Fixed-point metrics in millionths, identical on every target.
Mean squared error of (confidence in micro, was it right) pairs, in micro.
A CONFIRMED verdict naming anything but the true cause; any CONFIRMED on a control.
Mean reciprocal rank; an absent cause counts as zero.
1-based rank of the true cause, null when absent or when the scenario is a control.
Wilson score interval at 95 percent, as millionths.