Hacker News Reader

The issue isn't whether the sample is small. The issue is whether it is random.

Yes. If 1000 people toss a coin 10 times, only the ones with exceptional results are likely to blog about it.

So if we had 4 purely random macbooks, and 3 of them failed, then that would be a meaningful indication that the failure rate is higher than 10%?

The issue is both small sample and lack of randomness.

Yes, it would be a meaningful indication. I'll grant the lack of randomness in the sample, but the math is correct.