Hacker News Reader

Comment by bawolff | Hacker News Reader

Its still a fallacy because some people are just lucky instead of being skilled, and its entirely possible to confuse the two.

In the same way, an appeal to an authority is better than nothing, but its still a really weak argument. You can misunderstand authorities and authority can be mistaken or suborned. Its a reasonable argument if nobody knows much about the topic, but its never convincing if people are learned in the topic; the benefit of an expert is not that they are experts, its that they know how to defend their position with reasoning. As the motto of the royal society says "Nullius in verba".

View on HN

Survivorship bias is more about believing a task is safe/easy because you only hear from the survivors/winners. Not about saying someone is better at a job because they haven't died/lost yet.

Luck eventually runs out. That's not to say you're wrong, only that I expect the distribution of the lucky amongst the skilled tends towards zero over sufficient time.

"Beware old men in a young man's game."

Depends on how often you role the dice not just how bad the dice are. Someone that’s got a 1% chance of dying every fireworks show is extraordinarily bad at their job. But if they only put on 2 shows a year they have better than even odds of surviving a 40 year career.

The odds there are surprisingly high, but not better than even. 44.75%

Actually the OP is correct - the odds of dying are 44.75%, while the odds of surviving are the corollary of 55.25%, i.e. greater than even.

Edit: Nope I'm wrong, just redid the math. Apologies.

But the important odds in this situation is not the individual, but the group - if you start with a group of 100 (or whatever) people extraordinarily bad at their job, what is the probability at least 1 makes it 40 years.

Survivorship bias is a bias because the sample is biased - Its when you pick a group of people who have already won and then are surprised that they beat the odds and won because you confuse the probability that a specific person wins with the probability that at least 1 person wins.

Actually, the important odds are: given that you've met a 75-year old explosives professional, how likely is it that he was skilful instead of lucky?

To get a good estimate of this you need to come up with some priors about how many skilful and unskillful equivalents there were at the age of 30 and how much worse the unskillful ones are at staying alive.

It was a survivorship bias point but I didn’t make that clear.

I meant if you meet X people at a random point in their career then on average more than half will make it to the end because some of the original sample died before you sampled the population. Basically older people would have a lower chance of their luck running out even if they where all equally incompetent.

At least assuming the original population isn’t dying of something else vastly more often etc.

The real question is what amount of time is sufficient. Its probably longer than you expect.

Reply on news.ycombinator.com