When the Empty Cell Answers: What ‘No Data’ Really Means in Cricket Analysis

It is half past eleven at night at a desk in Rajshahi. A spreadsheet sits open: eight rows, and beside every one of them the same short answer — N/A. No format. No match type. No phase-by-phase performance. No teams, no players. Even the ‘time sensitivity’ cell reads: not assessed. The ground, however, was not empty. A ball landed on the 22 yards, a bat came down, an umpire raised a finger, a ball-tracking line was drawn on the third umpire’s screen. Something happened. Only the record of it never reached me. The most dangerous number in cricket is not zero. It is the empty cell into which someone writes an estimate and then calls it data. Zero is at least honest — it admits nothing is here. An estimate insists something is here, with no evidence behind it. I started on a daily newspaper’s sports desk in 2026, and the habit of keeping handwritten notes in the margin of a scorecard began there — which over the wind changed, which bowler’s run-up slowed, how long a batsman stayed at the crease. When I launched a football analytics newsletter from Rajshahi in 2026, the rule became written: source first, interpretation second. Building a live dashboard at the 2026 World Cup in Russia did not change it. I moved from a Rajshahi newsletter to live World Cup analysis, and the discipline never changed. Cricket’s record-keeping has never been neutral. Its official results have repeatedly been written by a clause in the rulebook rather than by play. On 13 March 2026 at Eden Gardens, when crowd trouble stopped the World Cup semi-final, match referee Clive Lloyd awarded the match to Sri Lanka. India’s scorecard says ‘lost’, but that loss was not the product of any bowling performance — India had halted at 120 for 8. The match never finished. On 28 April 2026 in Bridgetown, the World Cup final was settled by Duckworth-Lewis-Stern. Rain came, overs were cut, the target was revised, and the back end of the second innings was played in near-darkness. A result was declared, but no full 100-over examination sat inside it. On 14 July 2026 at Lord’s, a boundary-count clause decided the final. England and New Zealand both finished on 241 after 50 overs. Both scored 15 in the Super Over. The match was tied. Then a number settled it: England 22 boundaries, New Zealand 16. England were world champions. The ICC abolished the boundary count in 2026 and introduced repeated Super Overs, but the 2026 result stands unchanged. All three cases follow one pattern: the outcome was produced by an administrative decision process only partly related to bat-and-ball performance. The analyst’s job is to isolate that process. Otherwise he is not measuring what he believes he is measuring. Suppose you are building a ‘pressure handling’ model. Inputs: required run rate over the last five overs, wickets lost, Super Over performance. Output: win or loss. In that match England’s output is 1, New Zealand’s 0. But the decision process behind that 1 was not bat-and-ball. A model that does not encode the assignment mechanism learns the wrong lesson: it manufactures a myth called ‘luck in the final over’ with nothing underneath it. There is a subtler point. Win-probability models had New Zealand ahead for long stretches of that night. When Trent Boult’s delivery deflected off Ben Stokes’ bat to the boundary in the last over, the extra runs were an accident rather than a skill. Yet in post-match discussion it became ‘England’s mental resilience’. The spreadsheet remembers what the stadium forgets — but if the spreadsheet receives an empty cell at that moment, it remembers nothing at all. What expected goals are to football, expected runs and win probability are to cricket. Expected goals are confessions, not predictions: the model is confessing what it knows and what it does not. This is where the empty cell and zero diverge. If a bowler is not given five overs, his economy is not 0.00 — it is unknown. If a side does not bat 35 overs because of rain, its strike rate is not zero — it is absent. Put both in the same database column and the whole analysis is contaminated, because averaging will treat the missing value as zero. Feed the 2026 semi-final into a ‘home-ground pressure’ model and it teaches the model something false, because the match never finished. Feed the 2026 final into a ‘DLS skill’ model and a rain-shortened match carries the same weight as a full one. Feed the 2026 final into a ‘tie-breaker nerve’ model and a rule change is read as a psychological trait. DRS ‘umpire’s call’ is the cleanest example of the problem. How far the ball pitched inside the line is measurable; whether it would have hit the stumps is decided by a tolerance threshold — an administrative boundary. An analyst who waves that threshold away as ‘luck’ has discarded the most important input in the model. In Bangladesh’s domestic cricket the distinction is even more visible. Many National Cricket League matches carry no ball-by-ball tracking and no over-by-over line-length data. When a national-team match and a domestic match enter the same ‘player depth’ conversation, the two levels of evidence are entirely different — yet the reader places them on the same shelf. The fix is not complicated, only uncomfortable. My own method now rests on three pillars. First, I split every result into three columns: observed (what happened on the ground), inferred (what I am concluding from the data), and assigned (what was decided administratively). Second, a cell that is empty stays empty, and I say so in the report. Third, every forecast carries a confidence level, and I reconcile it afterwards. The third pillar is the most neglected. A written confidence level makes it easier to catch your own error later, and that is the only way to stay honest about your forecasting ability. Working on Morocco’s defensive structure at the 2026 Qatar World Cup, I ran into the same problem. Morocco — Root: 2026 Qatar World Cup and Morocco. One goal conceded across five matches, an own goal at that, 1.2 xGA, a PPDA of 13.5 — these numbers were thinly recorded, because broadcast cameras follow the valued team. Where data is scarce, the analyst fills the gap with story. ‘Morocco surprised everyone’ is pleasant to hear; ‘Morocco’s block was structurally disciplined’ is more useful, and much harder to write. What 55 Bundesliga matches in empty stadiums taught me in 2026 points the same way. Empty stadiums did not silence football; they exposed its skeleton. Cricket is no different. Missing information does not stay hidden; it breeds the most speculation. Sports data firms are now testing append-only ledgers. The blockchain property that matters here is not the token but immutability. Each match gets a timestamped entry, and if a cell is empty, the statement ‘this value was not available’ becomes a permanent record in itself. Nobody can quietly fill it in later. That is the real gain. The problem is not solved — the empty cell stays empty. But the question of who left it empty, and when, can no longer be erased. With an audit trail, the line between inference and observation stops depending on personal honesty and enters the system itself. Now an uncomfortable admission. The urge to fill empty cells is not an analyst’s weakness; it is the economics of the trade. The deadline arrives at 11:45 pm. The editor wants a number. The reader wants to know who is ahead. Writing ‘no data’ in an empty cell reads as incompetence; writing an estimate reads as expertise. The system rewards the estimate. A filled empty cell is far more damaging than a genuinely empty one. The first grants institutional legitimacy to an error. Once the error enters the report, it becomes ‘historical trend’ the following season. Two years later nobody asks where the number came from. Another trap waits. Teams and players whose data is thinly recorded — usually associate nations, women’s cricket, domestic circuits — invite romantic storytelling. I have fallen into that trap myself. My fix is to define in advance which conditions would make me say ‘this side has genuinely improved’ — a falsifiable criterion. Without a pre-committed criterion, the temptation to bend the data to fit it later is overwhelming. In franchise cricket the problem sharpens. The IPL has tracking data for nearly every ball, while a small domestic T20 tournament’s scorecard often lacks even line-length. Two sides of the same comparison arrive with entirely different grades of evidence, yet sit side by side in the table. The next time a pipeline comes back empty mid-World Cup — and it will — the question will not be which number to insert. It will be: did I declare this empty cell in advance? If I state before a tournament which data I will not have, the freedom to improvise mid-tournament removes itself. Without that rule, every match leaves us with a beautiful story and a lost record.