Empty Feed, Full Scorecard: When 'No Data' Becomes 'No News' in Cricket

The Lord's scorecard of 14 July 2026 does not read like a sentence about cricket. It reads like a receipt. Tied after fifty overs, tied after the Super Over, decided on boundary count — a World Cup final settled by which set of batsmen found the rope more often. The same night produced a second entry that appears in no scorecard at all: the ball that deflected off Ben Stokes' bat and ran to the boundary, scored by the official as two run plus four overthrows, six in total. That was a judgment call, half rulebook and half scorer's habit, and it bent the direction of the most-watched match in the format's history. I have spent two decades hearing people say the scorecard tells the truth. The scorecard tells a summary, and summaries are written under deadline pressure, inside one scorer's conventions.

The real danger in this trade is not a wrong number. It is an empty number that looks clean. Last week a desk exercise put a file in front of me that was technically complete and substantively hollow — no title, no source, an empty list where the facts should be. A junior could easily have filed it as 'no news.' That is the subject of this piece: in cricket's information system, the most expensive error is not a bad statistic but an absent one read as a finding.

The biology of a feed

Cricket data travels through five layers. On the ground sits the scorer, coding each delivery with his eye and his experience. Above him the feed provider maps those codes into a schema. Then the aggregator — Cricbuzz, ESPNcricinfo — republishes. Then broadcast attaches graphics to live commentary. Last comes the reader's memory, the most costly layer because it glues itself to any summary and never audits it again.

Every layer runs a filter, and every filter fails in one of three ways: it drops something, it alters something, or it fails outright and fills in a default. The third is the dangerous one because it is invisible. When a page renders through JavaScript and the scraper returns nothing, the system does not say 'no facts.' It says 'unclassified.' To a reader, unclassified and ordinary look identical. In preview language, unclassified quietly becomes 'no particular signal' — when the truth is that the signal never arrived.

The least-inspected layer of the whole chain is the first one: the scorer's decision. My entire method rests on one sentence. If the scorer's call changes, the conclusion changes — so audit the call before the conclusion.

In 2026, building a private ball-by-ball ledger for all sixty IPL matches on Cricbuzz's Delhi data desk, my first task was not modelling but reconciliation. I cross-checked more than fourteen thousand deliveries against broadcast records — phase, venue, batter-bowler matchup, wicket probability. The desk wanted quick takes; I handed over a twelve-page methodology note first. Its most contested section was the error log, not the findings.

The ledger is not neutral ground

This is the thing I learned latest and used most. Anyone who opens a ball-by-ball feed feels they are reading raw truth. They are not. Every ball is a human call — byes, leg-byes, wides, dropped catches, and the most elastic word in the vocabulary, intent. A delivery that hits the pad is leg-byes to one scorer and a miss to another, and that difference surfaces three overs later in a run-rate graph. A run-out is credited by convention to the thrower or the fielder who missed, and conventions are not physics.

The consequence is that any conclusion drawn from the ledger has to carry its error rate alongside it. Every long analysis I publish names where the number came from and what that feed is known to get wrong. When someone says a batsman made 42 off 61, my first question is not about strike rate. It is how many of those 61 balls came in the Powerplay, what the required rate was through the middle, and what the match state was. Without phase context, a strike rate is decoration, not evidence.

An illustration. Suppose a batsman makes 42 off 61 and his side reaches 148 in twenty overs. The scorecard says: slow innings, cost his team. Now split the 61 balls. Eighteen came in the Powerplay with two wickets already down and the required rate at 7.1. Thirty came through the middle against spin from both ends, a phase in which the team scored at 6.2 an over. The last thirteen came at the death, where he scored 21. After that reconstruction, the same 42 (61) is no longer a slow innings — it is an innings that protected wickets through the squeeze and changed gear at the end. One number, two stories. The only difference is the phase split.

In 2026 that method took a step forward. Studying all sixty-four matches of the Russia World Cup, I tried to translate football's expected-goals logic into cricket as expected runs and wicket probability, then tested it on 720 Powerplay overs of the 2026 IPL. What emerged was that 'possession' has no direct cricket analogue, because control of the ball alternates by design; in cricket the scarce resource is not possession but wickets. My first draft ran two weeks late because I kept refining calibration, until my editor imposed a cutoff. I then set myself a two-day modelling deadline and a rule that a colleague sees the calibration file before anything is filed.

Five shapes of an empty feed

From that experience comes the checklist I run on every new data source. First, missing metadata: no title, no source — the signature of JavaScript rendering or a paywall. Second, a minimum-length threshold, where the extractor treats short copy as noise and returns nothing. Third, default values: 'unclassified' in technical language, 'ordinary' in the reader's. Fourth, a silently swallowed exception — the parser failed, the log never showed it, the output looked fine. Fifth, the diagnostic one: every metadata field nulling at once. When title and source vanish together, the conclusion is not that the story was trivial. It is that the pipeline never started.

This is where cricket journalism usually slips. An empty feed is more dangerous than a loaded one, because a loaded claim can at least be disproved. Nobody challenges an absence. It converts quietly into 'no news,' and the next day an editorial builds an opinion on top of the gap. Previews are the worst offenders: 'this side's Powerplay scoring has dropped across the last three matches.' The question worth asking is how many of those three feeds arrived complete.

2026, empty stands, and a thesis that failed

When the IPL moved to the UAE in 2026, I had something rare: a natural experiment. Sixty matches, more than 7,200 balls, no true home venue, no crowds. I built a conditions-adjusted model separating designated home win rates, toss impact, and dew-adjusted death overs. The designated home side won below the five-season baseline. That is the number every commentator picked up that year — look, no crowds, home advantage is dead.

The problem was that the toss effect had not disappeared; it had moved. In my model, toss impact tracked dew rather than crowd pressure: a wet ball reduced spin grip in the second innings and made the chase easier. Mumbai Indians won the title, and Jasprit Bumrah's 27 wickets became a clean test case, because his contribution lived in death-over economy rather than in a wicket count. That model became my auction baseline, since it let me compare players across conditions rather than on raw totals. Venue, dew and team form must be separated in every preview — and the output should be a recovery path, not a slogan about home advantage being dead.

The contrarian side: do not call the outcome a plan

Here is the objection I have to raise against my own work. Writing with the result already known, every wicket starts to look like the execution of a plan and every boundary like the release of pressure. The bias runs easily alongside data, because data will always cooperate with the story you already have. My defence is to judge each decision only on pre-ball information — the field at that moment, the bowler's previous over, the required rate as it stood — and then to publish the balls where the plan visibly failed. Without that list, analysis becomes publicity.

A second, more counter-intuitive lesson follows. The riskiest number in cricket analysis is the empty one, because nobody interrogates it. A wrong statistic starts an argument; an absent one produces silence, and silence gets read as agreement. In the UAE study I had to break my own thesis once — 'home advantage is dead' looks right until dew is separated out, and then the same data says the opposite.

So I write the null hypothesis down before the analysis: if the conventional read survives, the conventional read gets published. Sometimes it does. Boring findings are credibility deposits, and without that deposit no surprising finding holds up.

What to watch next

The next real improvement in cricket data will not be a better model. It will be schema validation — a rule that rejects any output whose fact list is empty or whose title and source are both missing. Until that rule exists, every data desk needs a small habit instead: label an empty file 'the story never arrived,' not 'no news.' And the next time someone tells you a team's Powerplay has fallen away over three matches, ask one question. Did all three ball-by-ball feeds come through complete? If the answer is that nobody checked, you are not reading a statistic. You are reading an empty bracket.