23Z Page 109
done in connection with a motor rectangle (R0, 45, 48). Effects of this were
(i) General method of setting motor (instead of by using strokes) and calculation of expected score in case of Χ2 limitation (R0, 47-49).
(ii) Tendency to use ΔP statistics for finding the best runs (e.g. R0, 103, 105). But when we had set enough messages we felt that ΔD letter counts of messages set by us would not be misleading for runs statistics (Dec. 1943).
Other lesson learnt at the time of the Heath Robinson setting were
(i) The importance of checks at every stage, including two makes, hand checks and exact numerical checks.
(ii) Ability of Wrens to compute and use simple formulae for set totals etc.
(iii) Possibility of good slides on ΔΧ's.
(iv) Value of having sprocket guide near lamp on a Robinson to minimise the effect of the sprockets of two tapes not matching exactly.
(v) Importance of careful labelling of all work and of pigeon holes for tapes (Tapes were originally hung up on hot water pipes.)
(vi) Value of using Bayes' theorem rather than orthodox statistical outlook.
(vii) 'Dottery' method for setting Ψ's.
(viii) Use of Δ wheel tape to save plugging.
When the first 'Robinson' arrived we had not yet adopted the method of checking the result of every run before accepting it. This was done by using various standard lengths of tape. For example the 1+2 run was usually done with tape length of 3814, even if this meant putting in a thousand blanks. The object was first to make the Robinson 'readings' equal to the settings, and second so that the tape could have one letter removed and then be used for checking the result of the run. This idea of using a message tape of length a multiple of a wheel length was first devised for Heath Robinson and enabled Robinson de-chis to be done. The method really came into its own with the double Robinsons which could take four tapes. It then became possible to set all five Χ's without ever making a de-chi tape. For example, when setting Χ3 given the setting of Χ's 1, 2, 4, 5, a message tape of length 3813 was used with a ΔΧ1,2 tape and a special ΔΧ4,5 tape of length 3813 (and therefore non-periodic). It was necessary to have four distinct types of ΔΧ4,5 tape (R1, 58).
In the Robinson period a large variety of new setting runs were discovered and routines were improved to a point at which not many mistakes were made. However when Colossus 1 arrived it was found that it could cope with more material than all the three Robinsons then operating. This applied to wheel-
Back to General Report on Tunny. Contents.