Doctors in Finland had done a study from nearly 16k same sex twins.
All participants were healthy when it began in 1975.
Bare in mind these were twins. That's about as good as controlling the genetic variable as one can get.
This study was conducted over a 20 yr time period
More the 30 min exercise, at least 6 times a month were classified as conditioning exercises (brisk walking)
Those that exercised less were considered occasional exercisers
Those that did nothing were considered sedentary
Within that 20 yrs, 1,253 died. Even considering other risk factors, the conclusion was that exercise was proved strongly protective, reducing death rate of conditioning exercises by 43% and occasional exercisers by 29%
Twins who exercised regularly were 56% less likely to die during the periods then their sedentary siblings. Those that occasionally exercised had a 39% lower death rate then their sedentary twin.
Harvard had a health newsletter that covered the health effects of exercise, they reference that study as well as others.
They had also referenced a randomized clinical trial of cardiac rehabilitation to see if walking can help with those that already have heart disease. It goes own to say meta-analysis of 48 trials in 8,946 patients showed that moderate exercise -typically walking or riding a stationary bicycle for 30 minutes 3 times a week - produced a 26% reduction in the risk of death from heart disease and a 20% reduction in the overall death rate.
Now I don't know the specifics on that one, so I'm inclined to say anecdotal without knowing more about what went on with it, just an interesting read none the less.
Ok, let's do science. Your hypothesis is that exercise can extend the quality and perhaps the quantity of someone's life. Please describe your critical experiment.
If they didn't cove this in all of that hard science, perhaps a brief review: A critical experiment is one is capable of confirming an hypothesis and, at the same time, must be capable of disproving it. It must be predictive not descriptive. And, revisiting the notion of 'disproof', all it takes to disprove an hypothesis is one, as in one, counterexample. For example see the classic Michelson–Morley experiment in, I believe, 1897.
You forgot one important aspect. Repeatability. It must be able to be replicated. If it can't be replicated, even if it technically may be true, it fails on that alone.