5 out of 14 samples were rejected for technical problem (that is the CRAM wasn't working) not 6 out of 15. Not much of a difference as 9 samples were valid anyway.
But according to their own "presented paper":
Accuracy - the probability of the CRAM reporting the true state of the driver in any 15 sec window - ranged from 0.69 to 0,72 with a mean of 0,71.(SD 0,008)
You said 100% accuracy, I said 70%. Who's the one not getting it? Sigh ...
Market Data 
Markets 
AI
