Empty Cells, Heavy Duty: Verification Over Estimation in Cricket Data

Last month I opened a data pipeline. Match logs from three sources were supposed to merge in one place, but the screen showed only zeros — no player names, no over counts, not a single row of run-rate. The spreadsheet did not lie; it waited for me to catch up. And in that exact moment the easy path beckoned: fill the empty cells with a plausible estimate, drop in a smooth average, nobody will notice. Before I trusted the model, I charted forty-six matches by hand; so I know the catching is my job. Watching cricket on English grounds from Liverpool, a habit formed. Where each ball landed, who held their line, which over the field changed — I write it down. This is not mere memory. A ground story is usually built on one word — momentum, rhythm, confidence. But rhythm is not a number; rhythm is an explanation. And an explanation unverified sounds like truth, though it may be wrong. My method is plain but merciless. Before a match I write down what I want to see and which piece of evidence would prove me wrong. Pre-registering a hypothesis is rare in cricket analysis, because it leaves little room for later spin. For every series I keep a note — source, sample, cut-off date. When someone asks, “Do you actually watch cricket?” I open the workbook instead of the feelings. I trust numbers I have counted myself. Charting forty-six matches, I logged one thousand two hundred fourteen shots — distance, angle, which bat. Why the labour? Because when the numbers settled in one place, a popular story collapsed. What was explained all season as momentum was really a quiet shift in shot quality. That taught me to stop writing deserved and start writing how many. If the data says one thousand two hundred fourteen shots, I do not stop without checking the next one. Here is the real point. In cricket we often take a number as a verdict without asking how it was made. An average run-rate, an economy, a strike-rate — where did the measurement come from? How many overs of sample? Who logged it? Under what conditions? These questions hidden inside a match-up usually vanish. An analysis that does not audit its inputs is a slave to its inputs. I attach a short method note at the end of every piece — source, sample, cut-off date. At first I thought it was excessive caution. Later I saw readers quoting the note itself to argue with each other. Then I understood: an analyst's job is not to impose a view but to build a clean basis for debate. My hand-charted record has shown one thing again and again: over counts and time counts never lie. A spell that looks short can build pressure over a long stretch; an innings that looks fast can hoard runs in slow overs. Four hundred fifty minutes against three hundred sixty — that gap told the real story many times. Counting the knockout minutes of every team in one tournament, I saw the finalist arrive tired while the champion arrived rested. People were saying confidence; I was saying calendar. The accounting of bowling changes is a simple example. A side changing bowlers every over looks aggressive. But written by hand, the change turns out to be not pressure but a sign of being lost — because runs come most in each new bowler's first over. What looks like strategy is really instability. That difference is caught only by counting by hand, not on the scorecard. Beyond time and overs there is one thing nobody counts: unequal preparation. I grew up in Dhaka, work in Liverpool — the two places show that analytical access is unevenly distributed. Big cricket economies hold camera tracking, physio data, large analyst teams. Small economies often lean on two or three senior players' shoulders. Veterans like Tamim Iqbal, Shakib Al Hasan, Mushfiqur Rahim carry the side across formats year after year, because there is little time or room to build replacements. This is not a shortage of talent; it is a matter of infrastructure. Whoever has the workload tool can save a player; whoever lacks it runs on guesswork. I learned to measure this gap by hand, because big models are not built for small leagues. Judging a small side by a big side's rules is asking the wrong question. The data from eighty-one empty stadiums taught me that part of home advantage is really the accounting of noise — and noise, too, is unevenly distributed. A side used to full grounds and a side that sometimes plays to empty stands cannot be judged by the same yardstick. Now the thing data-lovers rarely want to say. My spreadsheet has testified against me. I coded crowd presence for eighty-one matches in one series and saw the home win rate dip somewhat. At first I thought I had found something big. Then I understood: the sample was small, the effect modest — and two things happening together do not make one the cause. This is my biggest caution: correlation is not causation. Seeing a relationship between over-rate and winning does not mean faster over-rates win matches. Seeing a relationship between field changes and wickets does not mean the changes caused them. Often both are the product of a third thing — pitch, light, fatigue, or calendar. An analyst who ignores this caution becomes an advertisement for his own model, not an analyst. I am always cautious with small samples. Building a trend from four or five matches in a series is easy, and wrong. So beside every trend I write how many matches, how many balls, and what was left out. What is left out often hides the real story. There is another trap I carefully avoid: filling gaps with estimates. An empty cell does not mean I may put whatever I like in it. An empty cell means absent evidence, and absent evidence can never be the basis of a claim. A transfer is not a rumour; it is a row of cells awaiting confirmation. Until that cell fills, I will not write about it. As a blockchain ledger accepts no transaction without verification, so does cricket's ledger. So what happened to that empty spreadsheet? I did not fill it. I went back to the sources, asked for raw logs, reconciled every match by hand. It took days. Then when the numbers came, they proved my hypothesis wrong — and that was the day's best result. Because if an analysis does not challenge my earlier view, it is not analysis; it is only decoration for my opinion. In the coming weeks, when a team wins a run, someone will say the rhythm is back. I will be looking at the over count, the rest days, the travel distance, and the age of the pitch. The question will remain one — are we measuring the number, or merely liking it?