Boxed In (JezzBall clone)

Before anything else about how this plays: the design it's rebuilding is Microsoft's, from the Entertainment Pack era, and the name it goes by is a registered trademark still technically active even though the original hasn't been sold in decades. What's here is a clean-room, from-scratch rebuild under the author's own MIT licence — no Microsoft code, no Microsoft assets — so it's genuinely fine to host, but it isn't going out under its old branding, the same way this shelf has handled other direct-title clones before. Call it what it plays like: a wall-building, ball-trapping area-denial puzzle, built from nothing but canvas and vanilla JavaScript by a solo developer who's been maintaining it on and off since 2020.

What actually shipped needed fixing first

Worth being upfront about, because it's the kind of gap a quick glance at the repo wouldn't catch: the committed index.html pointed its script tag at a webpack-built bundle.js that simply isn't in the repository, and separately referenced a stylesheet and two images from the author's own personal site — ../back_button.css, ../img/favicon.ico, ../img/back_arrow.png — that only exist in the context of that site's own folder structure, not this one. The fix turned out to be exactly as small as advertised: the page already had a plain ES-module entry point (main.js, importing nothing but its own sibling files) sitting right there, just commented out in favour of the missing bundle. Switching to it, and dropping the two dead asset references along with the "back to the list of fun stuff" navigation link that only made sense on the author's own site, was enough to get a fully working standalone copy with zero build step. None of that reflects on the actual game code, which was clean throughout, but a candidate that arrives unable to load without source changes is worth naming plainly rather than glossing over.

The other real gap: the game board was hard-coded to 720×640 pixels with no responsive behaviour at all. The author had clearly thought about this — there's a commented-out "miniature mode" block sized at 360×320 sitting right next to the bundle toggle — but switching it on as written would have shrunk the board on every device, desktop included, rather than just narrow ones. Scoping that same 360×320 size to a phone-width media query instead keeps the full-size board on desktop and gives a phone a board that fits its screen without forcing horizontal scrolling.

The touch controls are a real implementation, not an afterthought

This is the part that actually earns the game a place here. ui.js binds touchstart, touchmove, and touchend directly on the game's own canvas layer, and touchmove drives the wall-placement paddle continuously as a finger drags across the board — the exact same logic path mousemove uses, just fed touch coordinates instead of mouse coordinates. Lifting the finger, via touchend, is what commits the wall, mirroring what a mousedown does for a click. Testing this directly on a simulated 390×844 touch profile, a single tap-and-release on an open section of the board placed a wall that then grew across the screen on its own and sealed off roughly a fifth of the level, with the cleared-percentage counter updating to match. The same sequence worked identically with a mouse click on the desktop profile.

The one action a touchscreen genuinely can't do is a desktop right-click, which is how this game normally swaps which way a new wall extends. Rather than hide that behind a long-press or drop it, index.html puts a permanent, clearly-labelled "Swap Direction By Tapping This Box" button on screen at all times, wired to the same swapDirection function the right-click calls. It's a plain, unglamorous solution and exactly the right one — a dedicated control rather than a compromise.

A good design, plainly dressed

Once it's running, the actual game is worth the fixing effort. Two or more balls bounce around an otherwise empty rectangle, and the only tool available is drawing a wall — horizontally or vertically, direction chosen by whichever way you last swapped it — that grows from a fixed point toward both edges of the board and locks in whichever side ends up empty of balls when it completes. A wall struck by a ball before it finishes breaks and costs a life. That's the entire design, and it holds up: because the walls are visible and building in real time, the tension is in the decision of where to start one, not in reflexes, and the board gets meaningfully harder as it fills in and the remaining bounce paths for the balls narrow. It's a genuinely well-built puzzle idea that most players under a certain age have simply never had the chance to play, because the original hasn't been sold or bundled with anything in a very long time.

What it doesn't have is any visual ambition to match the design: solid-fill circles for balls, a plain black bar for walls, and one flat magenta background colour throughout, with no sound at all. That's a legitimate, honest trade a solo hobby project is allowed to make, and it doesn't get in the way of playing, but it does mean this earns its place on craft and control rather than presentation.

Where that leaves it

A 7.5 reflects a genuinely good, well-loved design rebuilt faithfully and with real attention to making it work on a touchscreen — exactly the profile this shelf is built to surface — held back by presentation that doesn't try for more than function, and by the fact that nothing here is a new idea, only a well-executed old one made accessible again.