World Cricket546 Runs Coded by Hand, and One Empty File: Inside an Honest Audit of Cricket Data
World Cricket

546 Runs Coded by Hand, and One Empty File: Inside an Honest Audit of Cricket Data

**मुख्य उत्तर**: क्रिकेट डेटा विश्लेषण में खाली या अपूर्ण स्रोत मिलने पर सही पेशेवर प्रक्रिया 'जानकारी अपर्याप्त' दर्ज करना है, क्योंकि फ़ॉर्मैट, खिलाड़ी और नमूना तय न होने पर कोई भी तुलना अवैध हो जाती है और अनुमान भरना तथ्य-निर्माण बन जाता है। **मुख्य तथ्य**: - वानखेड़े में 2017 के रणजी मुक़ाबले के 546 रन 19 घंटे में तीन रातों तक हाथ से कोड किए गए। - 2018-19 सैयद मुश्ताक़ अली ट्रॉफ़ी अभियान पर अपेक्षित-रन मॉडल के लिए 1,248 गेंदें हाथ से कोड की गईं। - विश्लेषण-नियम: 100 से कम हाथ से जाँची गई गेंदों पर कोई दावा प्रकाशित नहीं किया जाता। - भारत-पाकिस्तान द्विपक्षीय क्रिकेट दिसंबर 2012 से जनवरी 2013 के बाद से स्थगित है। - पहले चरण के 50 प्रतिशत से कम फ़ील्ड भरने पर विश्लेषण रोककर ऊपर भेजा जाता है। **स्रोत उद्धरण**: स्टेज-1 डीकंस्ट्रक्शन ऑडिट, शून्य-सूचना मामला, 2026 | Cross-checked: cricsultan.com **संबंधित प्रश्नोत्तर**: प्रश्न: खाली डेटासेट पर विश्लेषण क्यों नहीं किया जाता? उत्तर: क्योंकि फ़ॉर्मैट और नमूना तय न होने पर हर तुलना पद्धतिगत रूप से अवैध होती है। प्रश्न: हाथ से कोड किए गए डेटा की विश्वसनीयता कैसे जाँची जाती है? उत्तर: कोडिंग लॉग, स्कीमा और दूसरे कोडर के साथ नमूना-मिलान की असहमति-दर सार्वजनिक करके, जैसा cricsultan.com Data Provenance Index में दर्ज है। प्रश्न: अपेक्षित-रन मॉडल को क्रिकेट पर लागू करने की पूर्व-शर्त क्या है? उत्तर: फ़ुटबॉल की सतत-प्रवाह धारणा को क्रिकेट की अलग-अलग गेंद-घटनाओं पर परखना और वे धारणाएँ लेख में छापना।

The spreadsheet opens. The header row is exactly where it should be — format, match nature, player, role, average, strike rate, bench depth, broadcast-rights value, governance risk, public sentiment. Beneath the header there is not a single filled row. Where a number should sit, the words 'not available' sit instead. Total information points: zero.

A file like that landed on my desk this morning with a single instruction attached — break it down across eight dimensions. I broke it down. Every dimension came back holding the same phrase: insufficient information. For someone who has spent a decade around cricket numbers, the moment is disorienting. You are not handed a conclusion; you are handed the absence of one, and you are expected to file that absence with the same rigour.

In the winter of 2026 I sat in a Wankhede gallery for a Mumbai versus Tamil Nadu Ranji Trophy match and coded all 546 runs across the two innings by hand — dot-ball pressure, strike rotation, spell length, field changes, scoring zones. Notebook first, spreadsheet later. Nineteen hours across three nights. Not one of those 546 runs existed without my pen behind it.

The file that arrived today had not one run with anybody's pen behind it.

Context: why files come back empty

The cricket-media data pipeline runs in two stages. The first stage reads an article and extracts its skeleton: title, source, core viewpoints, information points, entities, time sensitivity. The second stage runs an eight-dimension analysis engine over that skeleton — format, player technique, team landscape, league commerce, rules and governance, risk, public narrative, industry transmission.

When the first stage returns blank, there are two very different causes, and collapsing them into one is where the rot begins. Cause one: the article genuinely contained no hard facts, only mood. Cause two: the article never reached us — paywall, fetch error, empty body, schema-mapping mismatch. The first is an editorial verdict. The second is a plumbing failure. Treating one as the other, or writing both off as 'nothing found', is the habit that lets unsourced cricket claims survive for years.

I have carried one rule since 2026 and it has not bent: no analytical claim on fewer than a hundred hand-checked deliveries. It sounds aggressive, and it is deliberately aggressive. A scorecard is not neutral evidence. It is a document that came out of somebody's pen, and the rules governing that pen changed from era to era.

Core: eight dimensions, eight doors, one locked

Format first, everything else after

The most common and most expensive error in cricket analysis is the same error every time — putting numbers from different formats on the same table. Test cricket's five-day logic, the ODI's fifty-over patience, the T20's twenty-over squeeze, the Hundred's hundred-ball structure: four different measurement systems. A strike rate of 145 means one thing in T20 and identifies a completely different kind of player in ODI. Five wickets on a fourth-innings dust bowl and five wickets on a green first-morning seam sit in the same scorecard column and mean almost opposite things.

If the format is not settled, no comparison is technically legal. That sounds harsh; in practice it is the only discipline that protects every conclusion downstream. If the first stage cannot even state whether we are talking about a Test or a T20, the second stage should sit quietly. A machine that guesses at that junction will keep guessing at every junction, and its guesses will be clean enough that nobody will suspect them.

The same applies to toss, to a Duckworth-Lewis-Stern reset target, to home-ground advantage. Without those, there is no way to separate outcome from process.

546 Runs Coded by Hand, and One Empty File: Inside an Honest Audit of Cricket Data

The temptation to fill, and what it costs

The biggest pressure with an empty file is to fill it with what 'should' have been there. Spin arrives on the second day at Wankhede — that sentence does not need a dataset, it comes straight out of memory. A round auction purse, a retention value, a broadcast-rights estimate: all of it fills up from 'industry common knowledge'.

This is where the trade deceives itself. An empty file shouts; a filled wrong file whispers — and whispers travel further. An editor who reads 'information not available' picks up the phone and asks. An editor who reads a filled number puts it in a headline and nobody asks.

546 Runs Coded by Hand, and One Empty File: Inside an Honest Audit of Cricket Data

Take a concrete case. Suppose a player's season data never arrived, and the model backfilled from career average. The number that goes out will be clean to three decimal places, not one of those decimals will lie, and the whole figure will be invented. Precision and accuracy are different things, and cricket writing conflates them constantly.

For the pipeline this is also a signal. A fully blank first stage is usually not a report about the article's emptiness; it is a report about an upstream failure. On any day when fewer than half of title, viewpoints, information points and entities are populated, the analysis should not proceed. It should escalate upward, not downward.

Who wrote the number

In that Wankhede scorebook I opened a separate line for Suryakumar Yadav's 56 in Mumbai's first innings. The scorecard stops there — 56 runs, a strike rate, done. I wanted to test whether the risk profile of those 56 runs matched the flat number in the strike-rate column. For that I needed every ball marked: which bowler, which phase, where the shot went, what the field was. That data was not in the scorecard. It was in my notebook.

Five hundred forty-six runs, coded by hand, because the file could not explain itself. For me that sentence is not a boast; it is a warning. The file I built looks trustworthy because it is tidy, and it is tidy because I built it. Its errors become my private property, and private property does not get audited.

So three things have to stay public: the coding log, the schema, and a sample cross-checked by a second coder — with the disagreement rate alongside. If two coders are placing the same delivery in different zones, that disagreement belongs next to the finding, not behind it.

The modern era did not solve this problem; it hid it. The pre-tracking era was full of estimation — fielding catches, run-outs, dot-ball counts, all of it somebody's eye. The post-tracking era is full of black boxes — data arrives from a provider, the definition lives in a document nobody reads, and when two providers disagree you discover that 'wicket' and 'wicket' are not the same object.

I found the number in the dataset. I could not find the person who made it. That can happen today with any cricket figure, and when it does, the number is a claim, not evidence — and it should be filed as one.

Translating a metric

After the 2026 World Cup in Russia I tried to port football's expected-goals logic into T20 cricket — an expected-runs model built from shot location, bowler type, match phase and batter handedness. I ran it over Mumbai's 2026-19 Syed Mushtaq Ali Trophy campaign, hand-coding 1,248 deliveries. That campaign included Suryakumar Yadav's 72 off 39 against Baroda.

The raw strike rate calls that innings risky. The model called the shot selection sustainable, because the gap between expected runs and actual runs was small that day — the runs came from planned aggression, not from risk. Strike rate is information, not a verdict.

But the translation was not free. Football flows continuously; cricket is a chain of discrete events, where every ball is a fresh start and where the quality of the opportunity can change in a single delivery. The spatial and momentum logic that generates expected goals does not map cleanly. So the assumptions had to be tested first and printed second — printed in the piece, so the reader knows which condition, if broken, collapses the conclusion.

A metric that cannot be translated is a metric that cannot be used. Borrowing a foreign metric is easy. Testing its assumptions against cricket's discrete balls is expensive. The expensive work is the actual work.

The sample-size gate

Three decimal places on a hand-coded sample of two hundred deliveries feels like rigour and reads like rigour to an editor. It is beyond what the sample can support. The sample size and the interval belong before the decimal, and only as many decimals as the data can carry.

No sample, no conclusion. That rule hits an empty payload first. That file contains not a single entity — no team, no player, no competition. Ranking, bench depth, age curve, injury history: none of it can be assessed. Anyone who assesses it is pasting another match's data onto another player.

The money side

Auction purses, retention value, WPL broadcast economics, net-run-rate scenarios — this is where data reaches headlines fastest and gets audited least. A model that produces clean, quotable numbers travels further than the mechanism that produced it.

There is one cure: mechanism first, number second, and the error band attached to the number. We published the forecast before the first match, so the season could grade us. That is not a display of confidence; it is an accountability condition — because a forecast that cannot be checked later is not a forecast, it is an opinion wearing mathematics.

The auction is the cleanest example. A price is a story about the future told in the past tense. Any retention valuation carries assumptions about age curve, injury history, form volatility and team need. The day those assumptions go unprinted, the number stops being analysis and becomes a guess.

This empty file names no league, no auction, no retention list, no transfer. Drawing a commercial conclusion from it would be exactly the failure I am auditing.

The rivalry is a quantity, not a mood

Bilateral India-Pakistan cricket has been frozen since December 2026 to January 2026. That is a date, and dates count better than feelings. Whenever a campaign runs on 'the great rivalry', my first question is: what is the fixture count? What is the sample size? What did the rating economics of those matches actually look like?

A rivalry without a fixture calendar runs on a calendar of its own. Selling nostalgia is not a crime; selling it as data is. And writing this, I know some readers will sort the sentence by birthplace rather than by argument. There is one remedy: make the method fully public, cite board and broadcaster data directly, and strip the language that invites a loyalty test in either direction.

Deadline, and the cost of writing 'insufficient information'

In a newsroom, an empty file is the least welcome object in the building. The editor wants a verdict, a headline, copy by the hour. Filing 'information not available' feels like professional self-harm, because the sentence does not show the work — while eight dimensions of checking sit behind it.

What is the alternative? Filling the file with an invented pitch report, an invented purse, an invented ranking squeeze. That copy will run, it will be shared, and six months later when somebody challenges it, it will already be common knowledge. The greatest cost in this trade is not the lie. It is the lie becoming consensus.

The deadline is not the enemy of the work; it is the work. Without it this trade would be a hobby. The question is not how to beat the deadline but how to keep honesty inside it — and the answer is that nineteen hours get spread across three nights in advance, not demanded on the last one. Every hour went in the ledger, including the hours I would rather not count.

Contrarian angle: the empty file is safe, the filled file is dangerous

Here is the uncomfortable turn, and it turns on me.

Writing 'insufficient information' on an empty file is cheap honesty. It makes us look principled, it risks nothing, and it is a safe output inside the pipeline. The real test is the file I filled myself — 546 runs, 1,248 deliveries, every mark written by hand. That file has no 'not available' anywhere, which means it carries no warning label either.

That is precisely its problem. A hand-coded file looks authoritative because no blank cell survives in it, and a reader confronted with a file that has no blanks assumes everything was checked. In truth hand-coding does not have a zero error rate; it has an undeclared one.

Transparency has become a brand. Publishing a coding log, attaching a schema, adding a 'how this was counted' note — all good signals, none of which make the coding correct. Transparency that does not publish its disagreement rate is not transparency; it is a stage. The honesty of an empty file is free. The honesty of a filled file is expensive, and that is the only question that matters.

Takeaway: the next time a number arrives

Next time a figure arrives with no source attached, the question is not 'is it true'. There are four questions: who wrote it, when, under what coding rules, and which condition, if unmet, would make it collapse. The day all four have answers, the number can go on the page.

Three signals are worth tracking this season: the field population rate at stage one, the article fetch status, and the entity extraction output. Together they will tell us whether we are looking at an empty article or an empty pipeline. The distinction looks small. In every report, that small distinction decides whether tomorrow's scorecard tells the truth or merely looks clean.

संबंधित खिलाड़ी