I don't think this is necessarily the case. You do need to understand the idea's ___domain really well, but that doesn't have to be because it's something you do already. You could try to solve the problem of someone you know well, or have a cofounder who knows the problem ___domain better than you, or pick a market that seems underserved and learn about it until you get an idea and can make sure it's not a bad one.
There are benefits to picking something you don't already do - or rather, something that most people who start startups don't do. People naturally think the most about optimizing things they do often, and when most of them are being told to "scratch your own itch", they do even more of that. So most software startups will be about things that people who write software and found startups do a lot of, and those markets will be the most competitive. If you solve your grandmother's problems instead of your own, there's a lot less competition.
There are benefits to picking something you don't already do - or rather, something that most people who start startups don't do. People naturally think the most about optimizing things they do often, and when most of them are being told to "scratch your own itch", they do even more of that. So most software startups will be about things that people who write software and found startups do a lot of, and those markets will be the most competitive. If you solve your grandmother's problems instead of your own, there's a lot less competition.