Don't use any portion of any written work as your private key! Brute-force attacks will use every bit of written literature to do this same search.

You must use securely-generated random key phrases, or you will lose your coins.

No key is safe from a brute force attack that covers the alphabet of key and the length of the key.

What you're describing is a dictionary attack and it trades comprehensiveness for speed.

There's many more words than characters in any language. Using words as your alphabet gets you ($WORDS)^N possible passwords of length N instead of 128^N for an ascii alphabet. For humans remembering random words is easier.

Choosing a grammatically correct sentence instead of random words is little different in terms of complexity reduction than choosing a word based password (e.g. Password1) instead of random characters.

Obviously if the attacker can build a target specific dictionary they'll have a higher probability of success.

Indeed. My caution is against keys that are already in the clear in a written work.

Bitcoin is cryptography's biggest bug bounty.

I said password for the private key, not the private key itself. I store my PKs in KeePass databases with a crazy number of rounds, so each password guess attempt takes 2-3 seconds on a modern CPU. Good luck bruteforcing that.