The Model That Knows How to Stay Silent — The Quiet Honesty of Null Results in Cricket Analytics

A small apartment in Sydney. Nearly two in the morning. A spreadsheet open on the laptop screen — over number on the left, run rate on the right, wicket probability, and the output of my own hand-built expected-runs model. Right then a client message arrived: "What's your read on tomorrow's match?" I opened the file and saw the feed was empty. No innings data, no venue report, not a single line on how the pitch is behaving, and no certainty about who is even playing. In that moment two paths lay open. One, fill the gap with imagination, spin a story so the client is satisfied. Two, say it plainly — there isn't enough information, so right now I won't say anything. I chose the second. This whole piece is an argument for that decision. The reason is simple, and dangerous. When you're handed an empty file, the easiest thing is to make something up. And a fabricated line tastes sweet at first, then costs dearly. In the betting market a wrong estimate is paid for in money, and a wrong analysis is paid for in reputation. What is cricket analysis, really? I see it in three layers. The first layer — raw data. Which format, which innings, which over, who is batting, who is bowling, how the pitch looks, whether dew is falling, how strong the wind is. The second layer — giving that data meaning. Strike rate, economy rate, dot-ball percentage, boundary percentage, the different behaviour of powerplay and death overs, the matchup accounting of spinner versus seamer. The third layer — the decision. Every step across these three layers needs an input. No input, no output, however elegant the model. My roots are in football. In 2026, during the World Cup, I built my first expected-goals model. Later, coming to cricket, I understood that cricket data and football data are not the same. In football you can price a shot; in cricket the value of a single ball depends on the over, the wickets down, and the state of the match. The same six balls, in the powerplay and at the death, are two different worlds. From my years of watching matches I've learned one thing — cricket's numbers never speak on an empty stomach. You have to feed them format, pitch, weather and match state. A Test strike rate and a T20 strike rate are not the same thing. A day-one pitch and a day-five pitch are not the same. An opener's job and a finisher's job are not the same. If someone throws out an average number without reconciling these differences, that isn't analysis, it's just noise. This is where I follow a rule of my own: small samples are loud, large samples are honest. If someone scores two hundred in two innings, a headline appears. But the analyst's job isn't to stand behind the headline, it's to find the twenty-innings sample. Take Don Bradman — 99.94 across 52 Tests. Take Sachin Tendulkar — two hundred Tests, more than a hundred international centuries. These numbers are so solid because behind them sit enormous samples and a clear process. A two-innings flash can never become such a number. A big part of my work is building weekly briefs for the betting market. And in that work I follow an eight-dimension filter. When a match or a piece of news comes up, I go through eight doors. Doors mean checkpoints — one question at a time, one answer at a time. The first door — format and match. Is it a Test, an ODI, or a T20? Which phase of the innings? Where is the venue, how is the pitch? Dew, wind, rain — is there a Duckworth-Lewis possibility? If this door isn't passed, the other seven are meaningless. Pulling a Test conclusion from T20 data means the right calculation in the wrong format. The second door — player technique and data. Whose batting average, whose strike rate, whose economy. Situational splits — what they do in the powerplay, what they do in the death overs, what they do at home, what they do on foreign pitches. Reading recent trend against career trajectory. Confusing a single innings' flash with a career is the biggest mistake. The third door — team profile and ranking. ICC rankings, home-and-away profile, batting depth, bowling combination, bench strength, age structure. Which style works against which team — this matchup accounting is part of this door too. If a team is weak against spin, that may not show in its overall ranking, but in a specific matchup it is decisive. The fourth door — league and commercial ecosystem. Broadcast-rights value, franchise price, player salaries, the gap between auction price and playing merit. If a player commands a huge price at auction, that is not proof of his cricket skill — it is a market estimate. The tug-of-war between league and national duty also shows up at this door, when a packed schedule erodes both a player's body and his form. The fifth door — rules and governance. Power and revenue distribution, controversies over playing rules, integrity and corruption, eligibility and selection, political influence. A small rule change can sometimes overturn an entire strategy. And a selection controversy can overturn an entire team's narrative. The sixth door — the risk side. Injury, schedule load, financial risk, reputational risk. Injury especially. When a player returns from a serious injury, the barrier in his head is harder to clear than the one in his body. On paper he is fit, but the data doesn't tell you what is going on in his mind. The seventh door — public narrative and expectation. How wide is the gap between market expectation and objective assessment. That gap is actually the most valuable information, because the betting market runs on expectation. The eighth door — industry transmission. From the grassroots to the national team, then broadcast and the betting market — how an event travels along this chain. When a star rises it doesn't just change the team's arithmetic, it changes broadcast arithmetic, sponsor arithmetic. These eight doors are not decoration for me, they are a model's foundation. And here is the real point — if there is no information behind one of the doors, there is only one honest answer: it cannot be assessed right now. This is called a null result. A null result is not a failure, a null result is a valid conclusion. Let me give a real case. Once, in a brief, I wrote a team down as favourites purely on five recent wins. Later, digging through full innings-level data, I understood that three of those five wins had come on easy home pitches, and against weak opposition. When the pitch changes, much of that form evaporates. That mistake taught me that none of the eight doors can be skipped over. Another example. Word comes from someone — "this bowler isn't playing the next match." Now the first question is, what's the source? Someone's tweet, or a selection committee announcement? If the source is a rumour, then it is a prior — an estimate. The real information comes after the announcement. A transfer rumour is a prior; the medical is the posterior. If an analyst takes the prior for the posterior, his model walks the wrong path. Another example. A home team has been winning at one venue for a long time. People say home advantage is overwhelming here. But if I don't separate the pitch, the dew and the crowd's role, I'll miss the real cause. I remember 2026, when stadiums across the world were empty. Then it turned out home advantage hadn't fully disappeared. Empty stadiums did not erase home advantage; they exposed its source. Meaning the edge isn't only the crowd's roar, it's also the pitch and the familiar environment. That same year I understood — the model said one thing, the empty stadium said another, and the gap between the two was the real lesson. This is why I say, I do not trust a number I cannot trace to a touch. If I don't know exactly which balls sit behind a strike rate, the number is just a made-up face to me. And where a data feed is empty, the most credible analysis of all is a plain admission. Now the other side. Honesty has a danger. If someone turns a lack of data into an excuse and always sits there saying "no information," he can never give a decision. The analyst's job isn't only to doubt, it's to build the best possible estimate within limited information — but to label the estimate clearly as an estimate. Here lies a subtle trap. Correlation and causation are not the same. Two things happening together doesn't make one the cause of the other. A batter scores with a new bat, so the bat is the cause — the story is sweet, but unsupported. I always remind myself that a number and a cause are not the same thing. In the betting market, not understanding this gap lets you keep losing, but understanding it clears the path to profit. Another trap — defending your own model past its limits. Every model has an assumption, an error margin, and a falsification condition. The analyst who writes them down makes his own word testable. The one who hides them makes his own story sacred. A sacred story and hard data don't run together. So the next time someone demands a strong opinion on a match, first ask — where did the information come from? What's the format, how's the pitch, how big is the sample? If the answer is empty, the bravest answer is a clear "I don't know." Because a model that knows how to stay silent earns more right to be heard. And spinning a story from an empty file isn't analysis — it's just words.