normhead's situation, IMO, qualifies as "sunrise problem":
Rule of succession - Wikipedia, the free encyclopedia
we can look at it from 2 perspectives: probability that the next shot will be successful, or probability that the next shot will be failure. I believe, we're interested in the latter.
The formula is: P = (S+1)/(N+2). S is success, N is total experiments. Since we're looking at the problem from the "other side", i.e., treating mirror flop as a "success", the numbers are like these:
S = 1 (out of all experiments, normhead managed to bring the camera into nirvana once), and N = 12400, total amount of experiments.
So, the probability, that Norm's K-3 will flop during the next activation is (1+1)/(12400+2) = 0.00016, less than 0.02%
for DRabbit, the situation is significantly worse. She's treating a set of activations as a single shot. here we have couple of possibilities: either treat each time lapse as a single shot - which is not exactly fair
, since the probability of failure seems to be a function of time (or activations) - the longer you shoot, the larger chance of running into the issue, or we should take the mean amount of shots per session into account and apply the following rule: if actual number of activations per time lapse is less than mean, it still counts as one successful time lapse; for the rest: round to nearest. in ##: if mean amount of shots per time lapse is 500, a time lapse with 400 counts as one, a time lapse with 700 counts as one, a time lapse with 850 counts as 2.
Regardless of methods, it is possible (to a degree) to predict the probability of a successful time lapse, but matters get complicated, if we take into account the hypothesis that the likelihood of mirror flop rises exponentially with longer series of rapid successive shutter activations. statisticians - welcome to participate in the discussion