[Zickzack]

Flavius Josephus writes in the History of the Jewish War against the Romans - an important source for the Roman army - that the first cohort of every legion was twice the size of other cohorts, because it included the veterans, i.e. those who had served 16 years at least.

The following calculation is either simplistic or robust, depending on your view. And I am aware that there are simplifying assumptions, like the one that the rate of attrition did not depend on the service age of the soldier. There are reasons for that even. Younger soldiers might have fallen prey to infections because they encountered them the first time. Older soldiers were possibly tougher in that respect. Young soldiers took two years to learn to fight like a legionary. Old soldiers may have balanced their ageing bodies (a decade or more of gruelling service) with their experience. And yes, not all veterans would be incorporated into the first cohort, because they had higher echelon jobs in their respective cohort. But then, the good ones (long lasting ones?) would have been members of the first cohort in the first place because it was supposed to be elite. So bear with me.

Josephus gives us that a Roman legion could be divided in eleven roughly even sized parts, with the veterans forming one of these eleven parts. Assuming a constant rate of attrition (see above), this means that about 8% of a legion were lost each year for a variety of reasons other than retirement, 92% made it. If we assume twenty years of service, 0.92^20 gives ~ 0.188, so roughly one fifth of the legionaries made it to retirement.

If the same attrition rate applied to auxiliaries - twenty five years of service -, then about one in eight auxiliaries made it. And finally, the Pretorian guard, where the assumption of the same rate may be a stretch: sixteen years of service mean that one in four enjoyed the fruits of his post-military life.