ยง Accuracy vs precision
I had a hard time remembering which is which, so here's how I do it now.
First, I think of it from a probabilistic lens, where one of them is the
mean, and the other is variance of a gaussian distribution as shown above.
We don't yet know whether accuracy is the mean or the variance.
Next, recall that it's linguistically correct to say:
you're precisely wrong
but not
you're accurately wrong.
Thus, we can be precise about something wrong. That is, we can
be very "precise" about "hitting the wrong target". So, precision ought
not care about the true value , just about how well we hit something .
This is exactly what the variance attempts to capture: how "spread out"
we are, or "how well we hit the mean".
The accuracy itself is the distance between the mean of our distribution and
the true reference value we want to hit.