You should see the games they play with statistics.
Find some 1 in 1000 match. Find another 1 in 1000 match. Claim they're independent, so the chances of both of these are 1 in a million.
What they don't say is that there are 5000 things you can test that each have a 1 in 1000 chance, so you should expect to find ~5 at random if you test them all. The 1 in a million chance is if you choose 2 of the 5000 at random, test only those and they both match; not if you systematically run tests that only notify you when they find one of the matches.
Find some 1 in 1000 match. Find another 1 in 1000 match. Claim they're independent, so the chances of both of these are 1 in a million.
What they don't say is that there are 5000 things you can test that each have a 1 in 1000 chance, so you should expect to find ~5 at random if you test them all. The 1 in a million chance is if you choose 2 of the 5000 at random, test only those and they both match; not if you systematically run tests that only notify you when they find one of the matches.