One number instead of a checklist
Strength meters that award points for capitals and symbols are guessing. Entropy is not. It answers a single question: how many equally likely secrets could this have been? Written in bits, an entropy of n means 2n possibilities, and an attacker who knows everything about how you chose the password except the choice itself must, on average, try half of them.
Two consequences follow immediately. Every extra bit doubles the attacker’s work. And entropy depends on the process that produced the password, not on how the result looks.
The formula
If each of L positions is filled independently and uniformly from an alphabet of N symbols:
possibilities = N^L
entropy = log₂(N^L) = L × log₂(N)
The per-symbol contribution, log₂(N), is what each alphabet gives you:
| Alphabet | N | Bits per symbol |
|---|---|---|
| Digits | 10 | 3.32 |
| Lowercase letters | 26 | 4.70 |
| Letters, both cases | 52 | 5.70 |
| Letters and digits | 62 | 5.95 |
| All printable ASCII | 94 | 6.55 |
| Diceware word | 7,776 | 12.92 |
Notice how little the symbol set matters: going from letters-and-digits to the full keyboard adds 0.6 bits per character. Adding one character adds about 6 bits. That single observation explains most modern password advice.
Worked examples
- 6-digit PIN: 6 × 3.32 = 19.9 bits, about a million possibilities. Fine behind a lockout; useless offline.
- 8 characters, letters + digits: 8 × 5.95 = 47.6 bits. A GPU rig against a fast hash finishes in hours.
- 12 characters, full keyboard: 12 × 6.55 = 78.7 bits. Centuries against the same rig.
- 5 Diceware words: 5 × 12.92 = 64.6 bits. Decades, and you can remember it.
- 20 characters, full keyboard: 131 bits. Beyond any attack, now or foreseeable. This is the default on our generator.
The entropy calculator fills the formula in for any shape you describe and converts the result to crack times for three attacker speeds.
The catch: humans are not uniform
The formula assumes every position was filled at random. When a person chooses a password, positions are anything but independent: the first letter is capitalised, vowels follow consonants, the digits are a year, the symbol is ! at the end. Charlotte2019! has the shape of a 14-character, 94-symbol password, 92 bits by the formula, and real entropy of maybe 25 bits, because it is a name plus a year plus the most popular symbol.
This is why the strength checker exists alongside the calculator. It searches for the patterns a cracking tool would exploit and charges each one at its real cost: a dictionary word costs log₂ of its rank in the wordlist, a year costs 7 bits, a keyboard walk a handful. What remains is charged at the brute-force rate. The result is a much lower, much more honest number.
If you chose it, assume it has less than 30 bits of entropy regardless of length. If a machine chose it uniformly at random, trust the formula.
How many bits do you need?
Enough that the attacker’s cheapest attack takes longer than the secret is worth. Three scenarios cover almost everything:
- Online guessing through a login form, rate-limited to perhaps 100 attempts per second: 40 bits means an average of 170 years. Any decent password clears this.
- Offline attack against a slow hash such as bcrypt at 10,000 guesses per second (the site leaked its database but stored passwords properly): 60 bits is 1,800 years.
- Offline attack against a fast hash such as SHA-256 or NTLM at 10¹⁰ to 10¹¹ guesses per second (the site stored passwords badly): 80 bits is about 2,000 years at the faster rate.
Because you rarely know how a site stores passwords, aim for 80 bits on anything that matters. A 13-character random password or a seven-word passphrase gets you there. Past 128 bits you are in cryptographic-key territory and further length is theatre.
Entropy is not the same as crack time
Bits describe the size of the haystack. Crack time also depends on how fast the attacker can check each straw, which is set by the hash function the site used and the hardware the attacker rents. The crack time calculator lets you vary both and see the same password fall in seconds or stand for millennia. The guide on how password cracking works explains where those speeds come from.