Null Result, Filled Template: The Real Cost of Empty Data in Cricket Analysis

21 May 2026, Rajiv Gandhi International Stadium, Hyderabad. Mumbai Indians had collapsed to 129/8 and the tea was going cold on a four-person analytics desk in Andheri East. My death-overs wicket-probability model had given Mumbai 58% at the innings break. The desk's collective number was 44%. Krunal Pandya made 47 off 38, Steve Smith's Rising Pune Supergiant stopped at 128/6, and Mumbai won by one run.

My reward for that night was a senior colleague asking a former Ranji player to "sanity-check the girl's numbers." I stopped arguing with narrative and opened the Ledger instead — a private spreadsheet logging every model call against its outcome and the reason for the error. I kept the ledger for eleven years; the final still came down to one run.

Last week a different kind of document landed in front of me. Its title field was empty. Its source field was empty. Its format was undeclared. Its information points numbered zero. And it carried eight sections — everything from format analysis to industry transmission — with a single line in every cell of every section: N/A.

The document was not wrong. It was honest, which makes it a strange minority in this industry. It did not say who would win, whose strike rate would fall, whose purse would open. It said: there is no evidence, therefore there is no analysis. In a trade where every template must be filled, returning an empty one is itself a statement.

Over the next four days at least four people asked me the same question: "So what do we take from it?" That question is the centre of the whole affair. When the input is zero, the correct answer is "nothing." Cricket's media ecosystem does not pay for nothing.

Cricket is possibly the only sport that already owns a vocabulary for the null result. A rain-ruined match is No Result. A tie is a different species. The points table gives them separate columns, net run rate treats them differently, and a Duckworth-Lewis-Stern revision is a third event again. On the field, cricket accepted long ago that "there is no answer" is still an answer.

In the studio we never accepted it. A night's cricket coverage has no No Result column in the rundown. The ten o'clock bulletin has a three-minute slot and that slot must be filled. This is precisely where the data desk and the commentary box part company — DRS, Impact Player usage, death-over matchups are settled in the analytics room and then re-litigated by instinct on air. The studio is the last un-modelled variable in Indian cricket coverage.

Now the document itself. One thing was present: a domain label, cricket_world. Everything else was absent — no title, no outlet, no dateline, no team, no player, no venue, no format. Zero information points. By the framework's own rule, every conclusion must trace to a numbered information point. Trace what?

An information point is the atomic unit the whole method stands on: "the second slip came out in the eighth over," "yorker usage at the death fell from 34% to 21% this season," "after retentions the purse held 18.5 crore." Those sentences are checkable. Without them, what remains is opinion — and I don't call opinion analysis.

The document called itself framework-complete, content-null. I think it is one of the more professional artefacts to come down the cricket data line in years, and I also know that if it had reached a producer's desk, it would have been filled in inside three minutes and sent back.

Break it open. The engineering cause is obvious: the domain label survived extraction, the content fields did not. That does not mean the article was empty. It means the source text was under roughly two hundred words, or paywalled, or video/embedded content the extractor could not read. This is an extraction failure, not an empty article. The difference matters: one is fixed by a re-fetch, the other means there was never a story.

The second cause is institutional. In cricket, a feed item, a headline, an image caption all circulate as though they were stories. Building an eight-dimension analysis on one is like reading a pitch off the toss report. The third cause is commercial: if your desk does not have the right to return a zero, your desk will not return a zero. It will fill.

And that is where the thing I fear most begins — confident, complete, entirely invented cricket intelligence produced under deadline pressure.

The most likely output of an empty input is a lie, because a lie is the only output that fills every cell of the template.

The cleanest example is format contamination. Cricket's three formats are not comparable on numbers, and the point is basic enough that it bears repeating. A 140 strike rate is ordinary for a T20 opener, extraordinary in an ODI, and close to irrelevant in a Test. An economy rate of 8 is acceptable in T20 and destructive in a Test. When a feed item does not say which format the match belonged to, every tactical conclusion built on it is contaminated — however clever it sounds.

My ledger holds many entries of this kind. The most expensive is dated March 2026. By then I had 1,300 back-tested matches and a home-advantage coefficient worth +4.2 percentage points of win probability. Sport stopped, the IPL ran in the UAE from 19 September to 10 November 2026, Mumbai Indians won, and the coefficient went to zero. Where nobody has a home, home advantage cannot exceed zero. I rebuilt around dew, toss and travel, and found that chasing at night in Dubai was worth roughly +8%.

Every coefficient has a climate. Mine died in the desert, and I stopped trusting backtests.

That is why every piece I write now opens with an assumption audit: what would have to be true for this to be wrong? With that question at the top of the page, filling an empty input becomes harder. If I cannot say what I am standing on, the number the model produces is not a number, it is a voice.

Small samples sharpen the same problem. The Women's Premier League has existed since 2026. A "knockout average" off three seasons is a curiosity, not an analysis. Knockout cricket hands you three-match samples, and one good innings lifts a number by forty per cent. My rule is plain: if the gap sits inside the error bars, I report the error bars, not the gap. A hot market plus a small sample is the pairing that turns somebody into a star every season, and the same star fails to justify the price the next one.

Now the part that does the damage: numbers built on invented inputs reaching decision-makers. Consider an auction-purse model that carries retention cost and remaining purse but has no input on where the player sits on the age curve. It will produce a clean figure — "12.5 crore is fair value." The figure will be elegant. It will be wrong. Its elegance is what hides the error.

Playoff qualification makes it plainer. A side needs 12 points from 14, net run rate is tight, two games remain. The model says 61%. What is that 61% resting on? Rain probability, the toss, the assumption that the same bowling attack turns up that turned up in the last four games. Lose one of those inputs and 61% stops being information and becomes an ornament.

I have an eleven-year ledger, and it says one thing clearly: a bad bet and a bad result are different species; the ledger knows the difference.

Through the 2026 qualifiers I built a possession-adjusted chance-creation model across 40 national teams. Croatia ranked fifth in my table while books priced them at 25/1 and 33/1. They reached the final in Moscow and lost 4-2 to France. The same model placed Germany 14th and flagged their full-back exposure; Germany went out in the group stage for the first time since 2026. The piece drew 11,000 reads, and in August I took a seat on the Mumbai pricing desk of an Isle of Man-licensed operator.

Croatia sat at 33/1 through the qualifiers. My model said the market was asleep.

That gave me a format I have never changed: the price, the disagreement, the reason. Every article names the closing line I believe is wrong and the number I would have set instead. It makes the work checkable. It makes editors nervous. Both effects are acceptable.

The same ledger records that my best essay after March 2026 went out three weeks late and cost the desk a client pricing window. Two rules entered my work that day. Open with an assumption audit. Ship at eighty per cent. I still fail at the second. The difference is that I now admit it in a footnote instead of an apology email.

So let us price the value of that empty document. It is valuable because it raises a question the cricket data line rarely asks: do we have the right to declare No Result on our own work? On a desk where a take must be filed every evening, that right exists on paper and disappears in practice.

The biggest analytical failure is never the wrong model; it is the filled template somebody submitted with no input at all.

One unsentimental observation belongs here, because it applies to my own work. The money that has poured into cricket over the past few years — broadcast bids, franchise valuations, auction purses — came out of the same habit that produces a filled template: the rush to declare a number when no number is available. What gets paid for rights is often not an answer to how many people are watching; it is an answer to somebody needing a figure by ten o'clock. Platforms buying rights and losing money are repeating old television's mistake, and that mistake was never about the model. It was about the input.

I have to testify against myself here. I have filled templates. In 2026 I had fifteen minutes to say something about a franchise's retention list and no confirmed source on a player's recent injury. So I talked about the age curve instead. It sounded fine, because the age curve is a respectable concept. What I did was substitute a concept for information. The ledger entry reads: "Input missing; take was good; that was the problem."

The other risk that comes out of all this is contrarianism hardening into a brand. Being right against consensus once is intoxicating. Doing it reflexively is just another prior. My rule is that the disagreement must clear a stated threshold; if the edge sits inside the error bars, I say so and drop the piece. The rule lowers my output and raises my hit rate. I will take both.

There is a trap on the other side too. It is easy to say "there is no data, so nothing can be said" and turn that sentence into a philosophy. That is as lazy as inventing a conclusion. A null result is a temporary state, not a destination. The professional response is: state what is missing, state how to retrieve it, then retrieve it. The document did the first two, with a clean list of what a re-extraction must return. The third belongs to someone else, and it should happen.

In my experience these null results are cheap to fix. If the source was a short news item or a headline, one clean pass will usually recover the title, the outlet, the date, three to five information points and named entities. Then all eight dimensions can run. This is a pipeline repair, not an analytical failure — and a pipeline repair is cheaper than an analysis. I say that from a desk where rebuilding one coefficient took four months.

Forward, then. The next time a graphic tells you a team's qualification probability is 61%, ask the question the commentary box almost never asks: what was the input, and what was missing from it? If the answer is that the format was never logged, then 61% is not a number. It is weather.

My next watch item is what I call the absence signal — how often an empty source field, a missing timestamp and an undeclared format arrive together in the same document. My estimate is that the fastest-moving stories of the coming season will carry all three blanks most often. If I am wrong — if analysis built on unsourced items proves as accurate as analysis built on sourced ones — I will write that down, because the ledger will carry the entry either way.

And if you take one sentence out of this piece, take this: the spreadsheet never flinches, which is why I let it speak before I do. An empty spreadsheet speaks too. It just needs a desk willing to sit through three minutes of silence on the ten o'clock bulletin.