Each key word has a specified length and a specified number, *N*,
of ELSs that will be generated at random.
The beginning position of each ELS will be chosen at random from text position 0 to the last
text position minus the number of characters of the key word. Then the skip of the ELS will be
chosen at random subject to the constraint that the ELSs entirely fit within the text length.
When the ELSs for both key words are generated with this policy, the resulting ELS set will,
by construction, satisfy the Null hypothesis of no Torah code effect.

To generate ELSs that satisfy the Alternative hypothesis, ELSs for each key word are first generated
to satisfy the Null hypothesis and then a fraction *f* are moved so that they are in a more
compact meeting consistent with the Alternative hypothesis. The ELSs of the second set of ELSs are
the ones moved. Which ones to be moved are chosen at random.

Each ELS of the second set of ELSs determines its best meeting with the ELSs of the first set. How?
It goes through each ELS of the first set and determines the set of resonant cylinder sizes.
If there are no resonant cylinder sizes, the next ELS from the first set is examined. If there are
no resonant cylinder sizes for all of the ELSs of the first set, the ELS from the second set
is not moved. If there is at least one resonant cylinder size, then the maximum cylinder distance between
the ELS of the first set and the ELS from the second set is computed. For one of the ELSs of the first set, this
maximum cylinder distance is smallest. That ELS from the first set is the best ELS for the ELS of the second set.
Suppose that this maximum distance is *d* between the positions
of the ELS from the second set to its best ELS of the first set. The ELS from the second set is then re-positioned so that maximum distance between its positions and the positions
of the first ELSs becomes *ad* for the given fraction *a*. This is done for each of the
ELSs of the second set chosen to be moved.