3 ms·
> if that’s the difference between catching a connection in SF or not then that’s a 20 minute savings it could be Here's a heuristic you should try applying: u
by repsilat 8y ago
> if that’s the difference between catching a connection in SF or not then that’s a 20 minute savings it could be
Here's a heuristic you should try applying: under some reasonable assumptions, getting to the platform X seconds earlier should shave X seconds off your commute in expectation and not affect the variance. (Those assumptions are basically "you're not trying to catch a particular train.")
Of course, you may have a nonlinear objective function in terms of time/delay, or you may be optimising for regret instead of arrival time, and in those cases the analysis may turn out differently.
- DominikPeters 8y agoA related puzzle: Should you recharge your transport card before or after riding? https://trefethen.net/2018/03/25/metro-mean-and-variance/ https://trefethen.net/2018/03/25/metro-mean-and-variance/
- scarejunba 8y agoThe right answer (in practice, though I admit the question is fun) is to "use your debit/credit card" (on the Tube, for instance) for the train or "set up auto-load if your transit system doesn't accept credit/debit cards" (on BART, for instance).
- scarejunba 8y agoThat's old-school thinking from a time where you only have a wide probability distribution. Modern commuters have BART timings on an app. The distribution is pretty tight: the train is arriving between 0 and 90 s from now and you're at the top of the stairs. Naturally, everyone's worked out that heuristic while waiting for the college bus. It's the problem that's changed.