A Hashed Error Never Becomes True: Data Provenance and the Limits of Blockchain in Asian Cricket

Hook

On 19 November 2026, in front of 92,000 people at the Narendra Modi Stadium in Ahmedabad, Travis Head made 137 off 120 balls. On my laptop, the boundary shots from that innings sat in two tagging sets — one pulled automatically from the live event feed, one built by hand watching the match back. The two sets did not agree; the gap came to 0.4 percent of total runs, across three deliveries. The number is trivial. The question is not: every figure we quote, every trophy we hang on it, every judgement we pass on a cricketer — where did it come from, and who will vouch for it?

For years a line has sat in my notebook: the eye test is a witness, the data is the cross-examination. If the cross-examination notes are wrong, the case collapses and nobody in the court notices. Right now, in Asian cricket, large claims are being made in the name of blockchain at exactly this weak point. My job is simple — build a table, then wait and see what arrives.

Context: Who Keeps The Match's Books

A single cricket ball, rolling away, generates data in at least four separate layers. The first is the ball-by-ball event feed — runs, batter, bowler, field position. The second is ball-tracking and replay technology, which drives DRS and broadcast graphics. The third is the match referee's and match-management's official log. The fourth is the weakest and most discussed: human tagging decisions. Which ball was a mistimed four, which was an edge, which was simply the wrong line.

In Asia, the commercial value of these quiet layers is enormous. ICC media and data rights, the IPL live event feed, the domestic broadcasts of Pakistan, Bangladesh and Sri Lanka — different vendors, different labelling conventions, different missingness. I have seen the same kind of line-and-length delivery mapped to different grids in two countries' domestic tournaments, because the two feeds use different pitch-mapping reference points.

Into that gap stepped cricket-focused NFT platforms. In 2026 FanCraze partnered with the ICC and entered the digital collectibles market, and in 2026 Rario announced long-term partnerships with Cricket Australia and several IPL franchises. Both arrived in the same packaging — if it is written on a blockchain, the record cannot be altered.

But it is worth saying plainly what a blockchain actually does. When you hash a record and chain it to the previous block, the only thing proved is this: the record has not changed since it was anchored. It is not proof that the record was right in the first place. That may sound like a small remark; it is the central flaw in the blockchain narrative in Asian cricket.

Core: An Audit of 407 Deliveries

Let me open up my own process, because without the method the numbers mean nothing. I manually re-tagged a portion of the 407 legal deliveries from the 2026 ODI World Cup final, watching television replays, writing down each delivery's line, length, bat contact and footwork separately.

First, the baseline. My pressure-adjusted model for batting first on that Ahmedabad pitch, having lost the toss, said 289 runs in 50 overs, standard deviation 24. India made 240. Deviation: minus 49 — roughly two standard deviations below. That is not a bad day; that is a structural failure, and it needs to be split by phase to see where it happened.

Powerplay, first ten overs: India 80 for 2. My baseline was 74 for 2. They were above baseline here, scoring at 8.00 an over. The match was not lost at the top.

The real break came between overs 11 and 40: 122 runs in 29 overs, a run rate of 4.20, four wickets. My baseline for that phase was 168. Deviation: minus 46 — that is, 94 percent of the final shortfall was deposited in the middle 29 overs. Not in the first powerplay, not in the death overs. In the middle.

Look at the mechanism and the picture sharpens. Australia's seamers did not bowl wide yorkers four times over inside ten overs; they bowled slightly back of a length, hitting the deck. India's middle order defended those balls instead of striking them. The dot-ball rate in that 29-over phase sat close to four an over. A dot ball is not merely a run not scored — the batter changes his shot plan on that delivery, and the bowler uses that information next over. This is what I call structural pressure, and it is a completely different thing from what is sold under the name 'momentum'.

Now look at Australia's start in the second innings. Forty-seven for three after ten overs — Warner, Marsh and Smith gone. My chase model gave Australia a 23 percent win probability from that position. Then the Head-Labuschagne stand added 192, and the match finished in 43 overs at 241 for 4. The deviation is positive, but the deviation belongs to one batter, and it too runs against the lower-probability path.

Now the real question: across those 407 deliveries, which of the four layers matters most? The first three are populated automatically. The fourth is populated by human judgement. And the model scores its own performance on that fourth layer.

Blockchain can make the first three layers immutable; it cannot make the fourth one correct. That is exactly where Asia's blockchain projects took the easy road: instead of auditing the whole pipeline, they built a separate market — 'moments' for collectors. A great catch, a six, a delivery. The token becomes the commodity, not the truth of the data. Two entirely different questions, sold under one banner.

Contrarian: Immutability Is Not Accuracy

This is where I see the biggest problem, and it is also a question of being honest about my own trade. Immutability's main limitation is that it closes the path to correction. Suppose a ball is written on-chain as caught behind, and the replay later shows it came off the bat's edge. An automated labelling model got it wrong; humans get it wrong too. But you cannot delete what is written on-chain — you can only write a correction in a later block. Five years on, the researcher who downloads that dataset gets the original error and never sees the correction. In baseline auditing that is fatal, because people like me train models on years and years of old ball-by-ball data every single day.

The second problem is who holds the keys. If the validator nodes are run by a consortium of a few boards and broadcasters, that is not a distributed ledger; it is a permissioned database — a bit of extra work and a great deal of affection. The balance of power in Asian cricket is so uneven that a consortium automatically means the big board's terms. If a small country's domestic feed is forced to adopt the big board's hash standard, the weakest pipeline in the provenance chain becomes permanent.

The third objection is not about transparency, it is about metrics. The price a token sold for is written clearly on-chain, but whether that catch was extraordinary or routine — which human, which rule, which confidence interval sat behind that decision — appears nowhere.

I do not chase narratives. I build a table and wait for them to arrive and sit down. But it is time to say one thing, because I am watching this from inside the ecosystem: anyone selling something new in cricket data should publish their tagging protocol first, and the token second.

Takeaway

Not every era of our cricket needs a blockchain — a blockchain is a hash, and a hash answers only one thing, quietly: this object has not changed. When someone tells you every data point from the final is anchored on-chain, ask two questions. First: what exactly is the hash placed over — the event log, or the tagging decisions? Second: before anchoring, who audited it, and where is the audit code?

On that November evening, as Head walked off with 137, I thought a familiar thought again — the first xG model I built did not predict football; it predicted my patience. Cricket now needs exactly that patience again, because data problems never begin with technology. They begin with human decisions.