The Null Result: Cricket's Data Desk and the Report That Was Filled In Everywhere Except the Evidence

The Null Result: Cricket's Data Desk and the Report That Was Filled In Everywhere Except the Evidence Kazan, 27 June 2026. Germany took twenty-six shots at South Korea; South Korea took eleven. The scoreboard read Germany nil, South Korea two. My expected-goals sheet that night was completely full — a number in every cell, a ball-by-ball timestamp behind every number, a location map behind every shot. The sheet was not lying. The sheet simply was not answering the question I was asking. Four days later, the first of July, at Luzhniki. Spain completed one thousand and twenty-nine passes — at that moment the highest figure in World Cup history — and lost to Russia on penalties. My sheet read Spain 2.4, Russia 0.6. Full again, useless again. The one thing that worked in Kazan was passes per defensive action: Russia's 28.4 told us Russia was not defending, it was passive. A passive team's expected goals look low; it is not that the chances were absent. A filled sheet and a true sheet are two different organisms. I learned that in Delhi, not Russia, eighteen years earlier. In April 2026, after Cronje's confession, and in December 2026 after the CBI report that ended Azharuddin's career with a life ban, I was a twenty-six-year-old scorer-statistician feeding ball-by-ball data to Cricinfo from a cramped vendor office in Connaught Place. I pulled eleven seasons of one-day scorecards — one thousand and forty-seven matches — and built a collapse index. Eleven innings were flagged anomalous. Two flags died under scrutiny: a rain-reduced chase, and a No. 8 batter playing with a broken finger. I retracted both in print, by name, in the same column. This week the reverse of that column landed on my desk. Twelve fields. All twelve populated. All twelve saying the same thing — insufficient information. Title insufficient. Source insufficient. Article type unclassified. Summary blank. The list of information points blank. Named entities not identified. Time sensitivity not assessed. Source quality not graded. One live signal survived the entire packet: a taxonomy tag, cricket_asia. I will not call this match reporting. It is an audit. In my experience, such an audit is worth more than a match report, because a match's truth arrives eventually; a dataset's truth never arrives if you have already read an empty cell as a zero. Context: the information supply chain and where it breaks A fact in cricket travels through three stations. First, the field: a ball is bowled, a bat swings, an umpire signals. Second, the data desk: a scorer turns it into a number, an analyst cleans it, an editor distributes it. Third, the market: broadcaster, franchise, auction table, selection room, fantasy app. The pipeline we are discussing sits between the first two stations, and it has two tiers. Tier one is deconstruction. The source is read, its facts are broken apart, each fact is turned into a separate citable sentence, and every name — team, player, board, league, date — is placed in a separate list. Tier two is analysis, and it runs across eight dimensions: format and match, player technique, team landscape, league and commerce, rules and governance, risk, market expectation, industry transmission. Tier two carries one contractual condition, and that condition is the root of this entire matter: every conclusion in tier two must trace back to a numbered information point from tier one. The information point is the currency of this trade. No coin, no shop; no argument. Which is why, when tier one returns empty, tier two has only two routes. It can declare analysis impossible. Or it can quietly invent. The second route is easier, faster, and the reader will never know — because fabricated evidence is indistinguishable from real evidence. Anyone who has sat on a data desk should understand that this is the most serious failure mode in the workflow, worse than an incomplete report. An incomplete report tells you something is missing. A fabricated report convinces you something has been found. There is a further layer. In 2026 the currency is being coined by the market, and the market is currently in transfer-window mode. The auction cycle, retention rules, agent manoeuvres, board-issued NOCs, Future Tours Programme congestion — hundreds of claims circulate daily, and behind most of them there is no document. Empty cells are most dangerous in this season, because this is the season in which stories are manufactured hardest. The tag is an address, not content cricket_asia was the only signal left in the packet. A classification label. It says the subject sits inside Asian cricket. It does not say what the subject is. Beneath that tag sit more than eleven boards, at three different tiers. India, Pakistan, Sri Lanka — top-tier power. Bangladesh, Afghanistan, the Asian edges of global cricket — middle tier. Nepal, Oman, the United Arab Emirates, Hong Kong, Malaysia — associate tier. A tag that puts all three in one pocket gives you an address, not a direction. Which is why every inference beyond the tag — suppose it is an Asia Cup story, suppose it is an IPL auction, suppose it is a PCB governance dispute — is an inference, not a conclusion. In my trade we have a name for this error: category drift. The analyst reads the tag, builds the whole story in his head, and then selects the numbers that confirm it. This is the same error that happens daily in cricket journalism, when a writer decides whose story it is first and collects favourable data afterwards. Starting from the tag and arriving at numbers is the wrong road. Starting from the numbers and arriving at a headline is the right one. The grammar of a null result: what an absence proves Language matters most here, because in cricket we merge three different things into one word. The first is a null result: the data required is not present. The second is a negative finding: the data was present, and it showed no signal. The third is a clean certificate: an investigation happened, nothing was found, therefore all is well. These three are as different as a Duckworth-Lewis win and a real win. My favourite cricket analogy is DRS. When a ball reaches umpire's call, the system does not say the batsman was not out. It says the system did not have enough evidence to overturn. The broadcaster sells it as a reprieve, the crowd reads it as a let-off, and in reality it is the system confessing its own limit. That confession is on record in this packet. The governance level could not be fixed because no governing body was named. A risk rating was withheld rather than assigned, because assigning one would imply a check had been performed. No piece of evidence was found missing — the entire scaffolding of evidence was missing. That difference is enormous. Treating an absence as a certificate of cleanliness is the oldest habit of the cricket data desk. When I built the collapse index in 2026, the easy route was to quietly drop eleven of twelve flagged innings and keep one. Eleven anomalies tell you the model's threshold is broken. One tells you something happened. The first story protects the model; the second protects the truth. I kept eleven and retracted two by name — a rain-reduced chase and a No. 8 with a broken finger. The model failed there. The column in which I printed that failure was not my most read. It remains, to this day, the one I consider most correct. The auction model: two of three calls held, and that proves nothing Mumbai, February 2026. The inaugural IPL auction. I built a replacement-level valuation model for one of the eight founding franchises. The training data came from two sources — ball-by-ball data from the 2026 T20 World Cup, and two seasons of the Syed Mushtaq Ali Trophy. My sheet said Chennai's $1.5 million for MS Dhoni was fair. It said Kolkata's $950,000 for Ishant Sharma was a forty per cent overpay. And it said Rajasthan's $475,000 for Yusuf Pathan was the best value in the room. Yusuf Pathan was Player of the Match in the 2026 final. Two of my three calls held. Now the part nobody prints: a sample of three is statistically worthless. Two out of three looks excellent, but it is no better than four out of six and no worse than thirteen out of twenty. I wrote that in my own notes that year and have repeated it many times since — an auction's real test is the next auction, not this one's final. One thing the sheet did read correctly, and it was about the market rather than any player. That year the purse structure determined that scarce positions — wicketkeeper-captain, off-spinner, finisher — would sell high, and a player who was alone in the market had his price set by scarcity, not by performance. Every auction since has repeated the pattern. A price is only a rumour until the data agrees. And the data agrees twelve months later, not that evening. I write all this so a reader understands that the first job of an information pipeline is not to manufacture headlines. It is to block them. The greatest achievement available to a model is to say: I do not have an answer to this question. Three levels: entity, price, sample Here is the protocol I run on every cricket claim, and it is not elegant. It works. Level one, entity. The claim must carry a name. Team, player, board, league. Pronouns will not do, because intent hides behind pronouns. The packet that produced this article listed the entity field as: identify from the points above. And there were no points above. It sounds absurd, and it is the daily condition of cricket data: an instruction issued to one desk survives an instruction unfulfilled, and then walks into the next report disguised as a fact. Level two, price. Every cricket claim has a price, whether in rupees or in wickets. An auction fee, a contract length, a retention slab, a match-fee split, a broadcast bucket. If there is no number behind the claim, the claim is marketing rather than sport. Level three, sample. How many attempts, how many matches, how many seasons? When a young batsman produces a hot strike rate across two T20 league seasons and a newspaper finds the next Tendulkar, this is the level I apply. Two seasons is form. Six seasons is a trait. This is not merely an editorial distinction; it is a selection-policy distinction, and a selection-policy distinction runs for ten years on the field. On all three levels this packet stands at zero. No entity, no price, no sample. And that zero must mean zero. The most dangerous thing is the aggregate: why an empty packet must not be counted as zero Suppose the empty packet travels quietly downstream. It is logged in a database. Someone builds an annual report: how many rule-violation signals were detected this year? The system counts zero. Anti-corruption alerts? Zero. Governance disputes? Zero. The count is not lying. The count is counting something it never saw. Counting an empty packet as zero is like treating a rained-off match as a completed one and adding its statistics to career records. In cricket we do not do that, because Duckworth-Lewis stops us. On a data desk we do it daily, because there is no Duckworth-Lewis there. One thing was clear in this packet: the item needs a flag that says the case has been excluded from analysis because its base was empty. It belongs in the unverified category, not the verified-zero category. Until that distinction stands, cricket's big aggregates will remain dishonest — whether they belong to the ICC, to a board, or to a broadcaster. In my reading, the real risk in this whole matter is procedural, not in the risk matrix. The six risk families — sporting, personnel, commercial, rules and integrity, public opinion, systemic — could not be populated because there was no material to populate them with. And that is the packet's most honest feature. Had the matrix been filled in anyway, we would never have known. The same old error, in the selection room This looks like a sports-journalism debate. It is a selection-room debate. India's second XI, as I keep writing, is the hardest team in world sport to pick, and picking it is an optimisation problem, not a talent problem. When a senior player is rested and a youngster plays in his place, fans call it a discovery. I call it a management decision in which opportunity cost is the largest unknown. How is that opportunity cost built? In the data the meeting sits on, the youngster who played the rested fixtures appears at the top of the run list. The senior, who played the hard fixtures, appears with runs earned in difficult conditions. A direct comparison between them is impossible, and that impossibility sits in the report as a zero. Then an official says: look, there is no comparable number for the two, so we go by current form. This is the exact error recorded in this packet. Where a basis of comparison is missing, we either invent the basis or we invent a zero. Of the two, the zero is cheaper. There is only one measure of how expensive that is for cricket: checking, twenty years later, how often the discovered player actually performed the role the first-choice player performed across twenty matches. Nobody's desk holds that answer today, and it is not one board's folly. It is the system's habit. Access capture: and this is where I distrust myself At fifty-two, with six career turns behind me, one dirty truth has to be admitted. In my profession the most valuable asset is not the statistic. It is the phone number. Board officials, franchise owners, agents — conversations happen, and the implicit price of the conversation is often that difficult questions are asked less often. The cure is not a moral lecture. It is a line. A standing rule is pasted into my notebook: no claim from any board or franchise runs unless it has an independent second confirmation. And in any report where I use such a claim, I state what my relationship with that side is. This empty packet also reminded me that second confirmation matters less for verification than for restraint. Until the second confirmation arrives, you have to wait. My instinct wanted a decision; the data wanted a delay. Delay is the most unpopular act in cricket journalism, because every delay costs a scoop, and every scoop is a win in a broadcaster's eyes. There is an uncomfortable question I have to raise about my own trade here. A very large share of Indian cricket coverage covers the BCCI or works for it; that question has been hanging since 2026 and has never been answered. Until a public grading of sources exists, a reader cannot know whether he read a report or a translation of a press release. In this packet the source-grade field was blank. Of the twelve blank fields, that is the most expensive. The contrarian side: my own reflex is the greatest danger here Now the hardest part, because it requires me to testify against my own tool. For thirty years I have been read for the thing others are scolded for — the call that runs against the eye test. That habit has a side effect: a belief that the number nobody trusts is therefore the true one. It is a reflex, not reasoning. Applied to this packet, the reflex immediately suggests a punchy story: a piece whose twelve fields came back empty was probably hollow in reality. The temptation to file that piece is strong — cricket journalism sells headlines like the story that isn't there is the story. In this instance, that conclusion is wrong. I have a simple test, and I am publishing it so people can prove me wrong. If re-extracted data yields five or more entity-bearing information points, the null result dissolves entirely and the original article was real. That is entirely plausible. A document that returns twelve empty fields is often sitting on a page a parser cannot read — text rendered as an image, a JavaScript-built page, or a truncated input. In my experience, across roughly three-quarters of such cases, the probability lies with an incomplete extraction rather than an empty article. The second thing that makes me uneasy is that publishing a null result is a losing trade in the market. Nothing found reads as no headline. But if a data desk has any credibility value at all, it lies precisely in that capacity — the capacity to say I do not have an answer. A desk that cannot say it is not a desk. It is a publicity department. What to watch next cycle The route forward is clear, and it starts with re-extraction rather than template-filling. I set three minimum conditions: title and named source, a resolved article type, and at least five information points, each carrying an entity. Without these, whatever analysis gets published will be a copy of numbers rather than a report on numbers. And four signals are worth tracking in the next cycle. First, the count of information points in the re-extracted packet — three or more entity-bearing points and the analysis can proceed. Second, populated source and type fields — both filled and source quality can be graded. Third, format and date markers — Test, ODI or T20 plus a date unlocks the timeliness dimension. Fourth, any parse failure in the ingestion log — that will explain the whole null and prevent a repeat. If the underlying document turns out to concern governance or integrity — ACU, spot-fixing, an NOC dispute, revenue share — then the priority next time belongs to those dimensions, not match analysis. And then this null result, which looks empty today, will read in hindsight as a safeguard. Last word In Delhi, statistics was never mathematics; it was patience in disguise. This packet is a test of that patience. Twelve fields arrived full, and I sent them back the way they came — empty, with only the one error I could not commit. Anyone who believes the greatest crisis in cricket coverage is betting should first answer this: in the evidence column of your own last report, when was the last time anything in it was real?