Some Information: The number of values 16 bytes can hold is 2^128.
Total = number of total values that can be represented
Actual = number of values that have already been used
The probability of inserting an ID into a table that already exists could be formulated as the following:
1 - (Total - Actual) / Total
I've done some preliminary calculations and I've come up with the following:
Say you had a table that had 10 million records already in the table, then the probability would be
1 - (2 ^ 128 - 10,000,000) / 2 ^ 128
which is appromixately
0.00000000000000000000000000000003
I think now I know why it's said that:
"While each generated GUID is not guaranteed to be unique, the total number of unique keys is so large that the probability of the same number being generated twice is infinitesimally small"
from the wiki