The 36 on the Board, the Empty SCG, and the System That Heard Nothing

In December 2026, I was sitting in my parents' living room when India collapsed for 36 in Adelaide. What was on the screen was barely thirty minutes of cricket. There was no commentator on the mic in my headphones — I keep a sound notebook, ninety seconds of ambient audio after every match. That night, my phone held only a heavy silence. Three weeks later, at a 25 percent capacity SCG with roughly nine thousand people, I watched Hanuma Vihari bat 161 balls on a torn hamstring — 23 not out. Ravichandran Ashwin survived 128 balls for 39. Nobody cheered. Nobody left. That was the first time I understood that a match is held together by endurance, not noise.

Last year I put that silence into a full report — around twelve hundred words, containing Adelaide's 36, the empty SCG seats and Vihari's block. The report went into a news-aggregation pipeline. What came back was startling: title N/A. Source N/A. Information points: empty list. Entities: unavailable. Domain label: cricket_world. The system had dropped all of that silence, all those deliveries, all that torn hamstring. It heard nothing.

I have been in sports journalism for twenty years, and for the past few years I have been running a team of more than fifty journalists at RevSportz. We have a commentary box, field reporters, audio recorders. But as the demand for publishing speed has risen, a large share of our material passes through a pipeline — one that first reads the game, then turns it into data points, then feeds those points onward. When this pipeline fails, it does not fail loudly. It fails quietly, and the most dangerous part is that it presents its own silence as a 'finding.'

Earlier this year I saw another case. One of our reporters filed a piece on a Ranji match, describing a complete change in a spinner's bowling — in the first innings he was pitching the ball outside the line, in the second he was attacking the stumps. The file went into the pipeline. The text that came back was clean, the average was correct, the economy was correct — but the technical shift was gone. Why? Because the system had no tag for things like 'outside the line' and 'at the stumps.' What it did not understand, it dropped, and what it dropped, it never reported.

This is not a one-off. The wave of automation that has hit sports journalism carries a problem that is philosophical rather than technical: the system does not distinguish between 'unknown' and 'risk-free.' If a title is N/A, that does not mean 'there is no story.' It means 'we do not know.' But when a downstream reader sees 'N/A,' they often read it as 'no problem.' That is the biggest deception.

I have a clean example. Two years ago we tried an aggregator for coverage of an overseas tour. By the first Test, the system had correctly produced the playing XI for five of eight matches. For the sixth, what it gave was right — but only half. It had removed an injured fast bowler and put someone else in his place. The team looked balanced, but in reality that player was not playing. When our field reporter told us on the phone, we discovered the system had swapped the injured player because its database kept 'injury update' separate from 'squad.' There was no rule joining the two.

Now imagine if that same pipeline were applied to a live blog, a fantasy suggestion or a preview. A whole edifice of precise numbers would be built on the wrong team — and it would look so credible that nobody would bother to challenge it.

For me this is not just a technical issue. At RevSportz we initially thought automation would give us speed. It did. But we quickly realised that in sports journalism, what matters more than speed is 'the speed of truth' — the right information at the right time. And the first step toward right information is accepting that there are things the system cannot read.

Now let us get to the concrete work of what is broken in this pipeline.

The first layer is extraction — pulling information from raw text. There are three big misses here. First, paywalled or JavaScript-rendered pages. Often the system opens the page, but the content never loads. Second, the page exists but its schema has changed — if a site alters its layout, the old extractor catches nothing. Third, language. Reporting done in Hindi, Tamil, Telugu or Bengali in India stays out of reach of English-trained models. My sound-notebook habit matters here — what is recorded on the mic survives in text only if someone deliberately writes it down. No system will spontaneously catch 'silence' or 'the direction of the wind.'

The second layer is entity recall — catching names. In my experience this is where the most damage is done. A match report contains ten players, two teams, an umpire, a venue, a toss — all names. If the system drops even half of them, the entire downstream analysis is incomplete. And the biggest problem is that you will not even know what was dropped. The report will read fine, the numbers will look right, but one crucial name will be missing.

The third layer is format identification. In cricket this is the most critical. Test, ODI and T20 numbers are not comparable. A batting average of 40 in a Test and an average of 40 in a T20 mean different things. If the system did not identify the format, then it has not placed a single number in the right context. In my experience this is the most common error — because often the format is not even written clearly; it is inferred only from context.

The fourth layer is tagging. This is where the cricket_world problem surfaces. In a correct taxonomy this should be 'Cricket,' but the system put it in a different, non-standard label. When the label itself is wrong, all subsequent processing on that label goes in the wrong direction. It is like putting a book on the wrong shelf — the book is fine, but now nobody can find it.

The fifth and most dangerous layer is downstream reading. When all fields are 'N/A,' the system should say: 'Analysis blocked, content unavailable.' But often it says: 'No notable risks found.' That is the language trap. 'No risk found' and 'no risk exists' are two different things, but to a reader they look the same.

What worries me most in this whole matter is one thing: in sports journalism we often focus on 'what happened,' not on 'what was missed.' I saw Vihari's block at the SCG, I saw the collapse for 36 in Adelaide. These were moments where the scoreboard spoke little and the ground spoke more. If our pipeline cannot even read such moments, then what game are we reporting on?

Now the point that stings me the most.

Over the past few years, sports media has carried a very confident story — that artificial intelligence and automation will make sports reporting faster, cheaper and 'more objective.' You hear it at every conference. But the real question is not how fast the system is. The real question is what it says when it cannot read something.

In my experience, most systems go quiet at that moment. And in the world of data, silence gets read as 'zero.' This is the same error as a miscalculated DLS target in cricket — the target looks fine, but the match has been taken away.

There is another side to this that the media talks about less. Automation hurts most those journalists who report from places where data is not clean — Ranji, local leagues, the lower tiers of women's cricket. In these places, getting accurate information requires making phone calls, going to the stadium, talking to a coach. The system prefers the report it can access easily. The result: stories that are hard to get slowly vanish. And that is not just a technical question — it is a question about the entire sporting ecosystem.

One more thing. I was born in India and work in Australia. I have seen that when a system reads Indian cricket, it often catches only the big names, the big scores and the big tournaments. The patience of Test cricket, the block, the story of the third-choice keeper — none of that enters the system's view. And that is why when a 'data-driven' report arrives, it often contains the same story we already knew. This is automation's greatest illusion: it does not give new information; it dresses old information in new clothes.

So what is the solution? In my view, three things matter most.

First, every pipeline needs a 'null-gate.' If information points are empty, if the title is N/A, if the format is undetermined — the system must halt and route back for re-extraction. It should never move forward empty-handed.

Second, 'N/A' should always be written as 'unknown,' never as 'clear.' It is a small thing, but it is tied to the reader's trust.

Third, every report should contain at least one thing only a human can provide — a sound, a smell, a silence. That is why I keep a sound notebook. After every match I have ninety seconds of audio — crowd noise, wind, sometimes a child's shout. That is the data that will never enter any pipeline. And that is the data that binds the reader.

The lesson is clear. The future of sports journalism will not be decided by 'who writes fastest.' It will be decided by 'who can prove they did not lose anything.' The media organisations that can prove that will survive. The rest will be fast, cheap and empty.

That night in Adelaide, when India collapsed for 36, the silence in my room was not recorded on any scoreboard. But it was in my phone. The next time a system claims to know everything, I will ask it just one thing: did you hear that silence? If you did not, then you did not read the match — you only read a voluntary description of it.