But the key question is whether the lead time needed to randomize the list cannot be longer than the lead time to go through it sequentially.
Perhaps what you really need to do is not really randomizing it, but rather reading it in some order, except from front to back. For example, if there are n elements in the list, perhaps you should handle 1, n, 2, n-1, 3, n-2, etc. Or 1, n / 2, 2, n / 2 + 1, 3, n / 2 +2, etc.
If you always have the same number of points or the number of points falls into a small finite set, you can make a point estimation order for each possible number of points once in advance, either truly random, or perhaps carefully calculated in some way.
Jay
source share