---
title: 'Why you can''t trust a test that''s 99% accurate'
source: 'https://youtube.com/watch?v=2QSkkhbl3v4'
video_id: '2QSkkhbl3v4'
date: 2026-11-11
duration_sec: 89
channel: 'Veritasium'
---

# Why you can't trust a test that's 99% accurate

> Source: [Why you can't trust a test that's 99% accurate](https://youtube.com/watch?v=2QSkkhbl3v4)

## Summary

This video explores the counterintuitive nature of Bayes' theorem through a practical example: a 99% accurate test for a disease that affects 1 in 1,000 people. The hosts discuss why a positive result still yields a less than 10% chance of actually having the disease, highlighting the impact of base rates on diagnostic accuracy.

### Key Points

- **The Test Scenario** [00:02] — A disease affects 1 in 1,000 people, and there is a test that is 99% accurate with a 1% false positive rate. The question is: if you test positive, what is the probability you actually have the disease?
- **Initial Guesses** [00:18] — The hosts guess wildly, with one suggesting 99% and another using 'weird math' to arrive at 10%. The correct answer is actually less than 10%, just barely.
- **The Math Breakdown** [00:45] — Out of 1,000 people, 100 actually have the disease (1 in 1,000). The test correctly identifies 99 of them. Among the 99,900 who don't have the disease, 1% (999 people) get false positives. So, 1,098 people test positive, but only 99 actually have it.
- **The Surprising Result** [01:11] — 99 divided by 1,098 is about 9%, meaning even with a 99% accurate test, a positive result gives you less than a 10% chance of actually having the disease. This is a striking demonstration of how base rates affect probability.

### Conclusion

The video effectively illustrates Bayes' theorem, showing that even highly accurate tests can produce misleading results when the condition is rare. It emphasizes the importance of considering base rates in any diagnostic or probabilistic reasoning.

## Transcript

thousand people and there's a test for it that's 99% accurate and has a 1% &gt;&gt; You take the test and it comes back positive. disease? &gt;&gt; Um,
&gt;&gt; You're a doctor? &gt;&gt; Oh, so you better get this right. &gt;&gt; Oh, &gt;&gt; no. I think like 99 &gt;&gt; 99%. Yeah, &gt;&gt; but it's 50%. You either get it or not.
&gt;&gt; Yeah, &gt;&gt; my very weird math gives me a 10%. &gt;&gt; No, &gt;&gt; only 10%. &gt;&gt; The real answer is that it is less than 10%. Just barely.
affects one out of a thousand people. So, we know that 100 people actually accurate, right? So, it correctly identifies 99 out of the 100 people. There are 99,900 people who don't have
to be identified as having the disease due to the false positive rate. So that means 999 people plus the 99 who had correctly identified. So it's identified,98 people &gt;&gt; as having the disease. 99 of them
&gt;&gt; as having the disease. 99 of them actually have it. 99 divided by,098 is &gt;&gt; Yep. Yep. Yep. Yep. But how crazy is that? You never would have thought, you &gt;&gt; it's like a 99% accurate test, but only but it's a less than 10% chance that if
have the disease. &gt;&gt; Obviously, I promise I studied
