the expert who felt rightly that it was no real assistance. Mathematical computation of the probability of cribs was a system which could not be ignored but results needed to be modified and analysed by the judgement of experience. Nothing really could replace the knowledge which was gained by experience.
A Cribster had to come to realize that a crib was not right because he had thought of it himself, that he should not be discouraged because a series of apparently good cribs had failed and start making wild assumptions about new wheels and new keys. He had constantly to decide between two or more cribs as to which was the best and in doing so had to rely as much on a very wide experience as on written records. He had to know which risks could be wisely, which unwisely, taken and had constantly to make decisions on the policy to adopt in breaking a key: would it be better to run 1 crib with a 60% chance of coming out and, if this failed, change to another message, or start on a programme of a 3 form crib on one message which would finish by giving a 95% chance of success? Problems like this would, of course, have been easy had it not been necessary to consider such other factors as the bombe time available, the intelligence and cryptographic advantages of a quick break, the probable pressure on bombes in 12 hours time and so on. On the face of it, straight cribbing appears to he impossibly tedious when compared with depth but, though depth had its great moments which straight cribbing could not touch, the problem did not become less interesting when Banburismus had died. With perhaps 6 to 10 keys to break regularly, the cribster was a busy man faced with an interesting problem in tackling which he had to consider not only the total number of keys eventually broken but also economy of bombe time and the demands of Intelligence.
Reencodements are repetitions in a cypher of messages which have already been transmitted on other cyphers, or indeed in plain language. The great advantage of reencodements over straight cribs is the factor they receive in favour of their being right owing to the length of crib which has been written in without its crashing against the cypher text. We have already seen that for each letter of a wrong crib there is a 1 in 25 chance of a crash so that a 50 letter crib which does not crash gets a factor of 7 in its favour, a 100 letter crib a factor of 50, a 200 letter crib a factor of 2500, a 300 letter crib a factor of 40,000 -- in fact there is really no chance of a long reencodement being wrong. We have met reencodements in a variety of fairly distinct forms which I will deal with separately.