The Empty Payload: The Night the Cricket Data Desk Got a Scorecard With No Score

The file arrived at 2:14 a.m. Eight sections. Eight headings. And under all eight, a single token — N/A, insufficient information. The title field at the top was N/A. The source field was N/A. And at the very bottom, where a single sentence normally carries the whole case, there was an empty bracket: Information Points: [].

I have been reading scoreboards at night for eleven years. A scoreboard never arrives empty. It can arrive incomplete, it can arrive wrong, it can arrive halted twice by rain — but never empty. An empty scoreboard means nothing, so nobody sends one.

That night, for the first time, a document arrived with everything in place and nothing inside. Reading it took me forty minutes, because the first twenty I spent looking for where the fault was. The fault was nowhere. Every section sat in its slot, every sub-heading in the right order, every checklist ready to be ticked. Only the thing that was supposed to be analysed was absent.

And that is the centre of the whole matter.

Context: how cricket's content factory came to be

Over the past eight to ten years the shape of cricket coverage demand has changed. A single year in India now holds 74 IPL matches, a domestic circuit of Ranji and Syed Mushtaq Ali and Vijay Hazare, the Women's Premier League, and on top of that bilateral series, the Asia Cup, the World Cup cycle. From outside: the BBL, South Africa's SA20, the UAE's ILT20, the Caribbean Premier League, Major League Cricket in the United States, England's County Championship. A preview for every tournament, a report for every match, a post-mortem for every series.

Cricket has one feature that even football does not carry at the same intensity — its calendar runs nearly twelve months, and on a single day three countries can be playing three formats at once. The value of one over in a Test, in a T20 and in an ODI are three different values. That difference is the foundation of the analysis. It is also the hardest thing for automation to hold.

This demand built an industry — the content pipeline. Extract text from a source, break it into information points, build an eight-dimension analysis on those points, convert it into Hindi and English, tag it, publish it. The pipeline makes economic sense because its marginal cost is close to zero. Once the template exists, a thousand articles come out of the same template.

And that is exactly where the thing that woke me that night lives.

A template lowers cost. A template does not add information. When a document has eight headings and not one fact under any of them, a reader cannot tell at first glance whether they are reading analysis or a blank form. The two look identical. Headings are bold, tables run in straight lines, every row appears to have a cell filled in.

I remember 2026, when I rebuilt every IPL venue's par score from 2026–17 ball-by-ball data — Chinnaswamy 189, Wankhede 172, Hyderabad 165 — the first thing that changed was that I stopped writing "batting-friendly surface". Every preview began to carry a number and a band: par 189, range 176–197, confidence moderate. Editors were irritated. Readers started quoting the numbers back at me. That season my mean absolute error on first-innings forecasts fell from 18 runs to 11.

Not once did a reader send back a document that contained no numbers at all.

Core: N/A is not a neutral word

First, the token that was repeated eight times has to be opened.

"N/A — insufficient information" reads as modest. It is not modest. It is a factual claim. It means: I looked, and I found nothing. That claim can be true in two very different situations — first, there genuinely was nothing in the source; second, the source was never reached. The distance between those two is the distance between earth and sky, and the document does not separate them. That is the root of the whole matter.

In engineering this is called fail-open. The system fails, but does not appear to fail. It returns a result that looks structurally valid — empty, but inside the frame. The opposite is fail-loud, where the system says clearly that it had nothing to work with.

Cricket has a direct illustration. If the scoreboard reads 0/0, a reader can work out that the match has not started or has been abandoned. But if the scoreboard does not display 0/0 and instead displays a full card with every player's column left blank, the reader's first assumption will not be that the data is gone. Their first assumption will be that the card is designed that way.

An empty form looks like a filled card. That is its entire strength.

Now the second layer. One line in that document stung more than the rest. "Entities Involved: identify from the information points above." The entities are to be identified from the information points above. And the information points above were empty.

This is not a small omission. It is a structural defect in the template. The instruction says to derive conclusions from material that does not exist. When the material is empty the instruction cannot be completed, yet it does not look incomplete — it sits there like a normal task.

Take an example from outside the sport that comes from inside it. In DRS, the third umpire is instructed to decide "based on the ball-tracking data above." If no ball-tracking camera was installed at that match, the protocol looks complete, except that one link is hanging in the air. And the only way to fill a link hanging in the air is for someone to fill it from their own assumption.

That is precisely where analysis starts sliding into narrative. Whatever downstream stage runs on a document with empty information points will have to supply players, teams, venues and dates from itself. And once you begin supplying from yourself, the document starts speaking. Then it does not read as a lie. Then it merely reads as imagination.

I saw a version of this in Russia in 2026. My base rate, built on four tournaments, was 0.26 penalties per match; Brazil 2026 produced 13. Russia produced 29. I did not re-base mid-tournament. That was stubbornness, and it cost me — my under bets on penalties and cards lost three straight in the knockout rounds.

But what came out of that loss was not money. It was a switch. I built a regime flag — a separate switch for changes in the refereeing environment, kept apart from the noise. It earned its keep in September at the Asia Cup in Dubai, where dew made chasing decisive and the 2026 toss data became worthless.

Now put that switch on the empty payload. The switch was not there.

There was not one line in that document saying: the source did not open, this is not analysis, do not forward it. The line that mattered most was the line least written. The document said plainly in three places that there was no analysable content in the input, and then in the same breath laid out the scaffold of eight dimensions.

This is the same failure I have seen many times in cricket's data desks, only now at the machine layer.

Take a bookmaker's live feed. The feed comes back empty. The desk reads it as "no news", and "no news" is itself a position — for the person holding the bet, that silence is already a stance. The real question is never what the feed reported. The real question is whether the feed knew, or whether the feed was never told. Those are not the same thing, and the desk's language has no separate words for them.

This is my oldest objection to the whole data industry. I have spent eleven years with cricket's numbers, and I think the most dangerous thing is not the wrong number that looks wrong. The dangerous thing is the empty space that looks filled. A wrong number can be pointed at. An empty space cannot be pointed at, because where would you point — there is no space there.

Third layer: cricket's numbers are the most checkable in sport, and that becomes an inverted weapon

Of all the world's games, cricket's statistical record is the densest. A batter's average, a strike rate, a bowler's economy, a catch-to-run ratio — these are so memorable that any reader can check one in a minute. I like this. It is the sport's best tradition.

But it has an inverse face, and that face was clear that night.

When a document carries not a single number, there is nothing to check. The reader has nowhere to put a finger. And even if a number is supplied, but supplied in the wrong place — a Test average dropped into an ODI column — the error escapes checking because the number itself looks right. Only the format is wrong.

Format context — Test, ODI, T20 — is the thing within which batting average, strike rate and economy rate carry no meaning on their own. Cross-format comparison is invalid, and the invalidity is so common that it no longer surprises anyone. When I read a number in an analysis, my first question is not the number's value. My first question is which format token is written beside it. If there is no token, then for me the number does not exist.

A number without a format is not a number.

In the same way, one more thing was missing from that document — sample size. How many matches? How many overs? Over what window? Without that, any rate is incomplete. In my own work I have made a rule of waiting for thirty matches before I move a rate, and of writing that wait down in the piece. The rule slows me down and wins me arguments.

That document had no sample, because there was no sample.

Availability-minded accounting, and the lesson that came out of it

From the 2026 auction beat I picked up one habit — I price players by matches available, not by highlight. Retention logic, purse efficiency, injury history, workload, and that awkward truth that India's hardest team to make is the second XI.

That night I felt that habit now had to be applied to documents as well.

The quality of a pipeline is not measured by how many articles it produced. It is measured by how many articles it produced, and how many of those contained the material the pipeline claimed they contained. Availability accounting.

A pipeline that produces 1,000 documents of which 30 are empty forms has a useful efficiency of 97 per cent. But if those 30 reach a channel where the reader is sitting with trust in the writer, those are not 30 — those are thirty lakh reader-minutes in which no information arrived. And next time it is 40, then 55, because empty documents are the cheapest thing to make.

I made this argument at a webinar in 2026, at a time when the whole industry had "data-driven" in the air. I think there is a difference between data-driven and data-dressed, and the difference does not show.

An old precedent comes to mind here. The year 2026, the King Commission, the Hansie Cronje affair in South Africa. I was very young then and it took me years to understand that case. But what has stayed with me is which element carried the most weight — the evidence, meaning the record. And which was weakest — the interpretation. When there is no entry in the record, interpretation starts building itself, and there is always someone available to build it.

The record stays with whoever writes it down. Whoever leaves it blank hands it to someone else.

(Here, stepping aside — and this is the only part of this piece that sits beyond the ledger. Walking back through empty stadiums after the 2026 World Cup I learned something I later began calling the price of silence. Ranji's knockout coverage is so thin it cannot be called coverage. Between the WPL's broadcast promises and its delivery there is a gap nobody counts. Silence is a line item in the ledger, and I learned to price it. The empty document that night told me that the same silence now exists at the machine layer, and is more expensive there because no witness sits beside it.)

The risk register, which contains one risk that is not about sport

What is shameful is that the clearest part of that document was its risk register. Seven categories — sporting, personnel, commercial, rules, public opinion, systemic — and six of them empty. The seventh read: "Stage-1 returned an empty payload; downstream output at high risk of fabrication." Level: high. Impact: high.

I found that line correct. And I also found myself thinking that the real question is where a document that knows it should not be published can be stopped from being published. The document recorded its own objection inside itself, and then the same document was forwarded. It wrote fail-loud. It performed fail-open.

Admitting it in the language is not the same as carrying it in the process.

Contrarian angle: an empty document is more honest than a false one, and that makes it more dangerous

Here I have to break my own argument, because a tidy answer across eight sections could say: "an empty payload is an honest failure; an empty analysis is better than a wrong analysis; therefore the system behaved correctly."

That is entirely wrong, and it is attractive enough that its wrongness does not show.

A wrong number carries its error with it. It is a fingerprint. If I write someone's average as 47.3 when it is 43.7, somebody will check it, and if my editor is strict I will be caught. That is good. Fear works.

An empty document is an error with no way of being caught. It makes no comment, so it cannot be contradicted. It makes no claim, so it cannot be called false. It simply exists, and borrows credibility from its format that it never earned.

Here is the second turn, and this is where I put my own view.

When something is empty there are two roads. One — start filling immediately. Two — name it as empty. Across my whole career I have seen the first road taken far more often than the second, and never once has it produced anything better.

But there is a third road I adopted later, and it is the one I would recommend to the makers of that document. It goes like this: decide the threshold before the work starts — how many information points below which this document will not be published. My number is three. Three real points drawn from the source, plus a publication date and a format token. Below that, the pipeline halts, routes to a human, and says it cannot proceed.

This threshold is not a principle, it is an operating rule, and its entire value lies in being fixed before the work begins rather than after. Like my three-match rule. The biggest problem with that document was not that it was empty. The biggest problem was that whoever built it had not decided in advance when to stop.

And this brings back the outlook I call the price of silence, at a second layer. Cricket media trained itself over the past decade to never be caught empty-handed. Something to say about every series, every auction, every selection dispute. It does not treat "nothing to say" as an option. But "nothing to say" is a valid outcome, and the day the industry accepts it as one, half its waste will disappear on its own.

The next thing to watch

I am not going to delete that file. I will keep it as a maintenance case — a specimen against which every new pipeline should be tested on the day it finds nothing. Because no system shows what it is in its best state. It shows what it is in its empty state.

The real question is not why it happened that night. The real question is what the document will look like the next time the data does not arrive. If it once again arrives with eight headings and not one line saying "I checked nothing", then we have learned nothing. And if it arrives with one line that says plainly that I have nothing and I know it — then that bad day will be more useful than all the good days with eight sections.

An empty ledger sells for less than a ledger that looks full, but the empty ledger is the only one you can open next year without shame.

And the number line I have been looking for over eleven years was never one line. It was two. First: how much happened. Second: how much was known. That night's document had eight places for the first and not one for the second. The next time a cricket data desk tells a reader that "this number shows", the reader has a right to ask — and you have a habit to build — where the desk was standing at the moment the number was recorded.