The Empty Data Sheet and the Discipline of Honest Analysis: A Process-First Lesson Against Guesswork in Cricket Modelling

Half past midnight in a Melbourne flat. A spreadsheet lies open on the laptop screen, and almost every cell is empty. Only one label survives at the top — cricket_asia. The rows below repeat the same line: insufficient information, cannot assess. The scene is not new to me. From an A-League xG thread to that cold Russian night, a large part of my analytical life has passed between empty cells and the pull of guesswork. Still, every time a blank page opens, the same question returns: what do I place where I know nothing? How far should an analyst's hands withdraw when the input is zero — that is today's subject. The first stage of my work runs almost like a machine. Before entering any match or event analysis, two layers are required. The first stage extracts the raw material — information points, quotes, sources, publication dates, the parties involved, time sensitivity. The second stage drops that raw material into eight dimensions and builds an argument: format, player technique and data, squad balance, league and commercial structure, governance, risk, public expectation, and industry transmission. The problem is simple. When the first stage returns empty, the second stage has nothing to sit on. Here I built a professional habit long ago — when there is no data, the line 'insufficient information' is itself a valid answer, not an embarrassment. Readers who have followed my betting-market notes and newsletter for years know I do not manufacture stories from nothing. A model stitched together from guesswork is a question of time — when it breaks, and how much damage it does when it does. I admit, though, that the inner engineer does not stay quiet when I see empty cells. The contextual modeller and the iterative overbuilder — these two tendencies inside me always whisper: 'Two more variables and the picture would be clean.' That whisper is what I fear most. One simple truth took me twelve years to grasp in sports analysis: the scoreline is not evidence of process. Germany took twenty-six shots in one match, built 2.4 xG, held seventy per cent of the ball, and scored zero. Looking only at the result, one would say the attack failed. But after the seventieth minute, 0.09 xG per shot was telling a different story — possession without penetration. Since that night my rule has been fixed: process before scoreline, data before process. In cricket I make this translation with expected runs and expected wickets. Weighting each delivery at ball-by-ball level by its pitch, length, line, matchup and match state produces an expected number. Suppose a side in a T20 finishes fourteen overs twenty runs below its expected runs, but its expected wicket probability spikes in the last three overs. The scoreboard then tells a story of defeat, while a repeatable pattern hides inside the process — death-over ball selection, or over-bowling a spinner. I look for the cricket analogue of football's PPDA in the time gap of pressure creation. How many balls pass between releasing the ball and building fielding pressure — that indicator says whether the pressure is real or merely posturing. Add phase leverage and it becomes clear that a dot ball in the sixteenth over is worth far more than a six in the last. Now to the real question: what do I do when the input is empty? There are three honourable paths. One, admit the limit — state plainly that there is no data and no decision is possible. Two, pre-commit the sample-size floor: how many matches, how many overs, how many balls — below that, no conclusion. Three, use a rolling window, so a single night's flash does not bend the whole model. Alongside that, regularisation — before adding each new parameter, ask whether it truly improves predictive accuracy or merely beautifies the story. I also keep context layers separate — pitch character, dew, weather, day-night, DLS probability, crowd presence. In the pandemic phase, across the first forty-five empty-stadium matches, home sides won only thirty-three per cent and averaged 1.2 points. That data was proof to me that if xG does not take in crowd, travel and rest inputs, it is incomplete. In the transfer window the same lesson returns in different clothing — how loan structures and obligation clauses eat away at a smaller club's financial planning is visible only in the wage bill and the release-clause design, not in the headline rumour. This is where a contrarian word is needed. A variance-first mindset slides easily into nihilism — 'nothing is certain, so there is nothing to say.' That is wrong on both counts. Process and outcome are separate things, but the process signal is always there, if there are enough matches and the right layers. The lesson from the German night is not that results are meaningless; the lesson is that treating a result as proof is dangerous. Zero input and zero goals are not the same. One has no information; the other has information but a different outcome. The second trap is clearest in the transfer window. At this time a flood of rumour drowns the real signal. Clubs, agents, the source-less 'close circle' — everyone is selling their own story. Someone may lean toward a back three only to dodge the reputational pressure of results, to protect personal standing; the papers call it tactical evolution. Without evidence, both sides' claims are equally hollow. And injury news? Clubs release it according to their own interest, so beside a shout of 'fully fit' or 'season over,' my first task is to hunt the date and the source, not the emotion. The empty data sheet is in fact a gift, even though it looks like failure. It reminds me that finding evidence before telling a story is the analyst's first duty. The next match, the next transfer window — whichever arrives, my first question stays the same: which fact is verifiable, and which is merely noise? If an analyst does not hesitate to pour imagination into an empty cell, how far away is their next bad decision?