Originally Posted by
Biomega
But you can't predict the future. You play for what is most likely to yield the best DPS, because most of the time it will. You say you want to reduce variance, but you have no way of knowing whether you hit the RNG well or poorly - all you can go by is expected average, and all you are doing by playing incorrectly is reducing that average. I know statistics can sometimes run counter to what people intuitively feel should be the case, but math is math.
You're right, though, that very short fights are an exception for several reasons, and may require entirely different approaches in the first place. An extreme example would be open-world questing where you're fighting one mob at a time, and something like MFD into BtE into (Deadshot/Opportunity) Pistol Shot kills things outright with no need to use RtB at all. Depending on how far apart the mobs are, even rolling it between mobs might not be optimal due to the long downtime.
In raids, it's fight-specific and can be very difficult to accurately math out due to the many factors involved. If there's a priority add, for example, damage on that add is one concern - but continuing damage on the boss afterward doesn't disappear completely. If it's a short-lived high-priority add that you aren't cleaving (and then why aren't you playing Sub), chances are that pooling energy and holding CDs like BR/MFD is going to be much more important than RtB for that short window. But if the add lives longer, or if you're cleaving off the boss, then sticking to the correct RtB logic is most likely the optimal way to go.