How Aircrew Survival Odds Were Calculated and Misread
A Bomber Command tour was thirty operations. That number was chosen in 1942 as a humane limit, a point at which a crew could be rested and moved to training duties, and it was explained to new arrivals as a finish line. What nobody explained clearly was the arithmetic that stood between an aircrew and that line. Over Germany in 1943 the loss rate per sortie ran at roughly four per cent, a figure that sounds survivable in isolation and is not survivable thirty times in succession. Aircrew survival odds were the product of that repetition, and the product is a different quantity altogether from the number printed in the briefing.
Aircrew survival odds are not intuitive to anyone. That is the whole difficulty, and it was not solved by publishing better figures.
Counting the Odds Over Germany
Take four per cent as the chance of not returning from a single operation. The chance of returning is ninety-six per cent, and the chance of returning thirty times running is that figure raised to the thirtieth power. It comes out near twenty-nine per cent. A crew starting a first tour in 1943 therefore had something close to a seven in ten chance of being killed, wounded or taken prisoner before the tour ended, and the men flying those aircraft understood the outcome long before they could express it as a calculation.
The totals bear it out. Of roughly 125,000 men who served in Bomber Command, 55,573 were killed, a death rate approaching forty-five per cent that has no equal among British formations of the war. The Eighth Air Force set its tour at twenty-five missions and met loss rates near five per cent during the unescorted daylight raids of 1943, which produced aircrew survival odds of about the same order. Records of individual squadron losses are held in the collections described by the RAF Museum research department, and the pattern in them is monotonous rather than dramatic. Crews vanished at a steady rate, night after night, until the rate itself changed.
Where the Arithmetic Goes Wrong
Two errors recur whenever aircrew survival odds are discussed. The first is treating a repeated small risk as if it stayed small, which is how a four per cent loss rate per sortie becomes, in conversation, a four per cent problem rather than a seventy per cent one. The second is reasoning from the aircraft that came home. Abraham Wald, working for the Statistical Research Group at Columbia in 1943, was asked where to add armour on the basis of damage maps drawn from returning bombers. He pointed out that the maps recorded where an aircraft could be hit and still fly, and that armour belonged in the places the maps left blank. That inversion is now taught as the standard example of survivorship bias, and the discipline that produced it, operational research, went on to shape convoy routing, depth-charge settings and search patterns for the rest of the war. Survivorship bias is cheap to name afterwards and very hard to notice from inside a problem, which is why the damage maps looked convincing to competent people for as long as they did.
The same failure turns up wherever people are handed compound conditions in writing. Consumer terms are the clearest everyday case, because the numbers there are printed in full and misread anyway: a Chilean guide to Fuego sets out the rollover multiple, the deadline and the withdrawal ceiling as three separate figures, and it is the three combined, not the headline offer, that decides what a player actually collects. Nobody is concealing the arithmetic. It simply stops feeling like arithmetic once it arrives as three ordinary sentences. The instinct that misreads a bonus is the instinct that misread a tour of thirty, and it is a good deal older than either.
Reading Numbers Written About You
What changed in the air war was not the arithmetic but the inputs. Escort fighters, better navigation aids and improved tactics pushed the loss rate per sortie down, and every fractional reduction compounded across a tour in the same way the losses had. By 1944 the same thirty operations carried aircrew survival odds that would have seemed fantastical two years earlier. This is the practical lesson of operational research: attack the per-event number, because the exponent will do the rest of the work in whichever direction you send it.
Vietnam-era carrier aviation inherited both the mathematics and the habit of underrating it. Fixed-wing losses over the most heavily defended areas of North Vietnam ran at roughly two aircraft per thousand sorties in 1966 and 1967, a figure that reads as negligible and is not. A pilot flying a hundred missions into that region faced a cumulative risk nearer one in six than one in five hundred, which is why tour length mattered as much as any single mission profile. The comparison between the F-4 Phantom and the MiG-21 is usually told through turn rates and missile performance, but exposure counted for at least as much. The analysts who first made that argument were working in the tradition founded during the Second World War and carried forward by organisations such as CNA, whose wartime predecessor did the original submarine-warfare work.
Aircrew survival odds were never a secret. They were published, briefed and, in Bomber Command, discussed openly enough that crews made their own estimates and were usually close. What was missing was not information but a way of holding the compounding in mind while climbing into an aircraft for the eleventh time. The men who flew with the American Volunteer Group in China made the same calculation with worse data and reached the same conclusion. Arithmetic of this kind is easy to perform and very difficult to believe, and that gap, rather than any shortage of numbers, is what the whole subject is about.