Umpiring AI Trial May 2026 — ICC's BCB-PCB Test Series Pilot Decoded

Share this article
The ICC's AI umpiring pilot ran quietly through the Pakistan-Bangladesh Test series. The system was used in parallel with the existing TV umpire process for no-balls and stumping decisions. It was not used as the primary decision-making tool. The pilot ended on May 12. The data is now being reviewed by the ICC cricket committee.
What the system actually does
The AI no-ball detection system uses high-frame-rate cameras at the popping crease and a trained model that flags front-foot landing position relative to the line. The system processes the delivery in approximately 0.6 seconds. The output is a confidence-rated flag that goes to the TV umpire, who retains the final decision authority. The system does not call no-balls. It flags them.
The stumping component
The stumping detection component uses the same high-frame-rate cameras to track the bail position relative to the foot crossing the line. The system reads bail movement to confirm dislodgement. The stumping component is the higher-confidence part of the system because bail movement is a binary event that the cameras can resolve with high accuracy.
The false-positive rate
Across the two-Test series, the AI no-ball system flagged 312 deliveries. Of those, 287 were confirmed no-balls by the TV umpire. The false-positive rate was approximately 8 percent. The false-negative rate, measured against post-match video review, was approximately 3 percent. The system is operating at a higher accuracy level than the existing TV umpire no-ball review.
The umpires' reaction
The two TV umpires assigned to the series, who used the AI flag as a parallel input, gave a positive on-record assessment. The on-field umpires gave a more cautious assessment. The concern from the on-field side is that the AI flag, once embedded, could erode the on-field umpire's judgment over time. The argument is procedural, not technological.
The latency question
The system's 0.6-second processing time is fast enough to flag no-balls before the next delivery is bowled. The TV umpire can intervene to call a no-ball immediately, which means a batter dismissed off a no-ball delivery does not actually leave the field. The latency is the single biggest improvement over the existing TV umpire review, which has a multi-second processing window.
What this means for DRS
The AI system does not replace DRS. It runs parallel to it. DRS continues to handle lbw and caught-behind reviews. The AI system handles no-balls and stumpings only. The cricket committee is considering whether to extend the AI to bat-pad catches and wide deliveries, but those extensions require larger training datasets and longer pilot windows.
The cost question
The system requires dedicated camera installations at both ends of every ground that uses it. The installation cost per venue is approximately USD 180,000, with ongoing operating costs of approximately USD 25,000 per match. The cost is meaningful but recoverable across the broadcast carriage value of ICC events. The cost is the procedural blocker for smaller associate boards.
The deployment roadmap
The ICC cricket committee's preferred deployment roadmap has three phases. Phase one: ICC events from late 2026 (T20 World Cup, ODI World Cup). Phase two: bilateral series at Full Member venues from 2027. Phase three: associate venues from 2028, contingent on funding. The phased approach lets the cost spread across multiple cycles.
The procedural concerns
Three procedural concerns are being flagged by the ICC umpires' association. One, the AI flag's confidence rating must be transparent to the on-field umpires. Two, the system's training data must include Tier-2 conditions, not just Tier-1 conditions. Three, the system's decisions must be auditable after the match for accountability. All three concerns are being addressed.
The fan-experience question
For viewers, the AI system means faster no-ball calls and fewer dismissed-off-a-no-ball reversals. The procedural shift is mostly invisible. The bigger question is whether the AI system will eventually move from parallel input to primary decision-maker. The cricket committee's position is that the parallel-input model will remain for at least the next two cycles.
What to watch next: whether the ICC AGM signs off on the phase-one deployment of the AI system at the T20 World Cup 2026, because that is the moment the system moves from pilot to live tournament use and the false-positive rate gets tested at scale.
Related coverage
Share this article
Sanjana Patel
Expert in: InternationalCricket analyst and content writer at CricJosh, covering International with 42 articles published.
Related Articles

4 min read · 21 May 2026

4 min read · 21 May 2026


5 min read · 21 May 2026