I don't remember the exact formula, someone else might know where to find it.
But, in essence, 1000pts is assigned to what would be the perfectly average result on that course. The formula uses the mean and standard deviations to calculate everyone's scores based on this.
Stuart Scott is the man who implemented it.
I believe it's the same (or at least very similar) as the Lakes-5-day system:
Runner's score = 1000 + 200 x (MT-RT)/ST
MT is the mean of the top 90% of each course (excluding n/c, disq, dnf, etc)
RT is the runner's time
ST is Standard Deviation
(Have heard different "facts" about whether the 90% is used for the mean or if it's just all finished competitors.)
In that case what is the key to getting points as high as possible?
According to that system, coming last might garner you more.
basically you want a load of scouts who are shit with a map to all run around together, with separate Si.
So, you get like 30 people finish very close together. Then you have a very small standard deviation, and the winner will be miles ahead points-wise.