Angband

There is a specific kind of software that stops being a product and becomes infrastructure. Vim is one. Emacs is one. Angband is one. It has been in continuous development for decades — a direct descendant of Moria, itself a descendant of Rogue — which means this codebase predates the World Wide Web as a consumer technology. It has outlived Sierra, Origin, 3DO, and multiple complete cycles of what the industry considered a “modern” game engine. It is still being patched. It is still free. It is still, in the most literal sense, unfinished, because a vast procedurally generated dungeon is never really finished.

The Android build in question is not maintained by the upstream Angband team. It’s the work of Cuboideb Development, a small independent porting effort that has done something more ambitious than a straight recompile. Rather than shipping vanilla Angband alone, this package bundles the base game alongside four major variants: FAangband 2, Sil-Q, FrogComposband, and NPPAngband. That’s not a bonus feature. That’s four genuinely distinct games with different design philosophies — Sil-Q in particular is a tight, elegant, Silmarillion-era reimagining that plays almost nothing like its parent — packaged into a single free app.

The market positioning is essentially nonexistent, and that’s the point. There is no iOS version, and there won’t be: the Angband community has cited GPL incompatibility with App Store distribution terms and understandable caution around Tolkien’s IP as the blockers. There’s no Steam page, no Switch port, no publisher. The official channels are rephial.org, GitHub, and browser play at angband.live. This app exists because someone wanted it to exist on a phone, and did the work. In a store dominated by energy timers and battle passes, an app whose entire business model is there is no business model reads as almost confrontational.

• Five games in one package: Vanilla Angband plus FAangband 2, Sil-Q, FrogComposband, and NPPAngband. Each has its own dungeon logic, class balance, and tone. FrogComposband alone is a maximalist sprawl of races and classes that could justify the download by itself.
• Genuine permadeath: Your character dies once. That’s it. No cloud rollback, no continue token, no ad-supported revive. Hours of accumulated gear and knowledge evaporate because you misread a monster’s speed rating.
• Unidentified-by-default itemization: Every potion, scroll, wand, and ring is an unknown quantity until it’s used. Drinking an unlabeled potion at low health is a real decision with real consequences, and this single mechanic generates more tension than most horror games manage with a full audio budget.
• A customizable virtual keyboard: The port’s headline engineering achievement. The on-screen command layout can be remapped to fit individual hands, which is the only reason a keyboard-native game of this vintage is playable on glass at all.
• ASCII or tilesets: Play it as raw glyphs on black, or switch to graphical tiles. Both are supported; neither is a compromise the other side has to accept.

The Gameplay Loop

Angband’s loop is easy to describe and structurally brutal. You roll a character — pick from humans, dwarves, high-elves, hobbits, half-trolls, dúnedain and more, then a class from warrior through mage, priest, rogue, ranger, paladin, blackguard, and necromancer. You shop in town. You take the stairs down. You come back up, or you don’t.

What separates this from the endless flood of “roguelites” that borrowed the word and dropped the concept is that Angband has no meta-progression. There is no permanent currency, no unlock tree, no character that gets a little stronger every run. The only thing that persists between deaths is what’s in the player’s head. That is an enormously unfashionable design decision today, and it is also the reason the game has survived for decades. The progression is the player. Learning that a particular unique monster breathes in proportion to its remaining health, that phase door is a positioning tool and not just an escape hatch, that stair-scumming is both a legitimate survival strategy and a tedium trap — these are the upgrades.

Resource management carries the entire mid-game. Your scroll of teleportation is not a consumable, it’s an insurance policy, and every turn spent deciding whether to burn it is the game working exactly as designed. Combat is turn-based on a grid, which sounds passive and is anything but; the turn system is where the strategy lives, because monster speed relative to yours determines whether a retreat is a retreat or a slow-motion execution.

Where It Breaks Down

Reverence for a classic is not the same as an honest assessment. Angband has real, structural problems, and age doesn’t excuse them.

The difficulty curve isn’t a curve, it’s a wall with occasional handholds. New players die repeatedly to things they had no reasonable way to anticipate, and the game offers almost nothing in the way of diegetic teaching. The knowledge required is externalized — it lives in wikis, forum posts, and spoiler files, not in the game. That’s a design failure by any modern standard, even if the community has decided to call it a feature.

Then there’s the mid-game grind. Descending through the middle third of the dungeon involves a great deal of generating levels, scanning them, and taking the stairs again — stair-scumming, in the parlance. It’s an emergent optimal strategy that the game has never fully designed around, and it turns long stretches of play into a slot machine pulled while waiting for a level worth exploring. Sil-Q, notably, solves this. Vanilla still hasn’t.

The Interface Problem

This is the part that will decide whether the app stays installed. Angband was designed for a full keyboard, where every command is a single keystroke and muscle memory does the work. On a phone, that model has to be rebuilt from scratch.

Cuboideb’s answer — a fully customizable on-screen keyboard plus touch controls — is the right architectural call. But “customizable” means the burden of designing a usable control scheme lands on the player before they’ve played enough to know which commands they actually use. That’s onboarding friction of the worst kind: you have to be experienced to configure it, and you have to configure it to become experienced. Community feedback bears this out; the most consistent complaint isn’t difficulty, it’s relearning key mappings.

Once that investment is made, though, the design pays off. A turn-based grid game with no reflex requirement is a natural fit for a phone. There’s no input latency penalty, no twitch skill lost to a touchscreen. Play can pause mid-dungeon and resume in the coffee line. Angband may be among the best-suited classic PC games for mobile, buried under one of the worst-suited control schemes.

The aesthetic is ASCII, and calling that “retro” undersells it. Raw glyphs are an information-density choice, not a nostalgic one — a p is a person, a D is a dragon, and once fluent, a full dungeon floor parses faster than any tileset allows. The optional graphical tilesets are competently done and helpful for newcomers, but veterans switch back for a reason. It’s the same argument as a text-mode terminal versus an IDE, and both sides are right.

As a 2D, turn-based game of this scope, Angband is light on hardware by design — instant turn resolution and fast level generation are inherent to the engine, and battery demands should be minimal. But the port isn’t clean. User reports include occasional crashes and, more seriously, save-file corruption. In a game with permadeath, a corrupted save isn’t an inconvenience. It’s an unearned death, which is the one thing Angband’s entire design contract promises you’ll never suffer. That’s the most serious knock against this build, and it’s not a small one. Back up the save directory. Seriously.

The store presentation is threadbare: a handful of screenshots, a functional icon, no marketing polish. Fitting, if not exactly inviting.

Android is the only mobile platform where this game legally and practically can exist, and the port leans into what the OS allows. Direct APK-adjacent freedom means the bundle can ship five GPL-licensed variants together without a store gatekeeper objecting — the exact thing that keeps it off iOS.

Touch handling is the whole ballgame. Tap-to-move works for traversal; the virtual keyboard handles everything else. Landscape on a large-screen phone or a tablet is dramatically better than portrait on a small one, because ASCII dungeon maps want horizontal real estate and shrinking glyphs to fit is how a d gets misread for a D and a character dies. A tablet is the right screen for this. A Bluetooth keyboard should transform it — at that point it’s the desktop experience on a device already carried, which is the strongest way to play.

Angband is not a polished app. It’s a great game delivered through a rough app, and those are different claims that reviews too often blur together. The port earns real credit for the customizable keyboard and for bundling four variants alongside the original — that’s generosity, not feature-padding. It loses real credit for stability problems that strike at the exact thing this genre cannot afford to get wrong.

But the thing being ported is close to irreplaceable. Angband offers a kind of consequence that the modern games industry has spent years engineering out of existence, because consequence is bad for retention metrics. Here, a bad decision deep in the dungeon costs you everything, and the game does not care, and that indifference is precisely what makes the victories mean something. Decades of volunteer maintenance have produced something no publisher would greenlight today.

We rate it a solid recommendation for the right audience only: roguelike veterans and patient strategists will find something singular here; everyone else should know exactly what they’re signing up for. Download it. Back up the saves. Expect to die.