My son and I tried a couple of things.
Scratch and its "block coding" ilk are one path. The good bit is that writing simple game-like things is easy: the UI stuff is very accessible.
The bad bits are that the ceiling is very low: you can only do so much before you grow past the platform; that it doesn't do anything other than simple UI games (and some turtle-ish robotics) very well; that it's very hacky -- you don't learn any good habits in Scratch; and that ultimately, you're going to have to leave "block coding" behind, and make the switch to "line coding" (as they seem to call it).
We moved from there to PyGame, which expanded the possibilities a little, but not a lot. It enables some better habits, but doesn't really encourage them.
We tried GameMaker and its myriad variants. That opened up better game creation abilities, but it's still a long way from doing anything that's "wow" compared to what they see other people doing.
We tried Roblox Studio, which was possibly the best route in terms of results for effort invested. But it's a lot of work for a kid to deal with.
Minecraft Education Edition wasn't worthwhile. We actually had more success with a mod-making tool for Java Edition that did all the code generation for the game interface based on filling out some forms in its UI. We could add Java code to extend that, but the learning curve was pretty steep. Still, adding new blocks and simple mobs was fun for a while, and easier than Roblox or full-blown Unity or whatever.
My son basically lost interest in trying to make games when he realised how much work went into making anything that he was proud to show his friends. 3D modelling, physics, music, sound effects, level design, etc, etc, all on top of the basic coding. Making something that he felt looked ok and moved ok and did something ok took a lot of work, and was nowhere near a playable game. A 2D NES-style game was about the peak of what he could do: sprites, collisions, scoring, levels .. which, tbh, he could also do in Scratch, so it wasn't really worth the effort to achieve something that was cute, but wasn't going sustain his or his friends' interest.
These days he does a bunch of HTML and CSS, and little bits of JavaScript. His interest is mostly in making web sites and publishing mostly text and some basic 2D images, and he finds that easy enough to do.
He uses GitHub to keep history (and sometimes to revert to early versions), and I set up a script that pushes the site to the server so he doesn't have to remember the command-line magic.
He found Code Camps were worthwhile: not as useful as 1-on-1 time would be, but still, a few days hacking on something with a bunch of like-minded other kids and an instructor who can answer questions -- they gave him a lot of confidence to try stuff, and to master the material, even if he wasn't really stoked with the outcomes.