random stops as about 1 in 450. If the P5 limitation is incorporated
(so that there could be no contraction where a motor cross were compelled
by X2 and P5 ) the number of random stops is reduced to about 1 in 720.
(These calculations have not been checked here - suggest this be done.)
It is true that the requirement of at least 3 repeats will cause a good.
number of correct positions to be missed. This involves no loss however,
because if a hand assumptipn yielded less than this number it would not
be regarded as good enough to go ahead with. For this procedure (as well
as for the primary function of the machine) the inclusion ot the P5
limiitation would be useful but not at all essential if it involves any.
constructional, difficulties. Hand elimination of stops rendered illegal
by the P5 limmitation is relatively simple and rapid. From my talk with
Mr. Welchman I conclude that the machine will be constructed so that this
limitation can be wired in required but that it will not be an integral
part of the mechanism.
7. Mr. Newman's section, with the aid of Colossus, is now able to
carry out by statistical or mechanical methods, all steps in the solution
except the initial recovery.of the Psi patterns.
8 The motor can, in theory at least, be derived statistically by
writing into a.61 x 37 rectangle all positions where there is a / in the
Delta D text. These probably result from motor dots. The technique used is
exactly the same as for the recovery of Mu37 and Mu61 from MT except that
.the entries in the rectangle represent probabilities rather than certain-
ties. Without the X2 limitation the method has been successfully used on
a depth of just over 2. The X2 limitation (and also the X2P5) makes the
necessary depth 4 tirnes as great because the procedure is based on
comparison of columns and elimination of half the data from each reduces
the available comparisons to 1/4. Mr. Newman thinks that with the limita-
tions it might be done with a depth of 6. This obviously limits the
applicability of the method. In any event Maj. Tester's methods seem
less laborious. Entries. in the rectangle need not necessarily be confined
to /; + is sometimes used (Delta89) and sometimes U (Delta+M). A
combination can also be used. The characteristics which serve as criteria
are determined by a frequency study of.the DeltaD text of the individuil
message being analyzed.
9. The setting.of a known motor pattern is quite another matter. This
is easily done after the.X patterns have been set, by setting upt Mu37 and
Mu61 on the Colossus machine and running through in all 2257 positions.
It. is not even necessary to make a DeltaD tape because the machine
operates with Delta Z tape and DeltaX patterns set up on the X wheels. The
machine records all /s in DeltaD which coincide with motor dots. It looks at 5
positions simultaneously on each of the 5 impulses and takes less than