How many bits of entropy does a password need?
There is no single official cutoff, but the practical reference used by security tools marks below 28 bits as very weak and above 80 bits as comfortable against an offline attack for the foreseeable future; the length × log2(alphabet) formula lets you compute the exact value for any combination before deciding.
Why can a mandatory symbol requirement weaken a password?
Because the composition rule does not truly enlarge the alphabet used, it just forces 1 character from a category inside a length that stays short; the gain from log2(88) over log2(26) is only 1.76 bits per character, far smaller than the gain from simply adding more characters overall.
Does excluding ambiguous characters (0, O, 1, l, I) reduce entropy much?
The option removes 5 characters from an alphabet of up to 88, a small loss in relative terms: log2(83) works out to about 6.37 bits per character against 6.46 bits with the full alphabet, a difference under 0.1 bit per position that is worth it to avoid typos when copying the password by hand.
Can a high-entropy password still be easy to guess?
Yes, because the entropy = length × log2(alphabet) formula assumes truly independent characters; a long password built from a predictable pattern, like repeating a block or using a date, has much lower real entropy than the naive count suggests, even though it passes the theoretical calculation.