Well, it depends on your darkest dark and brightest bright you want to capture without hitting the sensor's and signal processor's limit.
In general you will have to answer this: why do you do exposure bracketing? E.g. a Nikon 5100 has an EV range of 13.7 by itself, and also uses Active lighting, emphasizing details in shadows and bright areas, so why bother with exposure bracketing?
The answer is:
- you want to capture details in highlight areas and
- to capture details in shadow areas with acceptable noise.
Basically the highest and lowest EV is determined by these two considerations. You have to make shots until you are satisfied with the highlight details, and make shots until you have enough details in shadow without unacceptable noise.
This way you get the highest and lowest EV you need. If your camera has a decent EV (dynamic performance) you could as well live with two shots instead of three or more. E.g. with a typical EV performance of 10, two shots can cover an enormous dynamic range (you take them at +5 at -5 - a hardly improbable scenario :-) ).
So all the other shots are somewhat redundant, except for one thing: the sensor and the signal processing subsystem somewhat prefer light (obviously), and you will see a nonlinear response in the incoming light-to-pixel value mapping. This is partly physics, partly electronics, partly signal processing. So what you want to know is: where is the sweet spot of your camera, where a little change in incoming light is reflected by a large change in pixel luminousity/etc. value change. There, your sensor is the most sensitive. More important is: the range of that sweet spot.
Now, let's say you know that your camera has a 5 EV range of very nice detail capture, and the highlights and darks capture somewhat coarse. (You do not have to guess this, usually you can dig up documentation about this or measure it for yourself).
Also, you have established (as I briefly described previously) that for a specific scene you need 10 EV.
And let's say your camera has a native dynamic range of 11 EV (meaning the total, not the sweet spot dynamic range).
So if you make a single shot, in the photo, a range of 5 EV will have great details, the rest is not so nice, in the darks and highlights.
What you should do then is: the current example scene needs 10 EV, you divide that with your 5 EV camera sweet-spot range, and so that means: you make two shots, at +2.5 and -2.5 EV to the middle point (with the assumption that the manufacturer put the middle sensitivity point to the middle of the sweet spot of the sensor). Now, that is kind of tricky, to not have any overlap, so the usual way to go about this is: do +2.5, 0, -2.5. And that's all.
You could make 11 shots for this, but you are just collecting redundant data mostly...
The only other consideration you can make is that in low light condition, the longer you are open, the more thermal noise you will pick up, so making shorter shots there as well could be useful to remove noise from the final image except from the darkest parts.