In Cricket's Empty Data Rooms: An Analyst's Hardest Answer Is 'I Don't Know'

On August 12, 2026, after Burnley's 3-2 win at Stamford Bridge, I posted a thread: Chelsea 2.4 xG, Burnley 1.1; three goals from four shots on target is never sustainable. That thread pushed my newsletter 'Expected Noise' to fifteen thousand subscribers. But one thing nobody noticed: in that thread I deliberately left three boxes empty — Burnley's keeper's save value, the defensive line's height, and the referee's added-time decision. I had no reliable data on those, so I said nothing. Nine years later, I am certain those empty boxes were my best analysis.

I began writing about cricket twenty-one years ago at a sports desk in Dhaka. Back then, analysis meant match reports and repeating star names. Working from London today, I see the opposite picture: a metric for every ball, a shot map for every batter, a pressure index for every spell. Data is so abundant that the biggest risk is no longer scarcity — it is the temptation to slot a story into the place where data is missing.

Modern cricket analysis has three layers. The first is descriptive: who scored how many, who took how many wickets. The second is measurement-based: strike rate, economy, phase leverage. The third is predictive: which bowler will hold an edge over which batter next match. Most writing stalls at the second layer while pretending to reach the third. But entering the third layer demands a skill nobody teaches — the skill of saying 'I don't know.'

The first rule, which I keep like a monastery's rule: a metric only becomes meaningful when formats are separated. A Test average and a T20 strike rate cannot be spoken in the same language. In Tests, the accounting of balls rests on an innings' patience; in T20s, every dot ball is a cost. If someone says 'this batter averages 45, so he is the world's best,' I immediately ask — in which format, at which position, against which ball? Without answers to those three questions, that 45 is just a number, not analysis.

The structure of play and innings differs from football, so football's models cannot simply be imposed. In football time is linear; in cricket, progress is non-linear, measured in overs and wickets. A game can turn in one over, then fifty overs can pass with nothing happening. So in cricket my metric stands on three pillars: expected runs per ball, wicket probability, and phase leverage — how much that ball actually shifts the match outcome.

Phase leverage is my favourite, because it catches the small-sample trap. Fifty runs in the first six overs of an innings means twenty-seven runs are setting the match's tempo. But the same fifty in the last four overs means the game is nearly over. The number is identical; the context is not. Those who cannot reconcile this often reach the wrong conclusion.

The small-sample trap is the biggest enemy. A bowler averaging 12 across three matches — that number cannot carry a forecast. Dropped catches, wrong umpiring calls and toss luck push an average down across three games. So by rule I register a hypothesis first, then test it out of sample. Not the three-match average but the twenty-match trend — that is what speaks. At the 2026 World Cup I analysed PPDA in Russia versus Spain, where Spain's 8.2 and Russia's 31.6 said one thing — Russia would want penalties. They won 4-3. But that forecast was one match's fate, not a constant truth.

At the Qatar World Cup, Enzo Fernández's 2.3 progressive passes per 90 and 89% pass accuracy were the signal of a large sample, not a single-match flash. Pedri's 12.5 kilometres per game was likewise a consistent signal. The difference is here: a signal demands a sample, a flash does not.

Home-ground and toss luck. Venue bias in cricket is enormous. On a spin-friendly pitch a bowler's economy naturally looks better; away from home the number swells. Likewise, the toss decision and dew factor swing night-match results. Analyse without stripping these variables and we sell luck as skill.

DLS and added-time accounting. The revised target after rain does not always reflect the match's true balance. I have seen a revised target hand one side an extra advantage, only for the next day's report to turn it into a 'brave chase.' It was an algorithm's output, not heroism.

Now I come to where I am most careful. Correlation and causation are not the same thing. A team keeps winning, so we say it has momentum. But momentum is a description, not an explanation. Look at twenty matches of data and you find that winning streaks are often a mix of schedule, opponent weakness and luck — not an invisible force. Yet after every series the media writes that 'rhythm has returned,' because stories sell and fractions do not.

My second doubt arrived in 2026, when the Bundesliga returned to empty stadiums because of COVID. Tracking thirty matches, I saw the home-win rate fall from 43% to 33%. From this, many will say crowds sway referees. Possibly — but my small sample is only a signal, not proof. Holding that distinction is the analyst's job.

The xG newsletter was my first monastery; the Russian wall was my first doubt. The empty stadium was my second. An analyst who fears the empty box does not love numbers — he only loves stories. And once you learn to think in the language of balls and innings, you see that football's metrics are only guests in cricket, not residents.

As tournament pressure rises, every innings will bring a grand verdict — 'this team is the favourite,' 'this star is finished.' I would suggest holding one question: how much data stands behind the claim, and how much empty space is filled with story? The analysis that can admit its own empty boxes is the one whose forecasts deserve trust. Because cricket's most honest answer is never a number — sometimes it is silence.