Over the last few months, I noticed that I've become really obsessed with finishing watching things before midnight local time to avoid messing up my dashboard graph with stats that appear on the "wrong day".

It occurred to me tonight, after yet another movie finished just after midnight, that actually, software can solve this problem.

I propose that Trakt take the timestamp and duration of each item played into account when generating graphs, such that if (as happened tonight) I finish a 126-minute movie at 0:03, 123 of those minutes still appear in the graph for the previous day, and 3 of them appear on the same date as the play.

One could nitpick this and say the split should be 122 and 4, for example; I've no preference either way there. I can also see potential objections with regard to the genre & play type counts that appear on the graph: Does "1 movie" have to become "1 movie" + "1 movie", for example, or does it become "0.95 movie" and "0.05 movie"?

Whatever the answers to those (and other) questions, I think this sort of commitment to statistical accuracy can only help in the long run.