How Long Would It Take to Crack My Password?

It depends almost entirely on something you do not control: how the website you used it on stored it. The same password can fall in an afternoon or hold for centuries depending on that one decision, without a single character changing. The charts promising you three billion years are not exactly wrong, but every one of them rests on assumptions it does not print.

The estimate is one division

Every crack-time figure you have ever seen is the same arithmetic: how many possible passwords there are, divided by how many an attacker can test each second. The first term is set by your password. The second is set by the attacker and by the site, and you have no visibility into either.

combinations  = alphabet size ^ length
average time  = combinations / 2 / guesses per second

The division by two is there because an attacker finds the answer halfway through the search space on average, not at the end. Charts that leave it out are quoting the worst case for the attacker, which flatters your password by a factor of two. That is by far the smallest distortion involved.

Guesses per second is not one number

This is where estimates diverge by many orders of magnitude, and it turns on how the other end stored what you typed.

• Against a live login form: a few guesses per second at best, because rate limiting and lockouts exist. Almost anything survives this.
• Against a stolen database hashed with a fast function such as MD5 or plain SHA-256: tens of billions per second on a rack of consumer graphics cards.
• Against a stolen database hashed with bcrypt, scrypt or Argon2: perhaps tens of thousands per second, because those functions are deliberately slow and memory-hungry.

The gap between the second and third cases is around a million to one. That is the difference between a password lasting an afternoon and lasting longer than you will need it, decided by a choice made by somebody else years ago. You usually find out which one it was at the same moment everybody else does.

Brute force is the attack nobody tries first

The charts model an attacker working through every combination in order. Real cracking software does not start there, because it does not need to. It starts with lists of passwords from previous breaches, then dictionary words, then those words with the decorations people reliably add: a capital at the front, a digit and an exclamation mark at the end, a zero for an O, a three for an E.

This is why crack-time charts flatter human-chosen passwords so badly. Something like "Tr0ub4dor&3" scores well on a chart that assumes ninety-five equally likely characters in every position, and falls quickly to a rule-based attack that has seen that exact pattern a million times. The chart measures the shape of the password. The attacker models how people think.

A crack-time estimate is only honest when the password was generated at random. If a person invented it, the figure is an upper bound, and usually a wild one.

A number that survives changing hardware

Instead of asking how many years a password buys, ask how many bits of entropy it has and treat the answer as a category. Bits carry no assumption about hardware, so they stay meaningful as hardware gets faster — which is exactly what every crack-time chart quietly fails to do.

• Below 50 bits: assume it falls. Any offline attack recovers this.
• Around 70 bits: survives casual attacks and a slow hash, but not much else.
• Around 100 bits: no realistic offline attack reaches this today.
• Above 128 bits: past the point where the arithmetic matters at all.

Twenty random characters drawn from the full keyboard is roughly 131 bits. There is no meaningful security difference between that and a thousand-character password: both are already outside what counting can reach, so anything that gets you will not be a machine enumerating combinations.

What the estimate cannot tell you

A crack time only describes an attacker guessing. It says nothing about how passwords are actually lost: reused across sites, typed into a convincing fake login page, read by malware at the keyboard, or reset by a support agent who was talked into it. A hundred-bit password entered on the wrong domain is gone in a second, and the chart still says three billion years.

So use a generated random password, a different one for every account, kept in a password manager, with a second factor wherever it is offered. Those habits defeat the attacks that actually succeed. Length defeats the one that mostly does not — worth doing, because it is free, but it is not the part protecting you.