MCA's AI Coaching Pilot: The Gap Between Net Precision and Match Layers

A camera now sits at the Sharad Pawar Indoor Cricket Academy in BKC exactly where only a coach's eye used to be — just behind the bowler's release point. On the seventeenth ball, three numbers appear on the screen: release height, seam angle, revolutions on the ball. The coach looks at the tablet and tells himself the delivery was good.

When I rebuilt fourteen field diagrams from the January 2026 Ranji Trophy final in Indore, the first question was the same one: was the ball good, or was the ball good on that field? Inside an indoor net that question cannot even be asked, because there is no field. The AI-assisted coaching pilot the Mumbai Cricket Association has begun under president Ajinkya Naik, in partnership with Singapore-based Level Up Sportslytics, measures the ball. The match measures the over. They are being filed in the same report. This pilot is collecting coaching data and reading it as evidence for match decisions.

A field is an argument. Inside a net, that argument is never written down.

What the announcement says, and what it leaves out

The stated scope is AI in daily net sessions — training, not competition. Naik's reasoning is a change-management line: "We need to change as per the time." The association describes itself as the first to adopt AI in coaching, and the structure is explicitly a pilot: one academy, limited scope, feedback from coaches and players, then a decision.

It helps to place the announcement in its frame. MCA is a BCCI member unit and, historically, a Ranji Trophy powerhouse and a major feeder into the national side. Analytics used to sit with national teams and IPL franchises; it is now moving down to state associations. That diffusion is real, and it is where the actual significance of this story lives. The direction is clear, and a foreign vendor signing an Indian cricket association is evidence that a sports-AI vendor layer is forming beneath the elite game.

What is absent matters as much. No cost, no contract length, no pre-registered success metric, no data-ownership terms. No player is named. No performance metric is quoted. This is a top-of-funnel administrative announcement and should be read at that altitude.

Where the measurement is sound, and where the argument is not

Start with the honest part: three things genuinely travel from the net to the match. Workload — deliveries in a spell, their intensity, recovery gaps — because fatigue is physiology and an indoor net can count it. Repeatability of action — front-arm collapse through a spell, drift in seam angle, scatter in release point. And an underrated third: a shared vocabulary, which ends the "it felt like this" argument between coach and player because there is now a number on a screen. I have stood in enough nets over twelve years to know that half those arguments never resolve which pair of eyes is right. That benefit is real and should not be undercounted.

The trouble starts immediately after. The thing a net measures most cleanly — how repeatable a ball is — matters least in a match, because in a match a ball's price is set not by its cleanliness but by the cost of missing it. Miss your line in an indoor net and nothing happens. Miss it in the fourteenth over and the scoreboard moves, the captain shifts the field, and the plan for the next ball changes. The camera cannot see that difference because inside that room the difference does not exist.

This is where my 2026 mistake comes back. During the Russia World Cup I charted Croatia's compactness and spent a week trying to translate it onto cricket's powerplay ring, then rewrote the one sentence that was wrong: cricket's thirty-yard circle is not a pressing line. Football's rondo and basketball's small-sided games preserve the constraints of the match — numerical pressure, space, decision speed. Cricket's net removes all three: no fielders, no circle, no over limit, no one-bowler-per-over rule. So the football training-transfer logic does not apply here. The correct comparison is baseball's bullpen — isolated, repeatable, instrumented. And baseball spent a decade learning that bullpen tracking data mis-ranks pitchers until game-context data corrects it. Cricket's net is baseball's bullpen, not football's rondo. Naming that disanalogy in the same breath is not pedantry; without it the comparison becomes decoration.

Now the layer that actually matters. Mumbai cricket's constraint is not technical cleanliness; it is role definition. The Colts and Ranji pipeline produces technically tidy batters every year — straight blade, decent footwork, competent defence. What it does not produce reliably is roles: who bowls overs 15 to 25 in a one-dayer, who bats at five in a chase, who is the third seamer once the ball stops moving, who owns the slower-ball overs at the death. Those are structural questions, not technical ones. The AI has been bought for the technique layer while the crisis sits on the role layer.

As the resident counted-number skeptic, my question is not how many deliveries were logged. The question is how many roles were allocated differently because of them. If the answer is zero, the pilot improved practice, not outcomes.

One more item, boring until it isn't. Inside that BKC net, a fifteen-year-old's biomechanical signature is being captured by a foreign vendor. Who owns that data — MCA, the vendor, or the boy? Who took consent from a minor? What are the cross-border transfer terms? None of it appears in the announcement. A tightly scoped pilot is a good governance signal, but a low-risk structure is not a substitute for data terms.

The blind spot is the cleanest-looking part

The pilot's feedback loop will draw on coaches and players. That is the weak point. Feedback gathered in a room where the president has announced the initiative and the vendor is present is not data; it is courtesy. A coach saying "this helped" belongs in the same evidence category as a captain saying "we'll take the positives". I know that empty stadiums turn every voice into a tactical leak; an instrumented indoor net is the inverse — every ball becomes data and no ball carries a consequence. That is the cleanest dataset and the least transferable one.

The second blind spot is in the claim itself. "The first association to adopt AI in coaching" is an unverified superlative — the National Cricket Academy and IPL franchises have used biomechanics and analytics for years. Nobody's moral failure is at stake here; this is a communications error and should be fixed like one. Writing "among the first" costs nothing and removes the only reputational risk in an otherwise low-risk pilot. A rival association or a reporter can contest the claim, and the story flips overnight from adoption to exaggeration.

The test of this pilot is one bowler

The real verification is not a press release; it is a name. By the end of the 2026-27 domestic season, check whether any Mumbai age-group quick's front-arm path changed in the BKC net — then turn to a Colts match on a turf wicket in June and watch his seventh over, in the sun, under scoreboard pressure. If the change survives the field, the chalk was right and the legs arrived on time. If it does not, MCA has bought a camera, not a coach.