Doing some quick back of the envelope calculations. Consider this a thought experiment rather than anything too accurate. My server with a 2010 era quad core CPU can check about 300,000 keys per second. It could probably be pushed further with some tweaking. Let's say (conservatively) that a more modern quad core CPU can do 500,000 and use that as the reference. That means it can check 43.2 billion keys per day. Brute forcing the "correct horse battery staple" space One dictionary that includes a rank of how commonly a word appears on the web ranks the least common word "staple" at 16904. So let's use that hint (some mild cheating) and set our limits to the 20000 most common words. Total keys to check (20000 x 20000 x 20000 x 20000) = 160 000 000 000 000 000 And a server can check this many keys in a day:          43 200 000 000 So in this instance, we would need approximately 10,000 servers running for a year to brute force every combination of those 20000 words. Not practical, but certainly not impossible. But what if we use only the most common 1000? Total keys to check (1000 x 1000 x 1000 x 1000) =      1 000 000 000 000 And a server can check this many keys in a day:         43 200 000 000 In this case, we only need about 23 server days (one server running for 23 days, or 23 servers running for one day) to cover the space. And if we try the top 500: Total keys to check (500 x 500 x 500 x 500) =         62 500 000 000 And a server can check this many keys in a day:         43 200 000 000 Now a single server can cover the whole space in about one and a half days. That's actually (much) less time than to brute force a simple 6 character password. I'm not suggesting that everyone's four-word-wallet can be cracked wide open in a day, but it does mean that low hanging fruit - think simple, common words - will be quickly found. Aiwin+SHA256 brainwallets: the world's biggest encrypted password file...
Doing some quick back of the envelope calculations. Consider this a thought experiment rather than anything too accurate. My server with a 2010 era quad core CPU can check about 300,000 keys per second. It could probably be pushed further with some tweaking. Let's say (conservatively) that a more modern quad core CPU can do 500,000 and use that as the reference. That means it can check 43.2 billion keys per day. Brute forcing the "correct horse battery staple" space One dictionary that includes a rank of how commonly a word appears on the web ranks the least common word "staple" at 16904. So let's use that hint (some mild cheating) and set our limits to the 20000 most common words. Total keys to check (20000 x 20000 x 20000 x 20000) = 160 000 000 000 000 000 And a server can check this many keys in a day:          43 200 000 000 So in this instance, we would need approximately 10,000 servers running for a year to brute force every combination of those 20000 words. Not practical, but certainly not impossible. But what if we use only the most common 1000? Total keys to check (1000 x 1000 x 1000 x 1000) =      1 000 000 000 000 And a server can check this many keys in a day:         43 200 000 000 In this case, we only need about 23 server days (one server running for 23 days, or 23 servers running for one day) to cover the space. And if we try the top 500: Total keys to check (500 x 500 x 500 x 500) =         62 500 000 000 And a server can check this many keys in a day:         43 200 000 000 Now a single server can cover the whole space in about one and a half days. That's actually (much) less time than to brute force a simple 6 character password. I'm not suggesting that everyone's four-word-wallet can be cracked wide open in a day, but it does mean that low hanging fruit - think simple, common words - will be quickly found. Aiwin+SHA256 brainwallets: the world's biggest encrypted password file...

Doing some quick back of the envelope calculations. Consider this a thought experiment rather than anything too accurate. My server with a 2010 era quad core CPU can check about 300,000 keys per second. It could probably be pushed further with some tweaking. Let's say (conservatively) that a more modern quad core CPU can do 500,000 and use that as the reference. That means it can check 43.2 billion keys per day. Brute forcing the "correct horse battery staple" space One dictionary that includes a rank of how commonly a word appears on the web ranks the least common word "staple" at 16904. So let's use that hint (some mild cheating) and set our limits to the 20000 most common words. Total keys to check (20000 x 20000 x 20000 x 20000) = 160 000 000 000 000 000 And a server can check this many keys in a day:          43 200 000 000 So in this instance, we would need approximately 10,000 servers running for a year to brute force every combination of those 20000 words. Not practical, but certainly not impossible. But what if we use only the most common 1000? Total keys to check (1000 x 1000 x 1000 x 1000) =      1 000 000 000 000 And a server can check this many keys in a day:         43 200 000 000 In this case, we only need about 23 server days (one server running for 23 days, or 23 servers running for one day) to cover the space. And if we try the top 500: Total keys to check (500 x 500 x 500 x 500) =         62 500 000 000 And a server can check this many keys in a day:         43 200 000 000 Now a single server can cover the whole space in about one and a half days. That's actually (much) less time than to brute force a simple 6 character password. I'm not suggesting that everyone's four-word-wallet can be cracked wide open in a day, but it does mean that low hanging fruit - think simple, common words - will be quickly found. Aiwin+SHA256 brainwallets: the world's biggest encrypted password file...

Doing some quick back of the envelope calculations. Consider this a thought experiment rather than anything too accurate. My server with a 2010 era quad core CPU can check about 300,000 keys per second. It could probably be pushed further with some tweaking. Let's say (conservatively) that a more modern quad core CPU can do 500,000 and use that as the reference. That means it can check 43.2 billion keys per day. Brute forcing the "correct horse battery staple" space One dictionary that includes a rank of how commonly a word appears on the web ranks the least common word "staple" at 16904. So let's use that hint (some mild cheating) and set our limits to the 20000 most common words. Total keys to check (20000 x 20000 x 20000 x 20000) = 160 000 000 000 000 000 And a server can check this many keys in a day:          43 200 000 000 So in this instance, we would need approximately 10,000 servers running for a year to brute force every combination of those 20000 words. Not practical, but certainly not impossible. But what if we use only the most common 1000? Total keys to check (1000 x 1000 x 1000 x 1000) =      1 000 000 000 000 And a server can check this many keys in a day:         43 200 000 000 In this case, we only need about 23 server days (one server running for 23 days, or 23 servers running for one day) to cover the space. And if we try the top 500: Total keys to check (500 x 500 x 500 x 500) =         62 500 000 000 And a server can check this many keys in a day:         43 200 000 000 Now a single server can cover the whole space in about one and a half days. That's actually (much) less time than to brute force a simple 6 character password. I'm not suggesting that everyone's four-word-wallet can be cracked wide open in a day, but it does mean that low hanging fruit - think simple, common words - will be quickly found. Aiwin+SHA256 brainwallets: the world's biggest encrypted password file...