This wikipedia section is a good place to start seeing how it works:
https://en.wikipedia.org/wiki/Patent_claim#Basic_types_and_c...What I remember from law school is: let's say you're drafting a patent, and you want it to cover as much as possible without getting invalidated by prior art. So you're going to write some claims, and each claim can stand or fall on its own (each claim could be rejected by the patent office, or invalidated later because of prior art, without invalidating the more specific claims). So first you'll write a series of independent claims, and those will each be as broad as you hope you can get away with ("it's HTML on a computer!"). Having the broad independent claims stand is your best case scenario, because that makes it hard for competitors to work around your patent. But then for each independent claim you'll add dependent claims, which narrow down what you're claiming ("OK, it's HTML on a computer with foo," "OK, it's HTML on a computer with foo and bar"). That's your fallback scenario: maybe "HTML on a computer" gets rejected, and "HTML on a computer with foo" gets rejected, but "HTML on a computer with foo and bar" is original enough to make it through, so you still end up owning something of value. But maybe now that your broad claims are gone, your competitors can avoid your patent by doing "HTML with foo and baz" to accomplish the same thing.
So that's how the game theory works: you're basically leaving money on the table if you don't include some super broad claims at the top. Doesn't cost you anything if those ones don't stick. So we can't just read a patent and assume it actually protects every single broad claim -- but we can't assume it doesn't, either. We won't find out for sure which (if any) claims are actually valid unless there's a lawsuit and a judge or jury has to pick through the prior art and decide whether there are claims in there that are narrow enough to be original.