The Empty Spreadsheet Season: Where Does Cricket's Prediction Industry Get the Confidence It Sells?

On the eve of the Champions Trophy final, a data friend of mine turned his laptop towards me in a Dubai hotel lobby. Eight columns, thirty-two variables, every cell either 'not applicable' or zero. On top sat a confident conclusion: 'India ahead in the final.' I asked where the data was. The answer: 'The pipeline isn't set up yet.' I opened Excel to check a hunch, and a religion died — cricket's prediction industry now sells structure, not substance. The more professional cricket analysis has looked over the past two years, the emptier it has become. With graphics and slide-deck jewellery we have built a room with a padlock painted on the door and no furniture inside. In that Barbados final, South Africa needed 30 off 30; no slide deck called it, and none ever will.

For three years one sentence has held scripture status in cricket's market: 'Data changed everything.' Australia beat India in the 2026 World Cup final in Ahmedabad; India beat South Africa by seven runs in the 2026 T20 World Cup final in Barbados; India won the 2026 Champions Trophy in Dubai; RCB took their first IPL title in 2026. Behind every result someone has inserted the word 'process', as if defeat meant an immature mind and victory a flawless spreadsheet. Broadcast graphics now show a probability number after every ball, franchises run their own data departments, and teams build thousand-line models before auctions. Faith is born by force of repetition: prediction is now science.

The numbers do not stay on screen; they move money. In South Asia the fantasy-sports and betting-advertising market has multiplied over the past decade; every IPL season, tens of millions build 'probable XIs' for every match. The fuel for this demand is confidence, not accuracy. The broadcaster wants eyes stuck to the screen, the sponsor wants a flash of numbers, the team wants a tidy explanation after a defeat. Three interests combine into a slide economy where the model's job is not prediction but expectation management.

This column is the ledger of that economy. I am not an enemy of cricket analysis; I am an accountant. An accountant's job, when someone says 'my model called this', is to ask for the timestamp. I have done exactly this in football for two decades — when everyone was worshipping possession as a deity, I built an Excel model from 380 matches and showed that Chelsea's 93-point title came on just 54 percent possession; possession was the altar, the data was the hammer. Cricket needs the same test, because cricket has its own gods — 'intent', 'impact', 'momentum'. All three are revered, all three almost untested.

Test one: the base rate. The arithmetic reality of T20 cricket is that a ball, a catch, a toss control more than 40 percent of the outcome. At the 2026 T20 World Cup, the USA beat Babar Azam's Pakistan in a Super Over in Dallas on June 6, 2026, and Rashid Khan's Afghanistan beat Australia by 21 runs in Kingstown on June 22, 2026. Which pre-tournament model's probability distribution had those two results as normal? Almost none. But the models never admit error; they defend themselves in the language of 'it wasn't said', the way a boy who did not study before an exam seeks a blessing.

Larger samples make it starker. At the 2026 ODI World Cup, Australia began by losing two matches and then won nine in a row to take the title; Rohit Sharma's India won all ten league games and lost the final. The same team's fortunes swung so violently inside one tournament, and still nobody called it in advance. In the final, after India were bowled out for 240, Pat Cummins' side chased the target in 43 overs — that is not 'momentum', that is the cold arithmetic of the scoreboard.

My own database holds more than 1,500 T20 matches; the simple count: a clear favourite wins roughly 58 to 62 percent of the time pre-match. If the base rate is 58 and your model says 58, you have not made a prediction — you have quoted a ranking table. The gap opens only when a model states specifically that 'the favourite loses this one, because bowling triggers and venue patterns say the opposite'. Such falsifiable claims are rare in this industry.

The second god, 'momentum', is never defined and never measured. If the run rate rises over five overs, the team is deemed to be surging; yet one wicket and that momentum evaporates within an over. A word that can explain any result is not explanation, it is poetry. Twenty-one years of watching matches taught me that the thing called momentum is often two or three misread dot balls.

Test three: 'intent' is cricket's possession worship. In an auction, the batter holding a strike rate above 180 is bought first; on broadcast, boundary percentage becomes glittering graphics. Yet the match-winning statistic is avoiding dot balls and taking wickets in the right overs — ball-by-ball pressure, a structure exactly like possession. India won the 2026 T20 World Cup on Jasprit Bumrah's death-over economy; he was Player of the Tournament and that year's ICC Cricketer of the Year. Bumrah's value lies not in strike rate but in absorbing pressure.

The IPL's Impact Player rule, introduced in 2026, deepened batting and gave broadcasters a new graphic — but the fate of matches was settled by the pressure of the final over, not the rule. A side that buys 'intent' in search of a title usually loses its grip in the middle overs, just as possession worshippers used to boast of possession without goals.

Test four: the auction model versus agent noise. Before an IPL auction, teams build two-hundred-line spreadsheets — talent indices, age curves, venue splits. Then, in two minutes of bidding war in the auction room, the model becomes decoration. I went backstage at the transfer window with a calculator and no manners; I saw that price is set by an agent's story, not a team's arithmetic. RCB won their first IPL title in 2026, and that is real; but much of the 'auction model succeeded' coverage before and after the title was post-match storytelling — the model that ranked a different player first three months earlier quietly swaps its name after the final.

Test five: structure before the individual. At the 2026 Champions Trophy, under the hybrid model, India played every match in Dubai — same venue, familiar conditions, no travel fatigue. That is not a player's 'pressure handling', it is an accounting of advantage. At the 2026 T20 World Cup, the drop-in pitches and variable weather in New York and Dallas dragged the favourite hierarchy downward. I often say that learning to read a fixture list matters more than learning to watch a match. A prediction that ignores schedule, travel, dew, pitch and format is not a prediction, it is a wish.

Test six: the analyst in the dressing room. In recent years analysts' roles in franchises and national teams have grown — which is welcome, as long as they measure the rhythm of the match. But a match's rhythm and a spreadsheet's rhythm are not the same; the language of a bowler's shoulder in the 42nd over appears in no line. When an analyst turns statistics into instruction, cricket shifts from a game on the field to a game on slides.

And one uncomfortable truth: in 2026, during the pandemic break, I watched all 81 Bundesliga matches played behind closed doors and calculated that home wins fell from 43 to 33 percent — meaning nobody had measured how much a crowd actually controls results. Cricket has plenty of such unmeasured variables: dew, floodlights, boundary dimensions. A model that drops them gives confidence, not information.

So what is the work? Simple — pre-register the variables, then run the data. If a model claims 'the favourite loses', the reason, the conditions and the confidence level must be written down first. When it fails, correct it publicly. That is my method, and it is what most slide decks lack: they carry only conclusions, never a cell for admitting error.

Across these six tests, the conclusion is uncomfortable: cricket's prediction industry now manufactures confidence, not knowledge. Broadcast needs numbers, boards need alibis, agents need narrative — three demands combine into a slide civilisation with eight columns and no data.

I must stop here, because my claim is growing too broad. Honestly, the strongest case against me is this: models do work when samples are large and conditions stable. In Test cricket, in long leagues, in managing Bumrah's workload, India's data governance has genuinely delivered — I must concede that. My friend's empty model may be a broken pipeline rather than a philosophy; I may be indicting an entire industry for one lazy deck. I admit this is my old disease — I want to see the pattern so early that I overrun the evidence.

So let me make the claim falsifiable. The condition: if, before any upcoming ICC tournament, any public model can show accuracy consistently two points above a toss-and-ranking baseline, I will withdraw this column. And let me keep my own receipts file honest: before the 2026 T20 World Cup, my pre-tournament file scored 61 percent on picking match winners — which sounds good until I remember a 'favourite wins' bot scored 59. Two percentage points is not worth an entire slide civilisation.

My forecast for the next ICC event, with a date attached: on the morning after the tournament, when the aggregate accuracy of public pre-tournament models is measured, it will fall within two points of the toss-and-ranking baseline. If I am wrong, the receipt goes in the file, dated — a wrong timestamp is better than a hidden error. The question now belongs to the reader: do you want numbers, or a tidy confidence?