MCA's AI Pilot: The Road From BKC Indoor Nets to Match Conditions Is Still Unbuilt

Before reading the Mumbai Cricket Association's announcement, I opened an old file. Inside the Sharad Pawar Indoor Cricket Academy at Bandra Kurla Complex there is, right now, no scorecard being written, no toss, no DRS frame under review. What exists is an experiment: the association has tied up with Singapore-based Level Up Sportslytics to bring artificial intelligence into daily net sessions. One sentence in the release stops me — MCA is calling itself the first association to use AI in coaching. I went looking for the definition behind that claim and found a press note, not a coding log. Five hundred forty-six runs, coded by hand, because the file could not explain itself — the habit still hasn't retired. Where there is no data, asking the question is the reporting.

The Mumbai Cricket Association is a member unit of the BCCI and one of the oldest powers in the domestic Ranji Trophy structure — a large share of the supply line to the national team runs through its system. Under president Ajinkya Naik the association has taken the step, and his reasoning is plain: we need to change as per the time. According to the announcement the project is a pilot, meaning a limited trial, after which feedback from coaches and players will decide the next move. The venue is specific too — that same indoor hall at BKC.

Now the part that is not in the announcement. How many players, how many coaches, which technology, which metric defines success, what the contract length is, which budget line pays for it — none of these numbers is public. I found the number in the dataset; I could not find the person who made it. This time I did not find the number either.

The first task is auditing the claim. "First association" is a superlative, and a superlative is an assertion until the definition is fixed. What counts as AI — video annotation, pose estimation, workload modelling, or a spreadsheet logging ball-by-ball? The National Cricket Academy and several IPL franchises have run analytics tools for years, and so have private academies. It may well be that MCA is first, but without a definition the sentence is an invitation to correction, not a fact. The safer framing: among the first.

Second, and the real question: who grades the pilot? My own method has been the same for years — publish the threshold before the tournament so the season can grade you. Here the order is reversed: experiment first, criteria later. A pilot whose success markers were not set in advance cannot fail, and what cannot fail cannot inform. Feedback loop is a good phrase, but if it is unclear which question the feedback answers, the loop is not a loop.

Third, technical fidelity. The experiment runs indoors in nets — no weather, no pressure, no real pitch bounce, no crowd. A low-variability environment is safe for a trial, and that same quality makes it less transferable to the field. After the 2026 World Cup I tried to port football's expected-goals logic to T20: shot location, bowler type, match phase, batter handedness. I coded 1,248 deliveries from Mumbai's 2026-19 Syed Mushtaq Ali campaign, including Suryakumar Yadav's 72 off 39 against Baroda. The model flagged his shot selection as sustainable rather than reckless — the raw strike rate alone could not show that. But most of the work did not go into the metric; it went into testing the assumptions against cricket's discrete ball events. A metric that cannot be translated is a metric that cannot be used. How much of that indoor model survives outdoors is precisely what the announcement does not answer.

Fourth, who owns the data. A foreign vendor will hold Indian players' load, technique and video data — cross-border transfer, consent, and club-versus-player rights are all unresolved. That is not a legal footnote; it is the most expensive clause in the deal.

Fifth, the sample. One facility, zero named players, zero coded deliveries. I keep the Wankhede night in my notebook — three nights, 19 hours, 546 runs — because it reminds me that a tidy file is not ground truth, it is only tidy. Here there is no file at all.

The contrarian case is that I may be digging in the wrong place. The real story here may not be AI but procurement. A domestic association has handed a foreign sports-tech vendor access to a core coaching function, and the information asymmetry is obvious — the vendor knows the technology, MCA has to judge the value. The weakest point in such deals is always that the buyer has no benchmark to compare against. Second, the indoor choice may have been made to reduce risk, but that same choice caps the learning. And the real risk is not that AI fails; the real risk is that it succeeds at something nobody measured.

Four signals are worth tracking: whether the pilot outcome is published; whether the scope extends to outdoor pitches and age-group teams; whether Level Up Sportslytics' cricket client list grows; and whether a data-use policy is ever written. Three answers out of four would make this experiment assessable. The question is not whether Mumbai adopted AI — it is which number it can show six weeks from now that did not exist before.