Complete Schema, Empty Substance: When Absence Itself Is the Finding in Hockey's Archive
**মূল উত্তর:** হকির এক্সট্র্যাকশন রেকর্ডে শুধু ডোমেইন লেবেল "হকি" থাকলে সেটি সাধারণ ব্যর্থতা নয়, তথ্য-শূন্য একটি ফলাফল। স্কিমায় "ব্যর্থ" ও "শূন্য" আলাদা না থাকায় এমন রেকর্ড বৈধ দেখায় এবং পরের ধাপে ভুল বিশ্লেষণের ঝুঁকি তৈরি করে। **মূল তথ্য:** - ১৪ মার্চ ২০২৬ পর্যন্ত সংরক্ষিত ওই রেকর্ডে শিরোনাম, সূত্র, ধরন ও তথ্যবিন্দু — সবই ছিল N/A। - বাংলাদেশ প্রিমিয়ার ডিভিশন হকি ২৭ বছরে মাত্র ১৩টি সংস্করণ সম্পন্ন করেছে। - ২০২০ সালে ১,১৮০ ঘণ্টা আর্কাইভ ফুটেজ কোডিং করে Leagueের অনুপস্থিতিকাল ডেটা আকারে লিপিবদ্ধ করা হয়। - ১৯৯৯ সালে ঢাকায় এক ম্যাচে মোহামেডানের ৪১টি আক্রমণ-প্রবেশের ১৯টি এসেছিল বাঁ-ভেতরের চ্যানেল দিয়ে। **সূত্র উল্লেখ:** সাব্বির রহমানের হ্যান্ড-কোডেড আর্কাইভ ও ম্যাচ নোট, প্রকাশ ১৪ মার্চ ২০২৬। **সম্পর্কিত প্রশ্নোত্তর:** - প্রশ্ন: কেবল ডোমেইন লেবেল থাকা রেকর্ড কেন বিপজ্জনক? উত্তর: কারণ এটি বৈধ দেখায়, ফলে পরের ধাপে কেউ ফাঁকা ঘরগুলো নিজের অনুমানে ভরিয়ে ফেলতে পারে। - প্রশ্ন: বাংলাদেশ হকিতে এই সমস্যা বেশি কেন? উত্তর: কারণ League ও ম্যাচ-রেকর্ডের ধারাবাহিকতা কম; ২৭ বছরে ১৩টি সংস্করণ মানে বিপুল সংখ্যক অনুল্লেখিত সময়কাল। - প্রশ্ন: সমাধান কী? উত্তর: স্কিমায় "না হওয়ার কারণ কোড" যোগ করা, যাতে অনুপস্থিতি শূন্য হিসেবে নয়, কারণ হিসেবে সংরক্ষিত হয়।
Fourteen files went into an extraction pipeline from my Kuala Lumpur desk on a Monday morning. Thirteen came back in normal rhythm — headline present, information points listed, at least an inference available about who wrote the thing and why. The fourteenth returned as a tidy table: row after row of cells, each with a label beside it, and nothing inside. Title: N/A. Source: N/A. Type: unclassified. Core viewpoint, the one-sentence summary: blank. Author stance: absent. Article purpose: absent. Information points: none listed. Time sensitivity left as "not assessed." Source quality instructed to be judged "from the source fields" — fields that were themselves N/A.
The only substantive datum in the entire record was the domain label: hockey.
The table looked innocent. That is the danger. A broken pipeline normally shouts; it throws error codes, it goes red, somebody phones you. This record did not break. It filled every slot in the schema and put nothing inside them. It did not leave a single cell blank — it placed the words "nothing here" inside the cell instead. By the schema's own logic the job was complete. By the substance's logic the job may not have been done at all.
The strangest line read: "identify entities from the information points above." There were no information points above. The instruction was pointing at a void, and the machine quietly obeyed the finger.
One thing the record did leave behind, and it is not information — it is a warning. When a template can produce a valid output from no content, its biggest risk is not data loss. The risk is that someone downstream fills the empty cells with their own inference and never notices they are doing it.
Working on hockey, the scene did not feel new. It felt familiar. The model is not the match, but it can tell you where to look. This time the model pointed at one place only: the label cell. Everything else sent me back to my own archive.
What a template can and cannot carry
An extraction pipeline is not a journalism machine; it is a mould. The mould's job is not to make the finished piece but to arrange the raw material for whoever sits down to make it. Title, source, type, core viewpoint, information points, entities — fill those six and the next person knows where to look. The mould is fine. File fourteen showed that a good mould is not sufficient.
In database terms, two distinct events can occur here, and their treatments are entirely different. One: extraction failed — the text existed, the machine could not read it. Two: extraction succeeded and found nothing. The first is the machine's fault; rerun it and move on. The second belongs to nobody; it is a true result, and the next stage has to make a decision on it. The problem is that most schemas have no separate cell for the two. Failure and emptiness end up looking identical, because both get written as N/A.
I learned that distinction on a wooden bench in Dhaka in 2026, hand-coding the Premier Division for a client feed over six weeks. I was charting Mohammedan's imported Pakistani pair, Shahbaz Ahmed and Tahir Zaman. In one match I logged 41 Mohammedan entries into the attacking third, and 19 of them arrived through the left inner channel. That does not mean they played down the left that afternoon. It means the left inner channel was what I could see most clearly that afternoon. A scoresheet where nobody wrote the goalscorer, filled in with a zero, produces exactly that error. It looks like a small detail. The credibility of an entire archive rests on it.
The Premier Division ledger is the largest example of that distinction I have. Thirteen editions in twenty-seven years. I put the number into a table once — a spreadsheet four columns wide — and it became the most quoted thing I have written. The dark stretch from 2026 to 2026, the Abahani–Mariners joint-title dispute, a national system leaning on a single artificial surface at Maulana Bhasani: those are events. The rows beside them, the years with no edition at all, are not events. They are the schema's N/A. And this is precisely where hockey writing in the country goes wrong most often: it treats N/A as zero and builds a story on top.
In 2026 the stadiums emptied, the league went dark for a second straight year, and I retreated into film — March to December, 1,180 hours. 2026 Asia Cup tapes, 1990s Premier League footage, the AHF Cup finals of 2026, 2026 and 2026. One thing became clear in that room: 1,180 hours of archived film taught me that empty stadiums have geometry. Absence has to be written as data, or it fills itself with sentiment.
Where "missing" and "zero" collapse into one
If file fourteen were only a technical accident, there would be nothing to write about. The problem is that it is the first stage of a process, and the later stages of that process run exactly this way in our part of the world.
Picture a desk. Evening deadline. An editor receives a record with the domain marked "hockey" and everything else empty. He forwards it to someone down the line and asks for four hundred words. That someone now faces the empty cells. What does he do? He reaches into memory. And in memory, hockey means 2026, means Shahbaz and Tahir, means "we were once the best in Asia." Four hundred words appear. The record has moved from N/A to narrative, and no alarm sounded anywhere.
The habit has a name, and the name argues against my own profession: the compulsion to fill a void. I watched it from the other side in the press box at the 2026 Asia Cup in Dhaka. India 2–1 Malaysia. Colleagues beside me filed six-hundred-word match reports; I posted eleven hand-drawn frames showing how India's right-half kept dragging Malaysia's zonal press five metres out of shape before switching the point of attack. Two of those frames were reprinted by a Dhaka daily within a day. The mechanism was not complicated. Everything in those frames was visible in the footage. Everything that was not in the footage I did not draw.
Now back to the label. The record does not say whether the hockey is ice or field. It does not name a league, a team, a player, an event. It does not say whether the text was news, analysis, opinion or a stats sheet. What, then, is the domain label? Not a conclusion — an inference. A keyword-level guess saying that the word hockey appeared at least once.
One piece of information is buried inside that guess, and it is not cheap: the input was not zero. Something entered the pipeline, or the classifier could not have attached a label at all. So four possibilities remain, ranked by the order of blame.
First, the input was near-empty — a five-line note, or a feed that had been truncated. Second, classification ran and content analysis was skipped or crashed. Third, truncation happened upstream: the article existed but was clipped before the extractor saw it, most likely at a paywall.
Third is ordinary. The fourth is the dangerous one, and it is why this record interests me: the extraction failed, but the failure arrived disguised as completeness. The schema filled with nulls rather than errors, so a failed process came to look like a valid — if empty — record. There is hardly a worse integrity failure in a data system, because it operates silently. No stage triggers, no alert fires, the empty record simply floats downstream.
The imagination downstream
This is where hockey re-enters, and it enters at a very practical point: large parts of our archive are made of precisely these valid-but-empty records.
The 2026 logging gave me a method for it. Coding 1,180 hours, I had to fill four cells per clip: was the event visible, if so who produced it, if not what was the reason, and does that absence affect my conclusion. The fourth cell mattered most. On many clips the note read: scoreboard visible, scorer not visible, and this limitation does not affect any decision I am currently making. That single cell let me come out of a thousand hours of footage without superstition.
On another clip the note had to run the other way: the penalty-corner replay exists, the entry angle is measurable, but the deflection zone sits outside the camera frame. That absence does affect my decision, because I was about to make a claim about corner geometry. So the claim dropped out of that clip and survived only as a hypothesis in the notebook.
Small sample, large claim
Here I have to name a fault of my own. Hockey in Bangladesh hands me small samples: a handful of corners in a season, a few matches, one artificial surface. And my mind enjoys turning that little data into a beautiful law.
With corners the trap is straightforward. Entry angle, deflection zone, rebound geometry — all measurable, and all lovely once measured. But the single artificial surface at Maulana Bhasani is a variable in the equation, not a lament. When set-piece output is low, the first explanation is coaching and surface; that is an argument, not a wail.

So I have capped myself. Any claim that runs beyond the sample I actually coded gets written as a hypothesis, not a finding. Forty-one entries and nineteen through the left inner channel is one match's coding, not a rule. The AHF Cup hat-trick — 2026, 2026, 2026 — and the Junior AHF Cup titles show a trend, but a trend is not proof.
And the real question starts after those three titles. Where do those players go after twenty-one? The HCT 2026 franchise experiment was an attempt to answer exactly that. But why was the experiment never repeated? And in which file is that "why" written down?
Nowhere. That is the largest N/A in our archive.

The geometry of absence
One thing I keep returning to. Thirteen editions in twenty-seven years reads like a complaint to anyone who only scans the number. The number is a coordinate. It says our system's default state is "not running." Running the league here is the exception, not the rule. A system whose default is the exception will produce an archive of exceptions too: records occasionally, silence mostly.
Which makes file fourteen unremarkable. It is entirely consistent with our conditions. Empty stadiums have geometry, as they did in 2026. Empty records have a shape as well. The work is the same in both cases: measure the shape, then decide what it is telling you.
The repair that is actually damage
The instinct that arrives first is the instinct to fix it. Find the input, rerun the pipeline, populate the cells. Technically that is the natural response, and usually it is the right one.
With this record there is a counter-argument worth taking seriously. The record's only value is that it stayed empty. An empty cell tells the truth for as long as it stays empty: here, we do not know. The machine stopped halfway and did not cover its own stopping. If the input cannot be recovered, the empty cell is our only witness.
We cannot tolerate empty cells. On a desk, an empty cell is discomfort, and the standard cure for discomfort is an inference. Once inserted, it stops being an inference. It becomes a record, and next time someone reads that record and decides. That is how a blank cell becomes a wrong policy three stages later.
Hockey administration has the same disease. A season does not happen. How is it recorded? Not with a reason, a date, a level of decision. It is recorded with a press release promising the next season. The most valuable fact — why it did not happen — is lost permanently, the same problem returns the following year, and nobody can see that it is the same problem.
HCT 2026 works the same way. The franchise model was a packaging test. Whether it failed or succeeded remains an open question, because the answer sits in no record. Which meeting, on what date, on what reasoning, decided not to run it again? One line would have taught us a method. The line does not exist, so neither does the learning.
So the definition of repair has to change. What is needed is not a rerun but a new field: a reason code for non-occurrence. And that field cannot be left blank — left blank, it becomes another N/A and we start from zero again.
The geometry notebook was born in a Kuala Lumpur data feed, drawn first on a Dhaka bench in 2026, and it is now on its third version. The first was built on one turf at Maulana Bhasani. The second came after 1,180 hours of footage. The third came after these empty records, and it has overwritten the first two in many places. That is painful, because a person loves the thing he built as a thing. But the revision is what gets published, not the relic. A notebook that begins defending its own existence has stopped being an instrument and become an obstacle.
The next checkpoint
The next time a match report or a Premier Division scoresheet lands in front of you, do not look at the scoreline first. Look for the goalscorer's name. Look for the corner count. Where a cell is empty, check whether it was filled with a zero or left empty. The filled one is the suspicious one.

And in the next AHF cycle, or the next domestic season, one question needs an answer: can our record distinguish "did not happen" from "happened, but there is no data"? The day it can, our numbers stop being superstition. Until it can, every empty cell is an invitation — for someone to fill it with a story.
