The Religion of Empty Cells: When Cricket's Data Pipeline Fails Silently

2:47 a.m.

An Excel sheet open on a laptop screen on a balcony in Barishal. The cursor blinking in cell A1, rows of empty cells below it. I had sat down to test a hunch: a pre-draw collapse model for an upcoming major tournament. The file was named collapse_v1.xlsx. It was supposed to hold 460 ball-by-ball data points, pressing-intensity figures for 14 teams, pitch reports for six venues, and a six-month squad-availability timeline for eight sides. The pipeline returned exactly one thing: zero. Not a single number.

I opened Excel to check a hunch, and a religion died.

That religion is analytics. In modern cricket it is no longer merely a tool; it is a faith. Before every decision, the question 'what does the data say?' is chanted like a prayer. But that night I realised something: the biggest lie of cricket's data age is not a wrong model. The biggest lie is assuming the data exists.

This piece is an accounting of those empty cells. Because an empty cell is not an accident. It is a symptom, of a system in which we have forgotten the difference between evidence and decoration.

Context: The temple cricket built in twenty years

In 2026 I started a page called BDCricTeam. Back then, analysis meant one sentence and one scorecard. Today it means a dashboard, a data vendor, a model, a report, and thirty minutes of graphics theatrics on television.

What changed? Data has moved to the centre of cricket's internal decision-making. In franchise leagues, auction prices are set from ball-by-ball maps. Bowling changes are made from a spreadsheet called 'match-ups.' 'Match-up' and 'split' are now ordinary dressing-room vocabulary. A national side's support staff today includes more analysts than physios.

This temple rests on a single assumption: that data is reliable, complete, and arrives on time.

But what I have seen in 21 years inside and around this industry is this: cricket's data supply chain is a long, fragile, multi-owner pipeline, and on any given day some cell somewhere is left blank. Some people see it. Most don't. Those who don't are usually the ones showing you graphics on television.

A word on my receipts file. In June 2026 I published a piece ten days before the draw, 'The Confederations Cup Was a Trap.' The argument was simple: Germany's 2026 Confederations Cup win had masked a collapse in pressing intensity. Opponents' passes per defensive action against them had climbed from 9.1 to 13.4. Germany exited the group stage with three points. Four thousand furious replies followed, and a Dhaka radio slot, which I kept for two years before getting bored.

Since then I have timestamped every prediction and kept a public receipts file: every call, dated, graded. It is not a diary. It is a public ledger, a chain in which every block carries a date, a claim, a grade. Anyone can reconcile the whole chain. That ledger is my only asset, and it terrifies me daily.

The draw was days away, but the spreadsheet already had Germany in flames. That fire was no magic. It was the product of a complete input and an honest model.

And this time? This time I had no complete input. I had zero.

Core analysis: What an empty cell says

Layer one: where the supply chain breaks

Cricket data comes from three layers. Upstream: domestic and youth cricket, trials. Midstream: national teams, franchise leagues, international series. Downstream: broadcast, commercial product, fantasy markets, derivatives.

The problem is that information flow between these layers is not one-way, and the timelines do not match. Upstream data takes months to arrive, sometimes never. Ball-by-ball data for domestic leagues is incomplete in many countries, absent in others. Yet national-team decisions assume the upstream picture is complete.

Here is the first fracture: the model we build has a more incomplete input than the model itself. The cleaner the model, the better the gap hides. Because a model is not afraid of an empty cell; it fills it with a default, interpolates, averages. And averaging often means inventing.

In May 2026 I watched all 81 Bundesliga matches behind closed doors. Home wins fell from 43 per cent to 33 per cent. I titled the piece 'Empty Stadiums Are a Tactical Experiment, Not a Tragedy.' The argument: crowd noise had suppressed away teams' pressing triggers for decades. Editors called it tasteless. Readers made it my most-read piece of the year.

During that project I started logging my own match database, 1,400 matches by December, plus three other databases I never finished. Those unfinished databases taught me the most: there is a difference between the courage to work with incomplete data and the fraud of passing incomplete data off as complete. Cricket's entire analytics industry oscillates between the two.

Layer two: the wall between formats

Test, ODI, T20. Blending their data is the oldest sin in cricket analysis. An opener's Test average and his T20 strike rate are two numbers for one person, but they resemble two different players. A format changes not just the number of overs but the risk calculus, the ball's condition, the field, even the batter's footwork.

A model that estimates T20 scoring from Test strike rates is not a model; it is decoration on a guess. That is exactly why, when I sit down to build a pre-draw model, I first decide to use only the last two years of T20 international data. Everything else is discarded.

But even that two-year window is not complete. Against many teams there is no ball-by-ball data. Many matches are rain-shortened. Many series' pitch reports were never stored.

Layer three: data theatre, or the magic of the spreadsheet

Now the real discomfort. Part of my identity is 'receipt-keeping accountant.' That identity has a trap: a complex Excel model looks like proof, even when its assumptions are cherry-picked.

I fell into that trap once. In March 2026, while grinding a data-analyst job in Barishal, I built a homebrew xG model from 380 Premier League matches and published 'Possession Is a Vanity Metric' on Medium. The argument: Chelsea's 93-point title was won on 54.1 per cent average possession, the lowest of any champion in five years. It drew 210,000 reads in nine days. Three outlets offered columns; I took the smallest fee with the largest editorial freedom.

Possession was the altar. The data was the hammer.

That piece taught me a rule I still keep: a model is never proof, a model is a claim. And a claim needs assumptions, sensitivity tests, and counterfactuals. Otherwise it is not analysis; it is spreadsheet theatre.

And this is where my 2:47 a.m. story returns. An empty spreadsheet is a gift. It forces me to see how much of my 'analysis' was standing on data, and how much was standing on my own confidence.

Layer four: player technique and the small-sample trap

The biggest small-sample trap in cricket is the batter's 'split.' 'His strike rate against left-arm spin is 145.' The viewer hears one number and assumes it is proven. In reality, if he has faced 60 balls from left-arm spinners, that 145 conceals pure coincidence. A top edge, a dropped catch, a boundary line could push the number from 145 to 110.

An analyst who knows the trap asks: how many balls? What period? Which venue? Against which attack? One who does not shows a big number on television and passes it off as truth.

In my database I follow one rule: I never use a split under 500 balls as the basis of a prediction, only as a lead for a hunch. That rule has saved me from many tempting hot takes and pushed me toward many boring truths.

For bowlers the trap is subtler. Valuing a spinner by economy rate means ignoring which phase he bowls in. Powerplay overs inflate economy, death overs inflate it, middle overs reduce it. One bowler, three jobs, three numbers. Judging him by one average is not judging him; it is averaging him.

Layer five: team depth and the ranking illusion

The most popular way to measure team strength is the ICC ranking. It is also the least reliable. A ranking is a rolling average that smooths home and away conditions, blends formats, and ignores squad availability entirely.

In a tournament, a team's real strength is set by four things: batting depth, bowling combination, bench depth, and age structure. Three of these can be measured; one cannot, and precisely because it cannot, it decides the most. Bench depth is like a reserve bank. A side with a strong reserve survives the hard stretch of a seven-match tournament. A side without one breaks at the first injury.

Yet almost every pre-draw analysis is about the first XI. Not the reserves. This is my biggest objection: pre-draw models account for the first XI, but tournaments are won by the strength of the fifteenth man.

Layer six: leagues, money, and the noise of agents

Now the part where analysis meets commerce. The franchise auction is cricket's biggest data drama. A player's price is set by a 'projected value,' really a weighted average of domestic performance, international flashes, age, fitness record, and market demand.

One variable usually drops out: the agent's noise. A player's representative does not merely negotiate; he sells a story. He manufactures a number, spreads it, and it slowly becomes true because everyone starts quoting it.

I once went backstage at the transfer window with a calculator and no manners. What I found: much of what the market markets as 'data-driven' is really a narrative sitting on top of data, and the narrative is usually not owned by the player.

This is cricket's most hidden cost. In the data age we assumed information would defeat narrative. In reality, narrative has entered wearing the disguise of information.

Layer seven: governance and the silent frontier

Another crack sits at the governance level. Who owns the data? Who may use it? Who can install tracking hardware and who cannot? Between international boards, franchises, broadcasters, and data vendors, a quiet war over ownership is running.

If a player from a small member nation asks who holds his own ball-by-ball data, the answer is hard to get. Yet that data will set his future auction price. This is a structure of inequality, and it is not any player's personal failure; it is the design of the system.

Counter-argument: where I could be wrong

Now I stand against myself. An analyst who cannot break his own claim is not an analyst; he is a propagandist.

First, I accept the strongest argument of match television: before data, cricket decisions were made from memory and bias. A coach would play his favourite for four matches because he liked him. Data has partly reined in that bias. That is a real achievement, and if I play the anti-data purist I deny it.

Second, my 'empty input' episode may be a technical glitch, not a structural failure. Maybe the vendor API was down that day, maybe my own script had a bug. If I pass off one accident as a trend, I fall into the old trap of pattern-recognition arrogance, where incomplete data alone lets me build a grand narrative.

Third, I was a data analyst myself, and I trust my own models. That is a conflict of interest. Part of my criticism may be resentment against a system that no longer keeps its door open for independent writers like me, having pulled analysts into the dressing room instead.

Accepting all three, I shrink my claim. I am not saying data is false. I am saying: data has changed cricket, but cricket has not yet proven that the data pipeline is complete. This claim is smaller, and it is the true one.

Signals to watch

I keep a pre-registered indicator list, to see whether my fear comes true.

First: if, before the upcoming tournament, no major side publishes a squad-availability timeline, that is a gap in the supply chain. Second: if an analyst shows a split number without stating the sample size, that is spreadsheet theatre. Third: if, before an auction, a player's price does not match his format-specific data over the last two years, that is narrative winning, not information.

And the fourth, most important signal: if during the tournament a side collapses to injury and its bench cannot fill the gap, that proves pre-draw analysis spent its whole budget on the first XI and ignored the fifteenth man.

Closing: a dated claim

I place a new block in my public ledger, with a date. The claim: most pre-draw analyses published before the upcoming major tournament will be built on first-XI data, but among the sides reaching the knockout stage at least two will be separated by their fifth-to-eighth bowling options and bench batting, the areas with the most incomplete data.

I may be wrong. But the error will stay in the ledger, dated, graded. Because to me a wrong cell is better than an empty cell: an empty cell does not lie, but a dressed-up cell does.

And what that empty night at 2:47 a.m. taught me is this: the next great battle in cricket analysis is not over a new metric. It is over the courage to admit which cells are empty.

The analyst with that courage survives. The rest build graphics.