So to quote it's inventor Paul Hawkins in the above 2015 article - "2.2 mm is our tennis error. We have stated 5mm accuracy, and 10mm is some scenarios" (regarding cricket )
Those 'secenarios' I gather to mean where the ball has bounced less than 40 cm before impact. Anything above that and hawkeye has enough frames to accurately establish the current path, and thus the predicted path. So worst case scenario is still more than 3 X half a ball width.
Do they really accept that it is 100% accurate? Surely nobody would believe that any tracking technology is 100% accurate? The tolerance and expected margin of variation is certainly open to debate but I'm pretty sure the developers of the technology would not claim 100% accuracy.
Agreed, I didn't mean that Hawkeye is 100% accurate for tennis, the point I was trying to make is that the tennis community seem to accept hawkeye's results as accurate and have no qualms overturning the umpire's call if hawkeye shows the ball missed the line by a couple mm. 'Umpires call' counts for nothing in tennis!