OK, I suppose I 'asked for that' by mentioning bernard and myself! However, as is so often the case in relation to statistical information, in the absence of any other information about all those other factors affecting the risk to individuals, population probabilities ('averages') are all one can talk about.You cannot in any meaningful way say what your, my, or Bernard's chances are of that. We all use different roads, types of roads, in different ways, at different times, and for different amounts of time. Population outcomes cannot be used to infer individual ones.To put that into context, you (and I) have about a 1 in 2,700 chance of being killed or seriously injured on the UK roads in the next 12 months (killed about 1 in 37,000; seriously injured about 1 in 3,000)
If it makes you more comfortable, think about absolute figures, rather than ('on average') probabilities. If the population risk of being killed or seriously injured on the UK roads in a year is about 1 in 2,700, that means that about 23,700 people would be killed or seriously injured in a year - although, as you say, without a lot more information one could not say how likely it is that you, bernard or I would be one of those 23,700. Similarly, if (as I was hypothesising) there were a 1 in 1,000,000 chance of harm coming to persons or property as a result of an "unsafe" electrical installation, then about 64 people/properties would suffer that fate, even though we can't know which 64 individuals/ properties they will be. However, in terms of the population, one can legitimately compare those 64 with the 23,700 .
Your comments obviously apply as much to the "unsafe electrical installation" as to injuries/deaths on the roads. Just as an individual who never leaves their house would have a zero probability of being injured/killed on the roads, so would a person (like my late grandmother, who wouldn't even touch a light switch!) who would never dream of touching a light fitting have a zero probability of coming to any harm as a result of it being metal and unearthed.
The more data one has, the more that one can look at 'averages' in subgroups, but any statistical/probabilistic information inevitably has to deal with 'on average' - whether within those 'averages' relate to whole populations or subgroups thereof. Such is the nature of anything probabilistic. If we knew for certain whether or not you, I or bernard were going to be injured on the roads in the next 12 months, there would be nothing 'probabilistic' to talk about - only 'known facts'.
The same concepts apply to things we often discuss here. We may 'know' from available surveys that 'in the population' ('on average') 7% (or is in 1 in 7 - I can never remember?!) of in-service RCDs are 'faulty'. However, without knowing a lot more about the make/model of the device, environmental conditions, loads to which it is subjected, duration of period in service etc. etc., we cannot take that to mean that this represents the probability of any particular individual device being faulty. However, that does not mean that population statistics are not useful.
Kind Regards, John