simultaneously empty, the likeliest explanation is not classifier failure but zero-length input. A working piece of text is never left entirely unclassified — weak text still lands in some bucket. A zero-length body does not. If the fault is the classifier, the fix is a better prompt. If the fault is the input, the fix is the crawler. Misdiagnosis costs time, and time is the only asset journalism cannot recover.

Source tiering. Tier 1: documents — ICC and board scorecards, match referee reports, selection press releases, annual reports. Tier 1.5: vendor data — ball-tracking, pitch maps, GPS workload wearables, broadcaster speed-gun readings. Tier 2: named-agency reporting. Tier 3: unnamed sources — not useless, since many cricket truths surface here first, but never usable as Tier 1. Tier 4: social clips — proof an event exists, never proof of what it means. A transfer rumour is just a row waiting for a primary key; without club, fee, term and agent commission, the row will not join any table.

Time sensitivity. Team announcements before the toss. Franchise auction windows — IPL in December, BPL in January–February. ICC rankings: weekly refresh plus the annual recalculation in May, so comparing across May means comparing two baselines. Board election cycles: Bangladesh's 2026 election in October, similar cycles in Pakistan and Sri Lanka, each rewriting what "long-term planning" means. Injury windows. Weather and time zone — dew, humidity, cloud cover, all of which govern ball behaviour.

June 2026 was the month the crowd became a control group. Closed-door matches taught us how much of home advantage was crowd noise and how much was travel and rest. The variable you think you cannot control is often simply the variable you have not measured.

The entity graph. A player's name never travels alone: agent, franchise contract, central contract, sponsor, selection committee. A claim that a fast bowler is being rested may rest on injury, workload management, board-franchise pressure, or tactical experiment. Each has different evidence. When someone collapses the four into one explanation, they are telling a story with no primary key.

Workload. I ran the first xG audit because the eye test had no receipts. In October 2026 I published the shot map after Tottenham 4-1 Liverpool: Spurs 1.5 xG, Liverpool 1.7. Two colleagues said xG was a spreadsheet for people who cannot watch football. I kept the receipts. Nine years on, at 57, I still enforce the same rule on my own copy: no claim without a number. Cricket workload needs five measurements — overs and balls by format, days between matches including travel, the acute:chronic load ratio (7-day load against 28-day average, associated with injury risk though association is not causation), pace by spell, and line deviation on the pitch map. Together they permit a pre-registered forecast: if a given fast bowler plays three Tests in 21 days, his first-spell average pace in the third will fall by at least two kph on the same vendor's gun. That sentence can be proven false. "He looks tired" cannot.

Scorecard versus reality. A side wins by five wickets with a final-five-over strike rate under eight; the opposition loses two wickets but trails on run rate all innings. Run margin is an outcome, not a performance. Outcomes go in the table; only performance forecasts the next match.

Lord's, 14 July 2026. The match tied, the Super Over tied, and the winner was decided by boundary count. What the players played was a game; what was announced was a rule application. Tournament rule architecture is not a technical detail; it is a factory for producing outcomes. The ICC later changed the rule — an admission that the rule, not the game, was the problem. The same lens applies to governance: a board that withholds injury updates is not protecting selection freedom, it is managing an information market.

The information economy. Broadcast rights determine not only who sees the pictures but which data exists. A series without ball-tracking will have weak analytics — a capital decision, not a technical limit. Franchise leagues turn players into assets; an auction price is a valuation, not a measure of quality. Agents manufacture prices; their briefings are not evidence.

When the pipeline is empty. Admit the zero as a result. Diagnose the failing stage from the raw response — status code, body length, rendering, paywall. Return to Tier 1 without wasting time. Label estimates as estimates. Keep a ledger of which claims were verified and which later failed. The old rule survives in digital dress: look, then write — except that "looking" now includes understanding the pipe.

Contrarian: the crisis is not the empty feed, it is the empty explanation

Cricket's data volume has multiplied over a decade; analytical quality has not kept pace. The problem is not collection but joining. The most frequent failure is the join key: a player, a board announcement, a franchise contract — three databases with no shared primary key. Get it wrong and you write the wrong story, then decorate it with numbers. That is more dangerous than having no data at all, because abundance supplies its own alibi.

We panic at empty feeds and shrug at bad interpretation. The first is a process fault: visible, fixable. The second is a professional fault: invisible, contagious. The journalist who will admit missing data is reliable; the one who will not admit misread data may be popular, but he is not reliable. Sochi was not a defeat; it was a dataset with a cold press box. I filed Germany's exit before matchday three in 2026 because four years of tracking showed a pattern. Had that tracking been empty, I would have published no forecast — I would have written that there was no data, therefore no claim.

Takeaway

Three questions stay in my notebook before the next series. Does this series have ball-tracking? If not, I publish nothing on swing, seam or pitch profile. At what tier are injury and workload facts surfacing — board release or agent hint? And when is the next election or auction cycle, since no comparison across that date is valid?

One proposal, on the record: a null register for every tournament — a list of what was not disclosed. If one board's null register runs twice as long as its neighbour's, that is not a journalism failure. That is a governance signal. Who publishes the list next season?