When the Spreadsheet Goes Blank: The Trap of Inference and the Chain of Evidence in Cricket Analysis

A column, a cell, and inside it three letters — N/A. Late December 2026, laptop open on the work table at home in Rajshahi, a spiral-bound notebook beside it. Nearly half past eleven at night. I was cross-checking a pass-completion column from a match I had watched with my own eyes, and exactly there the cell was blank. The passes were still glowing in my memory. My hand tensed — should I just fill it in? Write 'roughly 88 percent' and be done. Nobody would verify the number. That urge to fill the gap is, over more than twenty years in this trade, the most dangerous habit I know.

That night I saved the file with the cell still empty. The next morning I pulled the data from the original scorecard and filled the column — 86.4 percent, after correcting a small error. The number was lower than my estimate. That small gap is the foundation of my entire method: if the input is blank, the output stays blank.

Context: From Rajshahi to the World Cup, the discipline never changed

I moved from a small cricket-and-football newsletter in Rajshahi to live World Cup analysis, and the discipline never changed. When I started the football analytics newsletter Expected Truth in 2026, the rule was already in place: four things beside every number — source, timestamp, definition, benchmark. The live xG/PPDA dashboard I built for Belgium versus Japan at the 2026 Russia World Cup stood on the same rule. Newsletter or World Cup dashboard, a number alone says nothing; the chain behind it does the talking.

In this profession we work in two stages. Stage one breaks a match or an article down into small information points — who played, what happened, in which minute, from which source. Stage two builds the structure of analysis on top of those points. The rule is simple and hard: stage two cannot take a single step beyond stage one. If stage one comes back empty, the honest answer at stage two is exactly one thing — insufficient information. That is not failure. That is the mechanism working correctly.

The problem is that this honesty does not sell. And that is where today's argument begins. I am writing this mid regular season, so the question matters even more: before we read the undercurrents beneath the table, we need to know what we actually know.

Core: The anatomy of a trustworthy number

I want to walk through seven cases to show how an analyzable number stands up — and what it becomes when you delete the inputs.

In La Liga 2026-17, Lionel Messi scored 37 goals from 26.3 expected goals, a +10.7 overperformance. The number is striking, but the real lesson for me is elsewhere: the +10.7 held because of shot-level data, a full-season sample, and a league-wide benchmark. Expected goals are confessions, not predictions — where the shot was taken from, under what pressure, from which angle; the model confesses that. Delete the shot data and the +10.7 disappears; what remains is a romantic notion that Messi was extraordinary. The notion is not wrong, but it is not analysis. It is memory. And memory cannot measure the future.

At the 2026 Russia World Cup, Belgium beat Japan 3-2. Everyone remembers the story: Japan went two goals up, then lost late. The dashboard tells another story. After the 60th minute Japan's PPDA rose from 7.9 to 14.3 — they could not sustain the earlier pressure and were forced to allow more passes per defensive action. PPDA measures how many passes you concede per defensive action; when it climbs, the pressing structure is going soft. With minute-level data, the explanation shifts from 'Japan ran out of gas' to 'Japan's pressing structure broke, and Belgium exploited it.' The spreadsheet remembers what the stadium forgets — the stadium remembers the pain of the defeat; the spreadsheet remembers which pass was let go in which minute.

In the regular season, that distinction is my core work. A team's PPDA creeping up over three matches is a signal that never shows in the table and never makes a headline, yet it hints at fatigue and structural decay. Title pressure, relegation fear, the pattern of a referee's decisions — these three currents can be seen before they become headlines, if the information points were gathered in advance.

Now Morocco, because that team is my recurring underdog stress test. At the 2026 Qatar World Cup, across five matches before the semifinal, Morocco conceded only one goal — an own goal — with an xGA of 1.2 and a PPDA of 13.5. — Root: 2026 Qatar World Cup and Morocco. If someone reads those numbers and says luck helped, the question should be: what does zero goals conceded across five matches, own goal excluded, actually mean? It means the defensive structure controlled the quality of the opponent's shots. The low block is not cowardice; it is architecture — but for that claim to hold, my definition of xGA must stay honest: which shots count as big chances and which do not. If the model's definition changes, Morocco's credit can change too. I say that risk out loud rather than hide it.

In the January 2026 transfer window I looked at Sofyan Amrabat — 89 percent pass completion, 8.7 progressive passes per 90, 2.3 tackles per 90. The January transfer window is a liquidity event for hope, and I audit the books. The 89 percent figure alone says nothing — a centre-back can reach 95 percent by rolling five-yard passes sideways. Only with progressive passes and tackles together do you see where the value of the passing lies and how much defensive work is involved. That piece was later picked up by a European scouting network, and a consulting offer followed. I treat that as external recognition of the method, not as proof — because even if an outside party picks it up, the responsibility for choosing the variables remains mine.

In 2026, during the pandemic hiatus, I analysed 55 Bundesliga matches played in empty stadiums: the home-win rate fell from 43.3 percent to 33.3 percent, while away teams' PPDA and distance covered rose over the same period. Empty stadiums did not silence football; they exposed its skeleton. But a caution is essential here — 55 matches is a small sample, and alongside the absence of crowds, scheduling, weather, and refereeing decisions may all be involved in the decline of home advantage. An honest model also states what it cannot isolate. Turning the pandemic sample into a final verdict would, to me, be a misuse of data.

At Euro 2026, in Italy versus Spain, I tracked Jorginho — 92 passes, 8 progressive carries; Italy's PPDA 11.2 against Spain's 7.8. Spain pressed harder; Italy controlled midfield with less pressing. Here is a subtle lesson: possession and progression are not the same thing. If those 92 passes are sideways circulation, that is the appearance of control, not control. At the Tokyo Olympics women's football final, Canada's xG was 1.1 against Sweden's 0.7. Tokyo Olympics without crowds was a controlled experiment in pure signal. Yet on a one-match sample my rule is ranges, not prophecies.

Each of these seven cases has five layers — source, timestamp, definition, benchmark, and a holdout sample. The blank input has none of the five. Analysis standing on zero is not analysis; it is certainty dressed in the clothes of inference. In cricket that risk is larger still, because T20 samples are small, match-up dependence is high, and the luck component of the toss and DLS is significant. Stripping out luck requires raw data — without raw data, luck simply looks like skill.

Contrarian: The blank cell is the most honest cell

Here is the strange truth: the market pays for certainty, not for blanks. A confident paragraph with no data behind it travels far; an honest 'insufficient information' gets scrolled past. This incentive structure works exactly the way the phrase load management becomes a polite name for a commercial tour. Analysis, too, can quietly become a polite name for invention.

There is a deeper problem here, bigger than a blank input. Much of our cricket opinion is actually a stage-two conclusion when stage one was never run at all. The most repeated phrases — in form, cannot handle pressure — are not information points; they are impressions. I call this the reverse risk of spreadsheet overconfidence: keeping your eyes on the pitch and forgetting the notebook. From years of watching matches in the ground and on screen, my experience says the eye and the notebook correct each other; drop one and the other's excesses go unchecked. So my rule: beside every model, at least one eyes-on scouting note and at least one live-updatable variable.

Another trap — underdog romanticism. Morocco is my favourite case because they showed a structure can be built against wealthier opponents. But precisely because I love that case, I must pre-commit falsifiable conditions: in which match does Morocco's model break? If xGA crosses a set threshold, the structure is not holding — that must be written in advance, not afterwards. Otherwise I keep adding variables while hunting for explanations, and eventually every outcome is explainable and nothing is predictable. Contextual overfitting happens exactly this way — a maximum of three core variables per piece, the rest context; I impose that ceiling on myself.

The contrarian truth is harsher still: the audience is complicit in this system. We share the confident take and scroll past the blank cell. So the responsibility is not the analyst's alone — it belongs to expectation too.

Takeaway

The next round's question should therefore be about trust in the chain of data. Beside any cricket claim I will want three things: source, time, and a confidence level. I will give numbers as ranges and keep forecasts in a public tracker — so that errors surface and correction becomes part of the method rather than a failure. Learn to read the blank cell as data, and it becomes our most valuable signal. From that small room in Rajshahi to the live World Cup dashboard, I have learned one thing: honesty has a cost, but inference costs more.

Before the next match, will the question be 'who wins' — or 'what do we actually know, and what do we not'?