[ffn]

In my opinion, if you're just a hobbyist programmer who want to experiment building games on the side, it makes more sense to build your MVP demo using humble html5 canvas (I personally use pixijs, but things like threejs or whatnot might be more appropriate depending on your idea). This is because building a good indie game is a big commitment requiring lots of time (which is money), and unless you're building this game for your grandchildren, it makes sense to test the market with a MVP and try to raise money for it on a kick-starter site before you commit to building it on a powerful engine like Unreal (or Unity).

In addition, worrying about powerful graphics as an indie game dev is like worrying about distributed database scalability as a minor e-commerce startup. Your priorities are messed up and, just like how you can just get away with mysql on a free heroku instance as long as you have interesting junk to sell on your ecommerce startup, so too can you get away with a shoddy 2d engine with box2d physics as long as your story is engaging (or your mechanics are fun).

[greggman]

It really depends on what type of game you're planning to make. Go to any game jam and look at the amazing games prototyped in Unity over a weekend. Compare the html5 games. It's often night and day. The amount of stuff you can get done in an engine you're familiar with so many tools and plugins at your disposal is a huge force multiplier.

Of course do whatever you want. Most of my projects have been HTML5. But at the same time I recognize that free importing of nearly any format of 3d assets into unity and built in animation, collision, physics, ability to edit nearly all parameters at runtime, even code. Is a huge boost once you get used to it.