General Report on Tunny
25A Page 154
| (1) Χ2 limitation phenomenon | |
| (2) Corruption represented by 9's | |
| (3) Random variations. |
These are the origins of the "R minus twice norm" controveray. (R4 pp 44, 54, 73, 86.)
Before considering these necessary refinements, the adaptation of 'excess of good over bad' to Colossus counting will be described. The outlines of this are desirable for understanding the sequel.
(b) Adaptation to Colossus counting.
To save Colossus time the excess (to revert to the earlier example) of /'s over T's is found from a single run, not by counting /'s and T's separately.
| ij | = | pippage of evidence that ΔΧi5, the ith character of ΔΧ5 is a dot | |||
| = | /'s against ΔΧi5 | - T's against ΔΧi5 |  |
||
| = | /'s against ΔΧi5 | + T's against other characters | ΔΧ5 | ||
| - T's against other characters | supposed | ||||
| - T's against ΔΧi5 | all dots | ||||
| = | (/'s against all characters, if ΔΧi5 is a dot, all other characters crosses) | ||||
| - (/'s against all characters, if all are crosses) |  |
||||
| = | (T's against all characters, if ΔΧi5 is a cross, all other characters dots) | ||||
| - (T's against all characters, if all are dots) |  |
||||
In either or the first term can be found on Colossus in one run, stepping chi 5; the second term (known as the NORM) can be found on Colossus as a single count without stepping.
The two description , are equivalent. Because it seems more natural to run for good letters, the theory, including the naming of runs, is in terms of the former.
Because it is better and easier to have strings of dots on Colossus than to have strings of crosses, the actual Colossus runs are those of the latter description.
This is said "Wheelbreaking runs are always run inside out on the impulse being run for".
Back to General Report on Tunny. Contents.