Earlier quoted context omitted.
Get 8 billion people to flip a coin 30 times in a row, and one is likely to get heads 30 times in a row. It means nothing.
He's been flipping the same coin for 20 years and it has never come up tails. That's quite the coin. There is no data set more relevant to a person than their own experience. Meanwhile I don't have enough fingers and toes to count up all the people I've known personally who have been killed or crippled in auto collisions.
That also means nothing, because you said nothing about how often the coin is flipped.
The first papers that show up in google say "Mountain biking athletes were found to have an overall injury risk rate of 0.6% per year and 1 injury per 1000 h of biking." and "75% of the injuries were minor, such as skin wounds and simple contusions"
4000 hours per serious injury would mean 2000 hours per coin flip. A hypothetical person that bikes twice a month, 4 hours each time, for 20 years with these accident rates, would have flipped the coin once.
But if the accident rates were 10x higher, you could easily still get the same results. One person is just not enough data here. Let's use the real data that show it's pretty safe.